Engineering Bifidobacterium longum Endo--N-acetylgalactosaminidase for Neu5Ac2-3Ga11-3Ga1NAc reactivity on Fetuin

-

Crystal structures of EngBF [7], as well as the enzyme EngSP from two Streptococcus pneumoniae strains, R6 8 and TIGR4 [9], have been reported. The EngSP(TIGR4) structure [9] confirmed interactions with the substrate that were previously inferred from docking experiments [7,8,10]. In addition, detailed substrate interactions of EngSP were resolved from complexes with both substrate and product that also confirmed the role of a dual tryptophan lid closing over the active site, as proposed previously [7].
N-glycans are removed by a single enzyme, peptide:N-glycosidase or PNGase, which is responsible for releasing the branched and matured glycan [11,12]. In an analogy for O-glycans and their release from glycoproteins and -peptides, endo-α-N-acetylgalactosaminidase is sometimes referred to as O-glycanase despite that the activity of this enzyme is the last to follow several enzymatic steps in mucin like O-glycan removal [13]. Not only for analytical purposes could an enzyme that can catalyze the release of full-size O-glycans seem attractive, but also if such an enzyme could facilitate transglycosylation by these O-glycans on to e.g. peptides or peptide synthesis precursors.
The present work describes the successful mutational analysis of EngBF to generate an enzyme releasing trisaccharide Neu5Acα2-3Galβ1-3GalNAc from fetuin based on structural models obtained by molecular docking of this extended glycan into the wild-type enzyme and a small set of in silico generated variants having single or double side-chain replacements. The template for the docking experiments was the crystal structures of the EngSP enzyme that allow for comparison of the glycan position in the docked complexes with that of the known crystal structure of the Galβ1-3GalNAc complex.

Experimental procedures
Materials. GalNAc for standard solutions and DMAB for colorimetric detection of GalNAc and for use in the enzyme assay were purchased from Carl Roth (Germany). Fetuin from fetal bovine serum and all other chemicals were from Sigma Aldrich (United States). Standards used for peak identification in product characterization by anion chromatography of saccharides released from fetuin were: Gal (Merck, Germany), GalNAc, Neu5Ac and Galβ1-3GalNAc (Carbosynth, Great Britain).
Expression plasmids, protein synthesis and purification. Expression plasmids for synthesis of the mutant variants of EngBF were prepared by the two-reaction site directed mutagenesis procedure [14] as described previously [15]  Protein synthesis in the E. coli Rosetta (DE3) strain transformed with plasmids encoding wild-type EngBF or variants and subsequent protein purification was performed as previously described [15].
A codon optimized reading frame encoding PmST1 sialyltransferase in a truncated form lacking residues 2-25 16 was ordered from Ther-moFisher (United States) for recombinant expression in E. coli. Sites for restriction endonucleases were added; NdeI, that overlap the start codon, and BamHI that immediately follows the last codon of the reading frame. The above NdeI/BamHI fragment was transferred to a pET11a derived plasmid encoding a six histidine residue tag added to the 3 ′ -end of a reading frame inserted between the NdeI and BamHI sites of this vector. The resulting plasmid, pDKH13, was sequenced to verify the coding region of the insert. Protein synthesis and purification of PmST1 using Ni-NTA chromatography was performed as previously described for EngBF [15]. NanI sialidase was produced and purified as previously described [15].
The ligand Neu5Acα2-3Galβ1-3GalNAc was built using the carbohydrate builder at Glycam.org [22] and prepared for docking in the Maestro LigPrep wizard with charges calculated for pH 6. Ligand docking used the Glide SP algorithm [19,20] with flexible ligand sampling, in which hydrogen bond formation is rewarded. The initial output was set to a maximum of 10.000 poses, which were subsequently energy minimized in a full force-field to optimize bond length, -angles and torsional angles before scoring of the poses using the modified Glide Energy score [18,19].
The RMSD values between the co-crystallized ligand, Galβ1-3Gal-NAc, and this moiety of the docked trisaccharide Neu5Acα2-3Galβ1-3GalNAc in the poses were calculated and plotted against the Glide Energy score. The poses were evaluated for lowest values of RMSD for the docked ligand compared to the co-crystallized disaccharide and the Glide Energy Score.
Surface electrostatic potentials were calculated using the Adaptive Poisson-Boltzmann Solver package [23] in PyMol (Shrödinger Inc.) and displayed using the electrostatic potential from ±5 K b T/e. Analysis of enzymatically released saccharides. The identity of the glycan released from fetuin by the EngBF variants was analyzed and established in two ways. The first method using HPEAC-PAD was performed as follows, 1 mL of 10 mg/mL fetuin was incubated with 1.5 μM of the E1294K variant at 37 • C for 72 h in 50 mM Acetic acid, 100 mM NaCl, pH 6.0. After incubation, the released saccharides were isolated from enzyme and fetuin, by filtering through an Amicon Ultra-0.5 mL 3 K MWCO centrifugal filter (Merck, Germany). The pH of 250 μL of the filtrate was adjusted to 5.0 by addition of HCl, and subsequently added 2 μL of 200 mM CMP and 36 μg of purified PmST1 sialyltransferase and incubated for 2 h at 37 • C. PmST1 was then removed by filtering as above. The filtrate of fetuin reacted with E1294K and the sample additionally treated with PmST1 as well as the standard markers to be used for anion chromatography, were all diluted to 50 μM in 10 mM NaOH Chemical structure of the glycans Galβ1-3GalNAc and Neu5Acα2-3Galβ1-3GalNAc. A) Galβ1-3GalNAc. B) Neu5Acα2-3Galβ1-3GalNAc. The sialidase activity of NanI and PmST1 sialyltransferase converts Neu5Acα2-3Galβ1-3GalNAc to Galβ1-3GalNAc. PmST1 sialyltransferase is specific for the Neu5-Acα2-3Galβ bond.
based on the concentration determined by the Morgan Elson colorimetric assay [24,25] for the glycans (see above) or the known stock concentration for small-molecule markers. The samples and standards were then analyzed on a ICS-3000 (Dionex) chromatography system equipped with a CarboPac PA200 column (Dionex, ThermoFisher Scientific) and a gold electrode. The samples were eluted steadily using a mobile phase (0.35 mL/min) of 100 mM NaOH and a stepwise gradient of sodium acetate (from 0 to 25 min, 0-75 mM; 25-30 min, 75-400 mM). Peaks were identified using QuickPeaks in the OriginPro 9.1 suite, which was also used for visualization.
The second method using TLC of products from incubation of fetuin and asialo-fetuin with E1294K was performed as follows: Fetuin (10 mg/ mL) was incubated with E1294K for 24 h at 37 • C. Upon completion, a fraction of this incubation was added NanI sialidase, prepared as described above, to a final concentration of 2.6 μg/mL and incubated for 24 h to remove the Neu5Ac moiety from the released fetuin glycans. The saccharides from samples of enzymatically treated fetuin were isolated from protein by filtering through an Amicon Ultra-0.5 mL 3 K MWCO centrifugal filter (Merck, Germany). The filtrate was loaded on a Silica gel 60 (Merck, Germany) stationary phase by pipetting 5 times 1 μL and allowing the spot to dry between loadings. The chromatogram was developed in a closed chamber with a 1-propanol:28% ammonia:water (15:1:6, v/v) mobile phase and subsequently dried before the migrated glycan moieties were visualized by submersion in a solution composed of 2 mL aniline, 10 mL phosphoric acid, 2 g diphenylamine, and acetone to a final volume of 100 mL, followed by heating at 100 • C for 10-15 min.
Steady-state kinetic analysis. Fetuin substrate was prepared as follows: 50 mg/mL of fetuin dissolved in 50 mM sodium acetate, 250 mM NaCl, pH 6.0 was dialyzed against the same buffer and centrifuged to recover the soluble fraction resulting in a solution with a concentration ranging from 30 to 35 mg/mL. For preparation of asialo-fetuin, the fetuin solution from above was added NanI sialidase to a final concentration of 2.6 μg/mL and incubated for 24 h at 37 • C, after which no further increase in Galβ1-3GalNAc release upon treatment with EngBF was detected. Subsequently, the asialo-fetuin preparation was dialyzed against several changes of 50 mM sodium acetate, 50 mM NaCl, pH 4.5.
Assays of glycan hydrolysis on fetuin and asialo-fetuin based on quantification of released Neu5Acα2-3Galβ1-3GalNAc or Galβ1-3Gal-NAc using the Morgan-Elson colorimetric reaction [24,25] were performed as described previously [15], except for changing the incubation temperature to 30 • C. Examples of data sets including progress-curves used in calculations of initial rates and glycan concentrations for the processing of fetuin and asialo-fetuin by the EngBF variant E1294K can be found in Supplementary Materials. Depending on the kinetic parameters of EngBF or the studied variants the concentration of fetuin or asialo-fetuin was varied up to 30 mg/mL, corresponding to approximately 1.8 mM cleavable Galβ1-3GalNAc or Neu5-Acα2-3Galβ1-3GalNAc in the assay incubation. The exact concentration range used for wild-type EngBF and its variants are listed in Tables 1 and 2. As stated in results, assays were performed either at pH 4.5 (50 mM sodium acetate, 50 mM NaCl) or pH 6.0 (100 mM sodium acetate, 50 mM NaCl). Assay time and enzyme concentrations were adjusted accordingly to obtain full progress-curves as described previously [15]. Fits to obtain initial velocities and substrate concentration from progress curves, as well as steady-state kinetic parameters and their reported errors from the derived saturation curves, were performed as previously described [15].

Results and discussion
Molecular ligand docking in the endo-α-N-acetylgalactosaminidase active site. To probe the accommodation of the Neu5Acα2-3Galβ1-3GalNAc glycan in the endo-α-N-acetylgalactosaminidase active site, molecular dockings with this ligand were performed using the software Glide as described in Experimental procedures. The highresolution structure of the EngBF homolog, EngSP, in complex with Galβ1-3GalNAc (PDB ID: 5A59) was used as a template for the docking. For each docking experiment the RMSD obtained from comparing the position of the docked Galβ1-3GalNAc moiety of the extended glycan to that of the co-crystallized disaccharide Galβ1-3GalNAc in the EngSP structure were plotted against the Glide energy score ( Fig. 2A). Intriguingly, the extended glycan could readily be docked into the active site to position itself close to the subsites of the Galβ1-3GalNAc glycan. Analyzing the best scoring poses for the extended glycan, demonstrated a marked favorable decrease in the Glide energy score concomitant with lowered RMSD values. In the docked complexes, the galactose C6hydroxyl in the extended glycan was either within hydrogen bonding Table 1 Kinetic parameters of O-glycan release for wild type EngBF and mutant variants using asialo-fetuin as a substrate at pH 4.5 a .
Enzyme  distance of the Asp-658 carboxylate, common to the crystal structure of the EngSP Galβ1-3GalNAc complex, or rotated towards the Asp-1254 carboxylate (Fig. 2B). Inspection of the best poses, judged by lowest values of Glide energy score and RMSD, showed that the Neu5Ac C1 carboxylate was either pointing towards the enzyme surface (Fig. 2C) or the solvent (Fig. 2D). Not surprisingly, the docking experiments with the Neu5Acα2-3Galβ1-3GalNAc ligand in the EngSP active site show that binding of this glycan is restricted by accommodation of the Neu5Ac moiety. And perhaps more important, the docking experiments indicate that the Neu5Ac residue exerts restrictions on correctly locating the galactose residue to its subsite as defined in the binding of the Galβ1-3GalNAc glycan. Thus, a simple metric for a rational substitution of side chains in the active site of the enzyme that facilitate docking of Neu5-Acα2-3Galβ1-3GalNAc, seemed to be the aim for a further concomitant decrease in the Glide energy score and RMSD.
To provide some guidance for the engineering of the EngBF active site, we tested both wild-type enzyme and different EngBF variants with alanine substituted amino acid residues for the ability to cleave the Neu5Acα2-3Galβ1-3GalNAc glycan of fetuin. Most variants are previously described [7,15] and form part of the active site lid, W750A, or comprise the binding site for the galactose moiety, Q894A, K1199A and D1295A (Fig. 3A). To complement these, we also constructed the E1294A variant, as Glu-1294 is also potentially blocking an extended binding site for Neu5Acα2-3Galβ1-3GalNAc (Fig. 3A). Despite recent findings with endo-α-N-acetylgalactosaminidases from other sources [6], no activity on fetuin was detected for the wild-type EngBF or the variants W750A, Q894A, K1199A, D1295A under the reaction conditions and assay method used in the present work. Nevertheless, the side-chain substitutions at the positions corresponding to Gln-894 and Lys-1199 in EngBF were recently shown to improve the activity of EngSP The side-chain numbering is according to EngSP and numbering for the corresponding residues in EngBF is shown in parentheses. A) The Glide energy score plotted as a function of the RMSD calculated for the superimposition of the Galβ1-3GalNAc moiety of the docked glycan, Neu5Acα2-3Galβ1-3GalNAc, and that of Galβ1-3GalNAc co-crystallized with the EngSP enzyme. The inset shows all the data using an extended xaxis. The data points in the square of the main plot represent the poses with lowest RMSD and Glide energy score. B) Superimposition of the co-crystallized Galβ1-3GalNAc in the active site of EngSP (dark grey) and selected poses (green) from docking this ligand into the EngSP structure with low Glide energy score and RMSD values marked within the square of the main plot in A). The figure was created using PyMol (Shrödinger Inc.). C) Docking poses of Neu5Acα2-3Galβ1-3GalNAc (green) in EngSP where the Neu5Ac C1 carboxyl moiety, shown in spheres, is solvent exposed. Side chains interacting with the Neu5Ac moiety are displayed as sticks (dark grey). The figure was created using PyMol (Shrödinger Inc.). D) Same as C) but where the Neu5Ac C1 carboxyl moiety is buried within the active site. Side chains interacting with the Neu5Ac moiety are displayed as sticks (dark grey). The figure was created using PyMol (Shrödinger Inc.). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)  A) The Glide energy score plotted as a function of the RMSD calculated for the superimposition of the Galβ1-3GalNAc moiety of the docked glycan, Neu5Acα2-3Galβ1-3GalNAc, and that of Galβ1-3GalNAc co-crystallized with the EngSP enzyme. The inset shows all the data using an extended x-axis. The data points in the square of the main plot represents the poses with lowest RMSD and Glide energy score for the model EngSP E1253K variant (homologous to the E1294K variant of EngBF). B) Superimposition of the Galβ1-3GalNAc co-crystallized with the EngSP enzyme (dark grey) and the best poses (green) from dockings of Neu5Acα2-3Galβ1-3GalNAc into the model EngSP E1253K structure. Side chains interacting with the Neu5Ac moiety are displayed as sticks (dark grey). The side-chain numbering is according to EngSP and numbering for the corresponding residues in EngBF is shown in parentheses. The figure was created using PyMol (Shrödinger Inc.). C) The Glide energy score plotted as a function of the RMSD calculated for the superimposition of the Galβ1-3GalNAc moiety of the docked glycan, Neu5Acα2-3Galβ1-3GalNAc, and that of Galβ1-3GalNAc co-crystallized with the EngSP enzyme. The inset shows all the data using an extended x-axis. The main plot shows the poses with lowest RMSD and Glide energy score for the model EngSP E1253K variant and the double variants, E1253K:Q868A and E1253K:K1156A (homologous to the E1294K, E1294K:Q894A and E1294K:K1199 variants of EngBF). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) on fetuin up to four-fold [6]. EngBF is almost identical to EngSP in the region of the active site (Fig. 3B) suggesting that our assay is not sensitive enough to detect activity at the level reported for EngSP under the reaction conditions that we have chosen. However, the E1294A variant was found to release the Neu5Acα2-3Galβ1-3GalNAc glycan when fetuin was used as a substrate. This observation was followed up by analysis of several in silico side-chain substitutions at the homologous position, Glu-1253, in the EngSP structure used for the docking experiments. Here, the E1253K variant was found to produce docking poses with the Neu5Acα2-3Galβ1-3GalNAc glycan having a markedly lower RMSD and Glide energy score compared to wild type enzyme and that of other side-chain substitutions at this position (Fig. 4A). Dissection of the best poses from docking of the extended glycan in the E1253K variant (Fig. 4B) shows the hydroxyl groups of C5 and C6 of the galactose moiety to be within hydrogen bond distance of the Asp-658 and Asp-1254 carboxylates. Thus, the interactions of the wild-type enzyme and the Galβ1-3GalNAc ligand found in the crystal structure of the EngSP complex and those of this moiety of the Neu5Acα2-3Galβ1-3GalNAc glycan in dockings performed on the E1253K variant, are in better agreement than what is found with the latter glycan in the dockings with the wild-type EngSP enzyme. In addition, the positioning of the Neu5Ac moiety is identical for all the dockings in this group of best poses for the E1253K variant with a solvent exposed positioning of the Neu5Ac carboxylate (Fig. 4B).
In docking experiments, we also tested the influence of substituting the side chains of Gln-868 and Lys-1156 in EngSP, corresponding to Gln-894 and Lys-1199 in EngBF, when combined with the E1253K substitution (E1294K in EngBF). The E1253K:Q868A variant showed poses of docking with Neu5Acα2-3Galβ1-3GalNAc with a decreased Glide energy score and reduced RMSD as also found for the E1253K variant above (Fig. 4C).
Product analysis. The finding that the EngBF variants substituted at Glu-1294 were able to produce a detectable product from incubation with fetuin, in contrast to wild-type enzyme, gave a strong indication that these variants were able to process the extended glycan, Neu5Acα2-3Galβ1-3GalNAc. To substantiate this we set out to identify the chemical composition of the glycan released by the EngBF variants, exemplified by E1294K.
The product released by E1294K upon incubation with fetuin was analyzed by HPEAC-PAD as described in Experimental procedures, before and after α2-3-specific Neu5Ac removal by PmST1 sialyltransferase in the presence of CMP [16]. Although the Galβ1-3GalNAc entity was fully hydrolyzed to galactose and GalNAc during the HPEAC-PAD analysis, as this is performed in 100 mM NaOH, distinct peaks that are likely to represent the full trisaccharide Neu5-Acα2-3Galβ1-3GalNAc or the partial hydrolysate Neu5Acα2-3Gal eluted at a position later than the Neu5Ac monomer (Fig. 5A). Supporting this attribution of the late eluting peaks, was the disappearance of these peaks after treatment of the released glycan with PmST1 sialyltransferase and instead the chromatogram showed increased amounts of Galβ1-3GalNAc hydrolysate (Gal + GalNAc) as well as free Neu5Ac (Fig. 5A). μM EngBF and (filled circles) 3.9 μM EngBF E1294K. B) Progress-curves from asialo-fetuin (10 mg/mL) incubated 90 min with a final concentration of (filled squares) 6.7 nM EngBF and (filled circles) 6.7 nM EngBF E1294K.
We also performed a TLC analysis (Fig. 5B) of product from incubation of fetuin with E1294K. The application of standards and the treatment of the product with NanI sialidase indicate that the released glycan is Neu5Acα2-3Galβ1-3GalNAc.
Steady-state kinetic analysis of wild-type EngBF and variants using asialo-fetuin or fetuin as a substrate. The EngBF variants altered at position Glu-1294 were subjected to a steady-state kinetic analysis on asialo-fetuin to evaluate the impact of the substituted side chains with respect to the native activity of EngBF of processing Galβ1-3GalNAc O-glycosylations ( Table 1). The steady-state kinetic parameters for all variants were close to those of wild-type EngBF (Table 1) and Glu-1294 appears not in any way critical for activity. In addition, this position seems functionally very tolerant to side-chain substitutions, both in terms of size and charge. With fetuin as a substrate, all the EngBF variants altered at Glu-1294 were active, but showed a linear increase in activity with increasing fetuin concentrations and no sign of substrate saturation at up to 30 mg/mL fetuin. This only allowed for the determination of k cat /K M , both at pH 4.5 and pH 6.0 (Table 2). It was also clear that all the variants were far better catalysts at pH 6.0 than at pH 4.5 -up to about 50-fold improved for the E1294A variant ( Table 2). Regardless of the pH of the incubation, no significant activity of the wild-type EngBF enzyme was detected ( Table 2). Inspired by the docking experiments and the previously published work on EngSP [6], we also investigated if the double substituted variants, E1294K:Q894A and E1294K:K1199A, showed any improved activity on fetuin over that of E1294K alone. Such an improvement, although limited to 1.5 to 4-fold, was observed for the E1294K:Q894A variant both at pH 4.5 and pH 6.0 ( Table 2).
Conclusion. In the present work, we have shown that abolishing the negative side chain on a single position in EngBF, Glu-1294, promotes the accommodation of the larger and negatively charged glycan, Neu5Acα2-3Galβ1-3GalNAc, irrespective of the size of the substituted side chains. However, the full potential of engineering EngBF to accept this extended glycan as a substrate, is most likely not fully exhausted with the variants described in the present work. The steady-state kinetic analysis with fetuin as a substrate, where no saturation was observed at the concentrations of fetuin available, indicates that much can be gained by improving the binding of Neu5Acα2-3Galβ1-3GalNAc to the active site of EngBF. We speculate that if: 1) positioning of the extended glycan in the binding site of EngBF can be achieved with the Galβ1-3GalNAc moiety exactly overlapping that of the native substrate, and 2) that the structural and chemical environment in the active site where the cleavage reaction takes place is fully preserved, then a similar k cat for the cleavage of both glycans may be expected. This is important, because if the problem of cleaving glycans extended beyond the Galβ1-3GalNAc moiety is reduced to making space for the glycan to position itself to overlap the native substrate, the task is by far more simple than having to manipulate the catalytic machinery. If we were to assume a similar k cat for release of glycans by the EngBF variants active on both asialofetuin and fetuin, the values of k cat /K M for Neu5Acα2-3Galβ1-3GalNAc (Table 2) indicate a K M for fetuin of the EngBF variants to be in the molar range at pH 4.5 and in the millimolar range at pH 6.0. Also, it may be that the greatly improved catalytic efficiency displayed by the variants at pH 6.0 compared to pH 4.5, points to a titration of the Neu5Acα2-3Galβ1-3GalNAc glycan resulting in a favorable uniform charge distribution at the higher pH. In the future, docking experiments with Neu5Acα2-3Galβ1-3GalNAc in the active site of EngBF, evaluated as here by superimposition of the glycan moiety and the native substrate, could be combined with e.g. possible neutralization of electrostatic repulsion ( Fig. 7A and B). This approach may pave the way for engineering endo-α-N-acetylgalactosaminidase variants with much improved activity towards this extended and negatively charged glycan. As for now, the enzyme variants described here may serve as a tool for identification of O-glycosylations and for determination of stoichiometry of the Neu5Acα2-3Galβ1-3GalNAc glycan in glycoproteins, as suggested previously [6]. But perhaps more interestingly, is the further development of a variant with similar activity levels on Neu5-Acα2-3Galβ1-3GalNAc as for wild-type enzyme on Galβ1-3GalNAc, which may find its use in applications involving transglycosylation, which the wild-type endo-α-N-acetylgalactosaminidase performs readily [2,3,26]. Such EngBF variants could show to be important tools in the synthesis of O-linked glycoconjugates.

Funding sources
This work was supported by the Danish Dairy Research Foundation, the Villum Fonden and the Independent Research Fund Denmark (9041-00126B) to MW. Additional funding was from a PhD scholarship to BS from the Ministry of Higher Education and Scientific Research of Iraq. Fig. 7. Surface electrostatic potential of the substrate bound EngSP(TIGR4) active site and Neu5Acα2-3Galβ1-3GalNAc calculated as described in Experimental procedures. A) Surface electrostatic potential of EngSP(TIGR4) (PDB ID: 5A59) in complex with Galβ1-3GalNAc shown as sticks (green) and displayed using the electrostatic potential from ±5 K b T/e. B) Surface electrostatic potential of the extended glycan, Neu5Acα2-3Galβ1-3GalNAc, found on fetuin and shown using the electrostatic potential from ±5 K b T/e. Both figures were created using PyMol (Shrödinger Inc.). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)