Mechanism of Bacterial Oligosaccharyltransferase

Background: N-Linked glycosylation is catalyzed by oligosaccharyltransferase (OST). Results: Specific amino acids in enzyme and acceptor substrate are identified as key determinants for substrate binding and turnover. Conclusion: Quantification of substrate binding and turnover reveal a delicate interplay between acceptor substrate, enzyme, and metal ion. Significance: The study represents the first quantitative analysis of substrate binding and turnover in N-linked glycosylation. N-Linked glycosylation is an essential post-translational protein modification in the eukaryotic cell. The initial transfer of an oligosaccharide from a lipid carrier onto asparagine residues within a consensus sequon is catalyzed by oligosaccharyltransferase (OST). The first X-ray structure of a complete bacterial OST enzyme, Campylobacter lari PglB, was recently determined. To understand the mechanism of PglB, we have quantified sequon binding and glycosylation turnover in vitro using purified enzyme and fluorescently labeled, synthetic peptide substrates. Using fluorescence anisotropy, we determined a dissociation constant of 1.0 μm and a strict requirement for divalent metal ions for consensus (DQNAT) sequon binding. Using in-gel fluorescence detection, we quantified exceedingly low glycosylation rates that remained undetected using in vivo assays. We found that an alanine in the −2 sequon position, converting the bacterial sequon to a eukaryotic one, resulted in strongly lowered sequon binding, with in vitro turnover reduced 50,000-fold. A threonine is preferred over serine in the +2 sequon position, reflected by a 4-fold higher affinity and a 1.2-fold higher glycosylation rate. The interaction of the +2 sequon position with PglB is modulated by isoleucine 572. Our study demonstrates an intricate interplay of peptide and metal binding as the first step of protein N-glycosylation.

N-Linked glycosylation is an essential post-translational protein modification in the eukaryotic cell. The initial transfer of an oligosaccharide from a lipid carrier onto asparagine residues within a consensus sequon is catalyzed by oligosaccharyltransferase (OST). The first X-ray structure of a complete bacterial OST enzyme, Campylobacter lari PglB, was recently determined. To understand the mechanism of PglB, we have quantified sequon binding and glycosylation turnover in vitro using purified enzyme and fluorescently labeled, synthetic peptide substrates. Using fluorescence anisotropy, we determined a dissociation constant of 1.0 M and a strict requirement for divalent metal ions for consensus (DQNAT) sequon binding. Using in-gel fluorescence detection, we quantified exceedingly low glycosylation rates that remained undetected using in vivo assays. We found that an alanine in the ؊2 sequon position, converting the bacterial sequon to a eukaryotic one, resulted in strongly lowered sequon binding, with in vitro turnover reduced 50,000-fold. A threonine is preferred over serine in the ؉2 sequon position, reflected by a 4-fold higher affinity and a 1.2fold higher glycosylation rate. The interaction of the ؉2 sequon position with PglB is modulated by isoleucine 572. Our study demonstrates an intricate interplay of peptide and metal binding as the first step of protein N-glycosylation.
N-Linked glycosylation is a post-translational protein modification that takes place in the secretory pathway of eukaryotic cells. N-Glycans have a central role in diverse biological processes such as protein folding, intracellular trafficking, regula-tion of protein turnover, or cell-cell recognition (1). Protein N-glycosylation is common in Archaea (2)(3)(4) and is also found in defined taxa of Bacteria (5)(6)(7). The molecular mechanism of N-linked protein glycosylation is similar in all domains of life. An oligosaccharide is assembled on a lipid carrier, resulting in a lipid-linked oligosaccharide (LLO), 3 and transferred en bloc to the amide nitrogen of an acceptor asparagine, resulting in an N-glycosidic linkage. The process is catalyzed by oligosaccharyltransferase (OST), an integral membrane enzyme that modifies asparagines that are present in a consensus sequence N-X-(S/T), where X must not be proline (8). The N-X-(S/T) motif is found in eukaryotes, Archaea, and, in extended form, in Bacteria (8 -12).
Although the presence of an N-X-(S/T) sequon is essential, it is not sufficient to define a substrate for N-linked glycosylation. In fact, only two-thirds of all existing sequons in potential substrate proteins are modified with a glycan (13). The underlying principles that govern N-glycosylation site occupancy are poorly understood. A statistical analysis of eukaryotic glycoproteins with an entry in the Protein Data Bank revealed a surprisingly large number of glycosylated asparagines of low accessibility (13) suggesting that glycosylation occurs before protein folding (14). Therefore the fate of an individual sequon might depend on its structural environment and the competition between glycosylation and protein folding (15). The OST enzyme of higher eukaryotes is a multiprotein complex, with the catalytic Stt3 subunit surrounded by seven other subunits, some of which have chaperone or oxidoreductase activities, suggesting that OST directly modulates glycoprotein folding (16 -20). However, another factor governing site occupancy is the primary sequence of the glycosylation site. Statistical analyses have shown that sequons with a Thr in the ϩ2 position are more frequently glycosylated than those containing Ser (10,13,21). This is consistent with biochemical studies revealing that N-X-T sites in acceptor peptides and proteins are better substrates than N-X-S sites (22)(23)(24)(25). On the other hand, a comprehensive analysis of the mouse glycoproteome revealed the existence of non-canonical sequons, among which N-X-C, N-X-V, and N-G-X sites appeared with statistical significance (21). In bacteria, a single-subunit OST termed PglB, which is homologous to the catalytic Stt3 subunit of eukaryotic OST, catalyzes the transfer of glycans onto acceptor proteins.
The X-ray structure of the PglB protein from Campylobacter lari has helped identify the structural basis of sequon recognition (26). It was shown that the ϩ2 Thr/Ser of the sequon is optimally positioned to allow specific hydrogen bonds to be formed with the WWD motif of PglB, which is strictly conserved in Stt3 homologues (18). The structure suggested van der Waals interactions as a possible explanation for the preference of Thr over Ser at the ϩ2 position of the sequon. It also revealed a possible salt bridge between an arginine side chain on the surface of the enzyme and the negatively charged aspartate in the Ϫ2 position of the sequon, which may explain the observation that bacterial glycosylation sequons are extended ((D/ E)-Z-N-X-(S/T), with X, Z P (27)) compared with their eukaryotic or archaeal counterparts. Nevertheless, glycosylation of non-canonical sites lacking either the interaction in the Ϫ2 position or the ϩ2 position have been reported for the PglB enzyme in vivo (28,29).
To understand the catalytic mechanism of OST, quantitative in vitro assays and detailed structure-function analyses are essential. We exploited the stability and activity of detergentsolubilized C. lari PglB to determine sequon binding affinities and glycosylation rates. We chemically synthesized fluorescently labeled peptide substrates and purified milligram amounts of PglB enzyme, where we introduced mutations guided by the crystal structure. After developing highly sensitive assays, we quantified sequon binding using fluorescence anisotropy and determined glycosylation turnover rates even for highly disfavored mutant/peptide combinations. We were thus able to gain insight into (i) the influence of active site residues and divalent metal ions for sequon binding, (ii) the specificity of the Ser/Thr binding pocket, and (iii) bacteria-specific requirements of sequon recognition. Our quantitative assessment of peptide binding and catalysis provides insight into the natural selection of the N-X-(S/T) sequon in eukaryotes and the extended sequon in Bacteria.
Construction of Plasmids-Mutations in the pglB gene of C. lari were generated by the QuikChange method on plasmid pSF2 (26). For the in vivo glycosylation assay pglB variants were subcloned into a pMLBAD plasmid (30). For in vitro glycosylation studies the two endogenous glycosylation sites were removed (N535Q and N556Q) to prevent autoglycosylation of PglB. The resulting construct was referred to as wild type construct and all subsequent mutations were based on this construct.
Mutation of the DQNAT sequon within the 3D5 acceptor protein was performed by ligation of phosphorylated, doublestranded DNA of oligonucleotides 5Ј-CTAGCGGTGG-TGGTGGTTCTGGTGGTGGTGCCCAGAACGCCA-3Ј and 5Ј-CCGGTGGCGTTCTGGGCACCACCACCAGAACCAC-CACCACCG-3Ј into the plasmid pCL21 (31) digested with restriction enzymes NheI and AgeI. This resulted in plasmid pCL64 carrying an AQNAT acceptor site. All plasmids were validated by DNA sequencing.
In Vivo Glycosylation Assay-In vivo complementation analysis of PglB mutants was performed as described before (26). Briefly, Escherichia coli SCM6 cells were transformed with three separate plasmids carrying: 1) the C. jejuni pglB mut cluster (containing an inactivated pglB gene) to generate LLO; 2) the glycosylation acceptor protein 3D5 containing a DQNAT or an AQNAT site; 3) C. lari PglB, wild type or mutants. Expression and glycosylation of 3D5 was monitored by SDS-PAGE of periplasmic cell extracts and visualized by mobility shift due to increased size in an immunoblot using anti-c-Myc antibody, or the reactivity of the glycoprotein in an anti-glycan immunoblot using hR6 antiserum.
Expression and Purification of PglB Mutants-Overexpression and purification of PglB mutants was performed as previously described (26). Shortly, proteins were overexpressed in E. coli BL21(DE3) Gold cells in a 30-liter fermenter or in 5-liter baffled flasks using Terrific Broth medium supplemented with 1% (w/v) glycerol. Cells were grown to A 600 of 10.0 (fermenter) or 3.0 (flasks) at 37°C and induced with 0.1% (w/v) arabinose for 2 or 4 h, respectively. Cells were harvested by centrifugation and pellets were stored at Ϫ80°C before membrane preparation. All subsequent steps were carried out at 4°C. PglB was solubilized in 25 mM Tris-HCl, pH 8.0, 250 mM NaCl, 10% (v/v) glycerol, and 1% (w/v) N-dodecyl-␤-D-maltopyranoside (Anatrace) for 1 h. All purification buffers contained N-dodecyl-␤-D-maltopyranoside. PglB was purified on a nickel-nitrilotriacetic acid Superflow affinity column (Qiagen) and desalted into 10 mM MES, pH 6.5, 100 mM NaCl, 0.5 mM EDTA, and 3% (v/v) glycerol. Desalted protein was analyzed by size exclusion chromatography (Superdex 200, GE Healthcare) and, if needed, concentrated up to 26 mg/ml in a 100-kDa molecular mass cutoff Ultra-15 concentrator (Amicon, Millipore). The concentrated sample was re-analyzed by gel filtration and protein concentrations were determined by absorption at 280 nm of diluted samples using spectrophotometry.
Synthesis of Acceptor Peptides Labeled with 5-Carboxyfluorescein-Peptide synthesis was initiated by loading Tenta-Gel S RAM resin (500 mg, loading: 0.24 mmol g Ϫ1 ) in a 10-ml polypropylene syringe fitted with a polypropylene frit, a Teflon stopcock, and a stopper. The resin was swollen in DCM (6 ml, 20 min). After removal of the DCM, the Fmoc protecting group of the resin was removed by using a solution of 20% piperidine in NMP. Stirring of the reaction mixture at any given step was performed by attaching the closed syringes to a rotating axis. The completion of the reaction was checked using the 2,4,6trinitrobenzenesulphonic acid test. Removal of the Fmoc pro-tecting group was performed by using a solution of 20% piperidine in NMP (6 ml, 2 ϫ 10 min). After filtration, the resin was washed with NMP (2 ϫ 6 ml), MeOH (2 ϫ 6 ml), and DCM (2 ϫ 6 ml). Coupling of amino acids was performed using Fmocprotected amino acids (3 eq), Benzotriazol-1-yl-oxytripyrrolidinophosphonium hexafluorophosphate (3 eq), and N,N-Diisopropylethylamine (6 eq) in NMP (6 ml). The resin was stirred for 2 h before it was washed with NMP, MeOH, and DCM (2 ϫ 6 ml each). Acetylation of the resin was performed after each amino acids coupling using a solution of acetic anhydride:DCM (1:1). Coupling of 5-carboxyfluorescein was performed by using 5-carboxyfluorescein (2 eq), HOBt (5 eq), and N,NЈ-Diisopropylcarbodiimide (5 eq) in NMP (6 ml). The resin was stirred overnight and protected from light using aluminum foil. The resin was washed with NMP, MeOH, and DCM (2 ϫ 6 ml each). A solution of 20% piperidine in NMP (6 ml, 6 ϫ 5 min) was added to the resin to remove the excess of free 5-carboxyfluorescein before the resin was finally washed with DCM (6 ϫ 6 ml). After coupling of the fluorophore, the compounds were protected from light with aluminum foil at all times. and plotted against total protein concentration. Data were fitted in Prism (GraphPad Software) using the following equation for the 1:1 stoichiometry, for the mutant R375A, where a 2:1 stoichiometry yielded a higher R 2 value. The protein solution from the cuvette was analyzed again by size exclusion chromatography after each anisotropy measurement to prove stability of the protein over the course of the experiment. Extraction of LLOs-Isolation of LLOs was performed as described before (32). Briefly, LLOs were extracted from E. coli SCM6 cells carrying the Campylobacter jejuni pglB mut cluster (containing an inactivated pglB gene) by a mixture of chloroform:MeOH:H 2 O, 10:20:3. Extracts were dried in a rotavap and reconstituted in a buffer containing 10 mM MES, pH 6.5, 100 mM NaCl, and 1% Triton X-100 (w/v). The concentration of reconstituted LLOs was determined by titrating various amounts of LLO against a constant amount of acceptor peptide in an in vitro glycosylation assay.
In Vitro Glycosylation Assay-In a reaction containing 10 mM MES, pH 6.5, 100 mM NaCl, 10 mM MnCl 2 , 3% glycerol (v/v), and 1% Triton X-100 (w/v), 10% LLO extract (v/v, corresponding to 35 M) and various amounts of PglB (1 nM to 25 M) were mixed. The mixture was preincubated at 30°C in a water bath for 5 min before the reaction was started by addition of fluorescently labeled acceptor peptide. Reactions were incubated in a water bath at 30°C. Samples were taken at different time points and reactions were stopped by the addition of 4ϫ SDS sample buffer. Samples were diluted 200-fold prior to analysis by Tricine SDS-PAGE in minigels (8 ϫ 8 cm) consisting of a 16% resolving gel with 6 M urea, a 10% spacer gel, and a 4% stacking gel (33). Fluorescent bands for peptide and glycopeptide were visualized by using a Typhoon Trio Plus imager (GE Healthcare) with excitation at 488 nm and a 526-nm SP emission filter. The amount of formed glycopeptide was determined from band intensities of fluorescence gel scans (ImageJ) where the sum of signals for glycosylated and non-glycosylated peptide for each lane was defined as 100%.
For Michaelis-Menten kinetics, 1 nM PglB was incubated with various amounts of acceptor peptide (0.25-30 M) for 10 min. Data were fitted in Prism (GraphPad Software) by nonlinear regression according to the Michaelis-Menten formula.
For turnover rate determination, various amounts of PglB enzyme were incubated with 10 M acceptor peptide. A total of 10 samples were taken in 3-min intervals and data were fitted by linear regression.
For M 2ϩ titration experiments, the nickel-nitrilotriacetic acid eluate of purified PglB was supplemented with 0.5 mM EDTA, but no EDTA was added to the desalt buffer. 4 nM PglB enzyme was incubated with 10 M acceptor peptide and various amounts of MnCl 2 and MgCl 2 , respectively. Four samples were taken at the appropriate time intervals (3 to 25 min) so that the reaction was in the linear range. Data were fitted by linear regression to determine turnover rates. These rates were plot-ted against the M 2ϩ concentration and data were fitted with a model for one metal binding site using the following equation. Fig. 1 illustrates a hypothetical reaction mechanism based on previous biochemical studies and on the X-ray structure of C. lari PglB. The physical separation of acceptor peptide binding and catalytic sites suggested that sequon binding could be studied as a discrete, initial step of the reaction. We synthetically attached a fluorophore and a Gly-Ser linker to the N terminus, and a phenylalanine to the C terminus, of a consensus acceptor sequon DQNAT (the underlined N representing the glycosylated asparagine and the zero position of the sequon). The resulting peptide, 5-carboxyfluorescein-GSDQNATF-NH 2 (the underlined region representing the sequon, Fig. 2A) was used both for binding and glycosylation studies. C. lari PglB was overexpressed and purified in detergent (n-dodecyl-␤-D-maltoside) as described before (26). The purity of all PglB variants was analyzed by SDS-PAGE, and the monodispersity of the concentrated (Ͼ15 mg/ml) protein preparations was ascertained using analytical size exclusion chromatography (supplemental Fig.  S1).

Novel in Vitro Assays Allow Accurate Determination of Sequon Binding Affinity and Glycosylation Turnover-
Peptide substrate binding was quantified by fluorescence anisotropy and yielded a dissociation constant, K d , of 1.02 Ϯ 0.06 M for wild type PglB and the DQNAT consensus sequon (Fig. 2B), determined in the presence of 10 mM Mn 2ϩ (see also below). Control experiments using an unrelated membrane protein in a similar detergent solution (Fig. 4A) indicated that binding of the fluorescently labeled peptide to PglB was specific.
To determine turnover rates, PglB was prepared in the same conditions as for peptide binding analysis, and LLO was added as a glycan donor. The glycosylated product was separated from the unmodified substrate by gel electrophoresis, and the educt/ product ratio was determined by measuring the in-gel fluorescence intensities of both educt and product bands. Michaelis-Menten kinetics could thus be determined for wild type PglB using increasing concentrations of peptide substrate (Fig. 2C). We determined K m and V max values of 2.60 Ϯ 0.28 M and 2.20 Ϯ 0.07 peptide/s, respectively (Fig. 2D). The turnover rate for wild type PglB and the DQNAT sequon was determined for the initial, linear range of the reaction and resulted in 1.5 Ϯ 0.04 peptide/s (Fig. 2, E and F). Because our assay system limited the use of substrate peptide to a concentration of 30 M and many mutations introduced in PglB or the acceptor peptide reduced the enzymatic performance drastically, we decided to determine and compare initial turnover rates instead of Michaelis-Menten kinetics, because impractically high peptide concentrations would be required to reach V max . After optimizing the system using wild type PglB and the consensus sequon, we concluded that the two methods allowed us to perform quantitative comparisons of sequon binding and turnover rates using any combination of PglB mutant or sequon variant.
The Role of Active Site Residues for Sequon Binding-The PglB structure revealed a cavity that contains the active site of the enzyme and is believed to bind LLO (Fig. 3A). Based on a tentative modeling of LLO into the crystal structure it was hypothesized that Arg-375 might interact with the pyrophos- phate moiety of the LLO substrate (Fig. 3A). To test this hypothesis, we mutated Arg-375 to alanine and lysine and analyzed the in vivo activity of PglB using glycosylation-competent E. coli cells (18) expressing an scFv fragment as acceptor protein (31). We found that although the activity of the R375K mutant was only slightly reduced, the R375A mutant had no detectable activity (Fig. 3B). We then analyzed these mutants in vitro. Anisotropy measurements using the DQNAT sequon yielded dissociation constants of 0.35 Ϯ 0.02 M for R375K and 2.10 Ϯ 0.16 M for R375A ( Fig. 3C and Table 1). A slightly improved data fit was obtained for the R375A mutant using two distinct binding sites, yielding K d1 of 4.58 Ϯ 0.68 M and K d2 of 0.02 Ϯ 0.036 M (Table 1). We concluded that the positive charge of Arg-375 was not required for peptide binding, but was essential for later steps in catalysis. Activity measurements at high enzyme concentrations revealed that the glycosylation rate of the R375A mutant was reduced 45,000-fold, and the R375K mutant 40-fold relative to WT PglB (Table 1 and supplemental Fig. S2). These results were consistent with the in vivo activity measurements (Fig. 3B). In addition, the data demonstrated that extremely low turnover rates could be determined in vitro even when no in vivo activity could be observed.
We next investigated the catalytically essential residues Asn-56 and Glu-319. The structure suggested that the side chains of these residues are in contact with the acceptor asparagine and simultaneously contribute to metal binding (26). Individual mutations of Asp-56 and Glu-319 to alanine strongly reduced C. lari PglB activity in vivo (26), and mutations of the equivalent residues in C. jejuni PglB yielded similarly reduced in vitro glycosylation activities (34). We performed peptide binding experiments with PglB D56A and E319A mutants and found substantially lowered affinities in both cases (Fig. 3C). Dissociation constants could not be accurately determined because at the maximum enzyme concentration (110 -120 M), the anisotropy signal reached only about 30% of the signal expected for saturation. Analysis of the corresponding glycosylation rates showed a 900-fold reduction for mutant E319A (Table 1 and supplemental Fig. S2). For the mutant D56A, we observed differences in activity in distinct batches of purified protein. Compared with wild type, the activity of D56A was reduced between 11,000-and 140,000-fold, with an average reduction of 42,000-fold (Table 1 and supplemental Fig. S2). Irrespective of this fluctuation, mutant D56A had a much more pronounced effect on turnover than mutant E319A, which is in agreement with the observed in vivo activities. We conclude that the side chains of Asp-56 and Glu-319 have a role both in peptide binding and turnover.
Dual Role of the Divalent Metal Ion in Sequon Binding and Catalysis-Divalent metal ions are essential cofactors of most GT-A glycosyltransferases (35) and several studies have described the requirement of Mn 2ϩ or Mg 2ϩ for the activity of OSTs from various organisms (14, 23, 36 -38). To understand the role of the metal ion for PglB function (Fig. 3A), we measured peptide binding in the presence of EDTA. The results revealed very weak binding for the DQNAT sequon, too low for accurate K d determination (Fig. 4A). We subsequently performed a series of fluorescence anisotropy measurements at various Mn 2ϩ and Mg 2ϩ concentrations (Fig. 4B, supplemental Fig. S3, and

TABLE 1 Glycosylation turnover rates and dissociation constants of PglB mutants D56A, E319A, R375A, and R375K
Turnover rates and dissociation constants were derived from curves shown in supplemental Figs. S2 and S4. The errors represent the S.E. of the fit for each turnover rate and dissociation constant. Values for relative turnover and relative binding refer to the ratio of an individual mutant and wt PglB.

Mutant
Turnover supplemental Table S2). The data revealed a concentration-dependent increase of acceptor sequon affinity, with the highest affinity observed in the presence of 10 mM Mn 2ϩ and 50 mM Mg 2ϩ , respectively. At higher concentrations of the metal ions, the observed binding affinity was reduced.
To quantify the effect of increasing Mn 2ϩ and Mg 2ϩ concentrations on catalysis, glycosylation rates were determined (Fig.  4D). The resulting turnover rates increased and decreased analogously to the observed sequon affinities, but with a slight shift of the maxima (Fig. 4, B and D). Whereas the strongest peptide binding was observed at 10 mM Mn 2ϩ and 50 mM Mg 2ϩ , the highest turnover was observed at ϳ50 mM Mn 2ϩ and between 15 and 50 mM Mg 2ϩ . Our data allowed us to determine the apparent metal ion binding affinities (Fig. 4C). We used data points at concentrations of up to 50 mM for both metal ions because of the reduced activity at higher concentrations, as described above (Fig. 4, B and D). The data revealed an apparent K d for Mg 2ϩ binding of 2.32 Ϯ 0.51 mM, which was ϳ3-fold lower than that of Mn 2ϩ (apparent K d 7.88 Ϯ 1.76 mM). However, the maximum turnover rate in the presence of Mn 2ϩ was almost 3-fold higher than when Mg 2ϩ is present (V max of 2.00 Ϯ 0.14 and 0.70 Ϯ 0.03 peptide/s, respectively). This is consistent with a ϳ3-fold higher activity of eukaryotic OST enzymes in the presence of manganese compared with magnesium, determined at 10 mM metal ion concentrations for yeast OST (37) and 3 mM metal ion concentrations for hen oviduct OST (39).
Specific Interactions of the ϩ2 Position for Sequon Binding-The PglB structure suggested specific interactions between the sequon and the conserved WWD-motif. The hydroxyl group of the ϩ2 Thr side chain can form hydrogen bonds to Trp-463, Trp-464, and Asp-465. An additional hydrogen bond is possible between Asp-465 and the backbone of the acceptor peptide (Fig. 5A). In addition, a van der Waals (vdW) contact between Ile-572 and the methyl group of the ϩ2 Thr was proposed. Ile-572 is not only conserved in bacterial homologues, but can also be found in eukaryotes and some archaeal homologues (Fig. 5B). A mutation of the corresponding isoleucine to alanine in C. jejuni PglB reduced its in vitro activity (40). To test the role of the postulated vdW interaction with Ile-572, we performed binding studies with different sequon and PglB variants. We observed that peptide binding affinity was reduced 4-fold when the ϩ2 Thr was replaced by a serine (Fig. 5C, Table 2, and supplemental Fig. S4). Intriguingly, a similar reduction in peptide affinity was observed when the DQNAT peptide was bound to the I572V mutant (Fig. 5C, Table 2, and supplemental Fig.  S4). This suggested that the removal of a methyl group either from the ϩ2 Thr of the sequon or from Ile-572 of the enzyme leads to a comparable reduction in sequon binding. The I572V mutant revealed a ϳ2-fold higher affinity for a ϩ2 Thr compared with a ϩ2 Ser, suggesting that the methyl group of Thr can still benefit from vdW interactions with a Val-572. A further truncation of Ile-572 to an alanine reduced binding of the DQNAT sequon 57-fold and that of the DQNAS sequon 165fold compared with the WT enzyme (Fig. 5C, Table 2, and supplemental Fig. S4). To compare these effects with catalytic activities, we determined turnover rates of Ile-572 mutants. We found a stepwise reduction in activity for PglB mutants I572V and I572A. In both cases, turnover was 1.2-fold faster for the DQNAT sequon compared with DQNAS (Fig. 5C, Table 2, and supplemental Fig. S2).
In contrast to the I572A mutation, changes in the WWD motif affected peptide binding much more. Mutation of the second Trp residue of the motif (W464F, Fig. 5A) resulted in a 1000-fold slower turnover rate and strongly decreased peptide affinity, making it impossible to determine accurate K d values (Table 2 and supplemental Figs. S2 and S4). Given the sensitivity of our assays, we were able to quantify binding and turnover of non-canonical sequons. We investigated binding and turnover for substrate peptides containing cysteine, valine, or alanine in the ϩ2 position, because the PglB structure revealed

TABLE 2 Glycosylation turnover rates and dissociation constants for PglP and peptide variants affecting interactions at the ؉2 position
Turnover rates and dissociation constants were derived from curves shown in supplemental Figs. S2 and S4. The errors represent the S.E. of the fit for each turnover rate and dissociation constant. Values for relative turnover and relative binding refer to the ratio of an individual mutant/sequon and WT PglB/DQNAT sequon. that these amino acids could fit into the binding pocket. Binding was strongly reduced for the DQNAC, DQNAV, and DQNAA sequons (supplemental Fig. S4 and Table 2), and no accurate K d values could be determined. The glycosylation turnover of the DQNAC sequon was reduced 400-fold compared with DQNAT (and 350-fold compared with DQNAS) ( Table 2 and supplemental Fig. S2). Turnover of DQNAV or DQNAA was reduced 7000-and 4000-fold compared with DQNAT, respectively ( Table 2 and supplemental Fig. S2).

Mutant
Strong Contribution of Ϫ2 Position for Sequon Binding-In contrast to archaeal and eukaryotic Stt3 homologues, bacterial OSTs recognize and specifically interact with a negatively charged amino acid at the Ϫ2 position of the sequon (11,13,27). In vitro studies of C. jejuni PglB expressed in E. coli membranes revealed a preference for a Ϫ2 Asp over a Ϫ2 Glu (41). The X-ray structure of C. lari PglB revealed a possible salt bridge between the Ϫ2 Asp with the side chain of Arg-331 (Fig.  6A), a residue only conserved in bacterial homologues. We have quantified this interaction in C. lari PglB. The affinity of the R331A mutant for DQNAT, as well as that of WT PglB for AQNAT, was virtually abolished (supplemental Fig. S4, Table 3,  and supplemental Table S1), and no K d values could be determined. In vitro glycosylation of the R331A mutant revealed a 400-fold reduction in turnover for the DQNAT sequon, but a 15,000-fold reduction for AQNAT (Table 3 and supplemental Fig. S2). Even though the R331A mutant was unable to form a salt bridge, it nevertheless, revealed a preference for a Ϫ2 Asp over a Ϫ2 Ala in the sequon. Notably, the turnover of WT PglB enzyme with AQNAT was reduced 50,000-fold compared with the consensus DQNAT sequon (Table 3 and supplemental Fig.  S2). We validated these findings using an in vivo glycosylation assay with an acceptor protein containing either a DQNAT or an AQNAT sequon. Wild type PglB glycosylated DQNAT-containing acceptor protein almost quantitatively, but was unable to process AQNAT. The R331A mutant showed reduced glycosylation efficiency for DQNAT but, intriguingly, a very low but non-negligible glycosylation of AQNAT, which was detectable in the anti-glycan immunoblot (Fig. 6B). We thus demonstrate that a mutant bacterial OST enzyme can glycosylate a eukaryotic sequon in vitro and in vivo. However, an approach to invert the interaction at the Ϫ2 position was not successful. Neither PglB mutants R331D nor R331E were able to glycosylate a RQNAT sequon in vivo (data not shown). Another strongly conserved the PglB residue in close vicinity of the Ϫ2 Asp of the sequon is Arg-147, which forms a hydrogen bond with the catalytic Glu-319 (Fig. 6A). A R147K mutant revealed reduced in vivo glycosylation of a DQNAT-containing acceptor protein. A double mutant R147K,R331A was completely inactive (Fig. 6B). FIGURE 6. Interactions between the ؊2 sequon position and PglB. A, ball and stick representation of acceptor peptide (yellow) and interacting PglB residues (cyan), as observed in the X-ray structure (PDB code 3RCE). Sequon and PglB residues are labeled in black and cyan, respectively. Yellow lines indicate N and C termini of the sequon. Proposed hydrogen bonds are shown as dashed lines. A potential interaction between Arg-147 and the Ϫ2 Asp is indicated. B, immunoblots of in vivo glycosylation reactions detecting acceptor protein 3D5 (top) or bacterial N-glycans (bottom). Glycosylation results in a mobility shift from the non-glycosylated (g0) to the glycosylated form of the acceptor protein (g1). PglB mutants are indicated above the lanes, the glycosylation sequons present in the acceptor proteins are indicated below the lanes.

TABLE 3 Glycosylation turnover rates and dissociation constants for sequon and PglB variants affecting interactions at ؊2 position
Turnover rates and dissociation constants were derived from curves shown in supplemental Figs. S2 and S4. The errors represent the S.E. of the fit for each turnover rate and dissociation constant. Values for relative turnover and relative binding refer to the ratio of an individual mutant/sequon and WT PglB/DQNAT sequon.

DISCUSSION
Despite its essential nature, the reaction mechanism of oligosaccharyltransferase is insufficiently understood. This is not only due to the scarcity of structural insight (only one structure of a full-length enzyme available), but also due to the challenges of studying dynamic and labile membrane proteins in vitro.
The crystal structure of C. lari PglB revealed bound acceptor substrate in the absence of the donor substrate, which distinguishes PglB from glycosyltransferases (35). Hence, this allowed us to study acceptor binding as a discrete step preceding catalysis. However, we are aware that the sequence of events of the glycosylation reaction might be different from the mechanistic model for the in vitro reaction as represented in Fig. 1.
Guided by our crystal structure, we have introduced specific mutations in PglB and acceptor peptides, and report the first quantification of sequon binding affinities and turnover rates obtained with purified enzyme. For this investigation, milligram amounts of monodisperse PglB retaining its stability at concentrations above 20 mg/ml were required, which was challenging due to the low expression levels and the inherent aggregation behavior of PglB in detergent micelles. PglB expression levels ranged from 0.08 to 0.12 mg/g of E. coli cells, and only protein variants that showed excellent stability and monodispersity at concentrations up to 300 M were evaluated. To prevent aggregation, all anisotropy measurements were carried out at 4°C. For wild type enzyme and consensus sequon DQNAT, binding was also measured at 23°C and yielded a K d value of 1.62 Ϯ 0.02 M (supplemental Fig. S5), indicating that the temperature increase only slightly reduces peptide binding affinity. The concentrated protein was analyzed by size exclusion chromatography before and after the measurements. The second experimental requirement was the preparation of reaction substrates. LLO could be prepared from E. coli cells transformed with the genes allowing it to biosynthetically produce C. jejuni LLO, which can serve as a substrate for the C. lari enzyme. Acceptor peptides with attached fluorescent dyes had previously been used for in vitro glycosylation (42). However, our synthetic fluorescently labeled peptides could also be used for independent measurements of peptide binding to PglB by fluorescence anisotropy, a highly sensitive technique.
For the consensus sequon (DQNAT), we performed both Michaelis-Menten kinetics and turnover rate analysis. The peptide concentration used in the time course (10 M) was ϳ4 times the observed K m value (2.6 M), and the obtained turnover rate was close to the observed V max . For all other reactions reported here, initial turnover rates were measured instead of Michaelis-Menten kinetics. This was the consequence of a technical limitation. As illustrated in Fig. 2, C and E, accurate detection of the ratio of unmodified acceptor peptide to the glycosylated product is only possible when both are measurable from the same gel. In addition, for very unfavorable combinations of enzyme mutants or sequon variants, the solubility of the fluorescently labeled peptides prevented the use of concentrations that guaranteed maximum turnover (V max ). We therefore determined turnover rates using constant concentrations of synthetic fluorescently labeled peptides and variable amounts of PglB mutants. For reactions with a low enzyme/ substrate ratio, the obtained turnover rates can be directly compared with wild type because the peptide concentration was above the K m value and the determined turnover rates were consequently a good approximation of V max . For the reactions of highly disfavored enzyme mutants or poor substrate peptide analogues, turnover rates were quantified at high enzyme/substrate ratios. Under these conditions, the determined turnover rates did not reach V max . However, not only were comparisons largely reliable, but studying the enzymatic reaction at a fixed peptide concentration also reflects the natural scenario inside a cell, where a constant flow of peptide segments are presented to OST. In addition to the caveat above, the absolute turnover rates cannot be transferred directly to the in vivo situation because (i) the concentration of divalent metal ions inside a cell (periplasm or lumen of the ER) is much lower than that used in in vitro assays, (ii) glycosylation sequons are embedded in larger polypeptide chains, and (iii) in vivo OST and LLO are surrounded by a lipid bilayer rather than detergent micelles. Nevertheless, the rates reported here are highly useful for understanding the reaction mechanism of PglB. Our study therefore provides the first direct determination of a dissociation constant of an isolated acceptor sequon peptide and a purified OST enzyme. Our data revealed an intricate mutual dependence of peptide and metal binding for successful catalysis. Several conclusions can be drawn.
1) Residues in the active site of PglB are essential for catalysis, but not necessarily for peptide binding. Mutations of the essential Asp-56 and Glu-319 to alanine not only decreased in vivo and in vitro rates, but also drastically lower in vitro peptide affinity. This suggests that, along with bound metal, these residues provide stabilizing interactions to the amide group of the acceptor asparagine. In contrast, the conserved Arg-375 was essential for catalysis, but not for binding. Whereas replacing Arg-375 with a lysine preserved peptide binding and catalytic activity, changing it to an alanine abolished catalysis, but without affecting peptide binding. This observation agrees with the hypothesis that Arg-375 provides critical interactions with the pyrophosphate moiety in the LLO substrate. It also validates the experimental approach of studying peptide binding and catalysis using distinct techniques to obtain mechanistic insight.
2) The bound metal ion in the active site of PglB is not only essential for catalysis, but also for peptide binding. The requirement of a divalent cation for OST activity is generally accepted (14, 23, 36 -38). However, it was primarily assumed to act as a Lewis acid to stabilize the lipid-pyrophosphate leaving group, in analogy to metal-dependent glycosyltransferases (35). As the metal ion is not in direct contact with the acceptor peptide, its strict requirement for peptide binding to PglB was unexpected. However, it is consistent with the finding that active site residues Asp-56 and Glu-319, which contact bound metal ion, also contribute to peptide binding.
3) The dependence of PglB activity on relatively high concentrations of metal ions has important implications for in vivo OST function. At 10 mM, bound manganese caused a 17-fold stronger sequon binding than magnesium. This effect may be due to different coordination properties of the two metals (43), because protein stability was not affected after exposure to ele-vated concentrations of either ion (supplemental Fig. S1). The preference for Mn 2ϩ over Mg 2ϩ was much less pronounced in the in vitro glycosylation reaction, and we conclude that tighter peptide binding induced by Mn 2ϩ might not necessarily correlate with more efficient catalysis. However, our assays cannot reveal whether metal ion binding precedes sequon binding or if these events happen simultaneously. At lower metal concentrations, we observed similar activities for Mn 2ϩ and Mg 2ϩ , which has implications for in vivo situation. Even though Mn 2ϩ has been suggested as the physiological cation (44), it is estimated that the concentration of free Mn 2ϩ in the ER is 10 2 -10 3 times lower than that of Mg 2ϩ (45,46). Similar estimates can be made for the periplasm of Campylobacter species, because in their natural environment, Mg 2ϩ is 10 3 -10 5 times more abundant than Mn 2ϩ (47)(48)(49)(50). The growth media used for in vivo glycosylation in E. coli (LB media) contains ϳ0.23 mM Mg 2ϩ , suggesting that OST operates far below its maximal activity. Therefore, it is not surprising that supplementing growth media with MnCl 2 is a successful strategy to increase the catalytic efficiency in CHO cells (51). However, additional studies are required to determine whether both magnesium and manganese are used in vivo.
4) Hydrogen bonds between the ϩ2 Thr/Ser of the sequon and the surface of PglB provide strong contributions to peptide binding. Only amino acids containing a hydroxyl group (Ser and Thr) can provide the specific network of hydrogen bonds formed with the conserved WWD motif. In principle, the thiol group of cysteine can also form such hydrogen bonds, but they are much weaker than those of the hydroxyl group. We observed strongly reduced in vitro affinity and glycosylation rates of acceptor peptides containing a DQNAC sequon. Nevertheless, glycosylated N-X-C sequons have been found in vivo, for example, in bovine protein C or in human von Willebrand factor (10). Also Bause and Legler (23) showed that a peptide containing an N-G-C site can be glycosylated in vitro using calf liver microsomes, albeit with low efficiency. A mapping of the mouse N-glycoproteome revealed that 1.3% of all N-glycosylated sites are N-X-C sequons, followed by 0.4% N-X-V sequons (21). 5) Isoleucine 572 modulates sequon binding by providing van der Waals interactions. The binding affinities we determined not only confirmed the preference of Thr over Ser (DQNAT versus DQNAS sequon) but could quantify the role of Ile-572 in this phenomenon. Removal of the stabilizing vdW interaction between the side chain of Ile-572 and the ϩ2 Thr was found to reduce both peptide binding and glycosylation rates. Different Ile-572 mutants, combined with either ϩ2 Thr or ϩ2 Ser sequons, showed a positive correlation between these two events (Fig. 7). The mutant I572A, in combination with a DQNAS sequon, revealed the lowest affinity and turnover, whereas WT enzyme and DQNAT yielded the highest binding affinity and glycosylation rates. This correlation illustrates the additive effect of the vdW interactions contributing to increased substrate affinity and hence to increased glycosylation efficiency. However, the effect was less pronounced for catalysis than for sequon binding. We conclude that the reduced vdW interactions could speed up the reaction, possibly by facilitating faster release of glycopeptide from the enzyme, thus counteracting the reduced binding affinity (Fig. 1). Ile-572 is conserved in ⑀-proteobacteria and a mutation to a Val in C. jejuni PglB (I571V) revealed a slightly reduced glycosylation efficiency in vivo (52). In ␦-proteobacteria, Ile-572 is occasionally replaced by another hydrophobic residue. For example, Desulfovibrio alaskensis PglB has a Leu in the corresponding position (Leu-597, Fig. 5B). Given that a Leu provides similar vdW interactions with the distal methyl group as an Ile, we expect that N-glycosylation in D. alaskensis is efficient as well. Identifying the analogues residues of Ile-572 in eukaryotic Stt3 homologues is not straightforward. If biased sequence alignments are used, Ile-572 of PglB corresponds to Ile-593 in the Stt3 subunit of Saccharomyces cerevisiae OST. Intriguingly, the mutation I593A of S. cerevisiae Stt3 indeed resulted in a temperature-sensitive phenotype, whereas the mutation W517Y was lethal (20). 6) An aspartate in the Ϫ2 position of the bacterial sequon contributes strongly to the binding affinity. N-Linked glycosylation in Bacteria is more specific and less frequent than in eukaryotes. Also, only strains representing the clade ⑀and ␦-proteobacteria have been shown to N-glycosylate proteins (6). We studied the effect of removing either partner of the proposed salt bridge between the Ϫ2 Asp of the DQNAT sequon and Arg-331 of PglB. We observed abolished binding affinity and strongly reduced catalysis. Because removing the negatively charged Ϫ2 Asp of the sequon resulted in a much stronger effect than removing the positively charged R331A of PglB, we conclude that the recognition of the Ϫ2 Asp is not exclusively accomplished by Arg-331, but requires additional interactions. One candidate for such an interaction is the side chain of Arg-147, which also forms hydrogen bonds with catalytic residues (Fig. 6A). Its mutation to a lysine indeed reduced sequon binding and catalysis, but it is unclear if the effect is specific.
The 400-fold reduction of glycosylation activity of the R331A mutant correlates with the observation that K m values of sequons of eukaryotic OST enzymes are roughly 2 orders of magnitude higher than those of bacterial sequons binding to PglB (23,41,53). With the extension of the glycosylation sequon, Bacteria probably increase their substrate specificity, FIGURE 7. Correlation of peptide binding and glycosylation turnover. Values for peptide binding and turnover rates presented in Fig. 4C are correlated in a two-dimensional plot. Peptide binding is presented as log (1/K d ). Error bars indicate the S.E. of each fit, but are smaller than the symbols in most cases.
achieving higher stringency. This might be required to recruit acceptor sites in exposed and flexible regions of folded proteins in the periplasm (32,54). In contrast, eukaryotic OST enzymes are proposed to be localized next to the secretion machinery in the ER membrane, resulting in a high local concentration of sequons in the nascent polypeptide chains (15,17,19,55). In summary, our data highlights the details of the complex network of interactions between the sites of peptide, metal, and LLO binding on the surface of PglB, allowing sequon recognition and glycosylation to be studied as distinct steps.