Glycoprotein In Vitro N-Glycan Processing Using Enzymes Expressed in E. coli

Protein N-glycosylation is a common post-translational modification that plays significant roles on the structure, property, and function of glycoproteins. Due to N-glycan heterogeneity of naturally occurring glycoproteins, the functions of specific N-glycans on a particular glycoprotein are not always clear. Glycoprotein in vitro N-glycan engineering using purified recombinant enzymes is an attractive strategy to produce glycoproteins with homogeneous N-glycoforms to elucidate the specific functions of N-glycans and develop better glycoprotein therapeutics. Toward this goal, we have successfully expressed in E. coli glycoside hydrolases and glycosyltransferases from bacterial and human origins and developed a robust enzymatic platform for in vitro processing glycoprotein N-glycans from high-mannose-type to α2–6- or α2–3-disialylated biantennary complex type. The recombinant enzymes are highly efficient in step-wise or one-pot reactions. The platform can find broad applications in N-glycan engineering of therapeutic glycoproteins.


Introduction
Protein N-glycosylation is an important post-translational modification that affects the structure, property, and function of glycoproteins including folding, solubility, stability, localization, trafficking, molecular recognition, and interactions, etc. Glycoprotein Nglycans are attached via the innermost N-acetylglucosamine residue to the L-asparagine residue (GlcNAcβ1-Asn) in the Asn-X-Ser/Thr sequon (where X is an amino acid that is not an L-proline) of the protein with a β-linked N-glycosidic bond. All eukaryotic glycoprotein N-glycans share a trimannosyl chitobiose (Man 3 GlcNAc 2 ) core and can be classified as high-mannose, complex, and hybrid types based on the glycan structures extended from the terminal mannose residues on the core [1,2].
Many therapeutic proteins and enzymes are N-glycosylated. The level of N-glycosylation and the structure of their N-glycans can directly affect their solubility, stability, safety, function, efficacy, delivery, pharmacokinetics, immunogenicity, and dose frequency [1,[3][4][5][6][7][8][9]. Therefore, N-glycosylation is a critical quality attribute (CQA) of glycoprotein therapeutics considered by regulatory authorities [3,5,10]. Homogeneous glycoproteins with preferred N-glycoforms are highly desirable for their pharmaceutical applications and for exploring the fundamental understanding of their functions at the molecular level [7].
Nevertheless, N-glycosylated glycoproteins are intrinsically heterogeneous with variations on the N-glycan location, site occupancy of N-glycosylation, and N-glycan structures at individual N-glycosylation sites [11][12][13], and the N-glycosylation process is influenced by many factors [3,14]. Cells of different origins have been used to produce N-glycosylated glycoproteins [4]. Numerous strategies including cell line engineering, as well as the addition of inhibitors for metabolic glycosylation and glycosidase have been developed to reduce glycoprotein N-glycan heterogeneity [15,16]. Synthetic methods including total

Glycoprotein In Vitro N-Glycan Processing Route Design and Enzyme Selection
To establish an efficient platform for processing glycoprotein N-glycans from highmannose-type to disialylated complex biantennary N-glycans, commercially available bovine pancreatic RNase B was chosen as a model. It is a relatively small glycoprotein of 124 amino acids with a single N-glycosylation site at Asn34, which is attached with high mannose-type N-glycans (Man [5][6][7][8][9] GlcNAc 2 ) containing five to nine mannose residues [32][33][34]. As eukaryotic glycosyltransferases involved in N-glycan processing are well known and the corresponding bacterial alternatives have not been fully identified, attempts to express the former in E. coli are one of the focuses of our enzymatic glycoprotein Molecules 2023, 28, 2753 3 of 18 in vitro N-glycan processing strategy. Meanwhile, bacterial alternatives of eukaryotic Nglycan processing mannosidases and glycosyltransferases that have already been identified with desired functions are great choices for developing the efficient glycoprotein in vitro N-glycan processing platform. As shown in Figure 1, the heterogeneous Man [5][6][7][8][9] GlcNAc 2 N-glycans on RNase B can be processed to a homogeneous Man 5 GlcNAc 2 N-glycan by removing all α1-2-linked mannose residues, using a recombinant Enterococcus faecalis α1-2-mannosidase (EfMan-I) expressed in E. coli that we reported previously [34]. EfMan-I belongs to the carbohydrate active enzyme (CAZy) glycoside hydrolase [35,36] family 92 (GH92) and requires a divalent metal cation, such as Ca 2+ or Mg 2+ for activity [34]. After EfMan-I treatment, the natural N-glycan processing steps [2] can be followed in vitro using recombinant glycosyltransferases and glycosidases from different origins to form target homogeneous glycoproteins with the desired disialylated biantennary complex-type N-glycans.
amino acids with a single N-glycosylation site at Asn34, which is attached with high mannose-type N-glycans (Man5-9GlcNAc2) containing five to nine mannose residues [32][33][34]. As eukaryotic glycosyltransferases involved in N-glycan processing are well known and the corresponding bacterial alternatives have not been fully identified, attempts to express the former in E. coli are one of the focuses of our enzymatic glycoprotein in vitro N-glycan processing strategy. Meanwhile, bacterial alternatives of eukaryotic N-glycan processing mannosidases and glycosyltransferases that have already been identified with desired functions are great choices for developing the efficient glycoprotein in vitro N-glycan processing platform. As shown in Figure 1, the heterogeneous Man5-9GlcNAc2 N-glycans on RNase B can be processed to a homogeneous Man5GlcNAc2 N-glycan by removing all α1-2-linked mannose residues, using a recombinant Enterococcus faecalis α1-2-mannosidase (EfMan-I) expressed in E. coli that we reported previously [34]. EfMan-I belongs to the carbohydrate active enzyme (CAZy) glycoside hydrolase [35,36] family 92 (GH92) and requires a divalent metal cation, such as Ca 2+ or Mg 2+ for activity [34]. After EfMan-I treatment, the natural N-glycan processing steps [2] can be followed in vitro using recombinant glycosyltransferases and glycosidases from different origins to form target homogeneous glycoproteins with the desired disialylated biantennary complex-type N-glycans. Human N-acetylglucosaminyltransferase I (hGnT-I or hMGAT1) in the CAZy glycosyltransferase [37,38] family 13 (GT13) was chosen to add an N-acetylglucosamine (Glc-NAc) residue β1-2-linked to the α1-3-linked mannose residue on the trimannosyl chitobiose core of the Man5GlcNAc2 N-glycan on the EfMan-I-treated RNase B to form a hybrid-type GlcNAcMan5GlcNAc2 N-glycan. The hGnT-I uses uridine-5′-diphosphate Glc-NAc (UDP-GlcNAc) as the donor substrate and requires a divalent metal cation, such as Mn 2+ as a cofactor. It is highly selective toward Man5GlcNAc2 N-glycan with dramatically decreased activity for Man3GlcNAc2 and other high-mannose-type N-glycans [39]. This acceptor substrate preference is beneficial for our one-pot multienzyme (OPME) N-glycan processing approach described below.
To process the GlcNAcMan5GlcNAc2 N-glycan on glycoproteins further to form Glc-NAcMan3GlcNAc2, the reported Bacteroides thetaiotaomicron α1-6-mannosidase Bt3994 and α1-3-mannosidase Bt1769 [40] were chosen. Similar to EfMan-I, they are Ca 2+ -dependent CAZy GH92 bacterial mannosidases [40]. In contrast to EfMan-I, which is an α1-2mannosidase, Bt3994 was reported as an α1-6-mannosidase to catalyze the cleavage of the Human N-acetylglucosaminyltransferase I (hGnT-I or hMGAT1) in the CAZy glycosyltransferase [37,38] family 13 (GT13) was chosen to add an N-acetylglucosamine (GlcNAc) residue β1-2-linked to the α1-3-linked mannose residue on the trimannosyl chitobiose core of the Man 5 GlcNAc 2 N-glycan on the EfMan-I-treated RNase B to form a hybrid-type GlcNAcMan 5 GlcNAc 2 N-glycan. The hGnT-I uses uridine-5 -diphosphate GlcNAc (UDP-GlcNAc) as the donor substrate and requires a divalent metal cation, such as Mn 2+ as a cofactor. It is highly selective toward Man 5 GlcNAc 2 N-glycan with dramatically decreased activity for Man 3 GlcNAc 2 and other high-mannose-type N-glycans [39]. This acceptor substrate preference is beneficial for our one-pot multienzyme (OPME) N-glycan processing approach described below.
To generate the second antenna in the glycoprotein biantennary complex N-glycans, human N-acetylglucosaminyltransferase II (hGnT-II or hMGAT2, CAZy family GT16) was chosen. It requires a divalent metal cation for catalyzing the transfer of GlcNAc from UDP-GlcNAc to form a β1-2-linkage to the α1-6-linked terminal mannose on GlcNAcMan 3 GlcNAc 2 to produce GlcNAc 2 Man 3 GlcNAc 2 . hGnT-II has been shown to have a substrate binding pocket that interacts with both the α1-6-linked terminal mannose as the glycosylation site and the other GlcNAcβ1-2Manα1-3Manβ branch as the additional "recognition arm" [41]. This high acceptor substrate selectivity, again, is advantageous for the OPME N-glycan engineering approach presented below.
The in vitro N-glycan processing can be completed by a final sialylation step using a suitable sialyltransferase in the presence of CMP-sialic acid to form RNase B containing homogeneous α2-3or α2-6-sialylated biantennary complex-type N-glycans (Sia 2 Gal 2 GlcNAc 2 Man 3 GlcNAc 2 ). Different sialic acid forms can be introduced in the enzymatic sialylation step [46] and the most common sialic acid form, N-acetylneuraminic acid (Neu5Ac), is introduced as an example in our study presented here.

Enzyme Cloning and Expression
To facilitate the purification of enzymes that will be used for glycoprotein in vitro N-glycan processing, an His 6 -tag was introduced at the C-terminus of each recombinant enzyme to allow its easy purification by Ni 2+ -affinity columns. Furthermore, we found that fusing a maltose binding protein (MBP) at the N-terminus of the target recombinant protein and removing the N-terminal transmembrane domain of mammalian glycosyltransferases by truncation worked well to improve their soluble expression in E. coli [47]. These were the strategies that guided our general design to construct the plasmids for expressing target recombinant enzymes. In addition, E. coli Origami B (DE3) strain harboring pGro7 for chaperon expression was found to be a better choice for expressing mammalian enzymes than E. coli BL21 (DE3) strain, which was used to express recombinant enzymes from bacterial origins.
As we reported previously [34], EfMan-I was expressed as a C-terminal His 6 -tagged soluble and active enzyme in E. coli BL21 (DE3) cells with an expression level of 85 mg/L LB culture (Table 1).
Although Bacteroides thetaiotaomicron mannosidases Bt3994 and Bt1769 were cloned previously in pET21a vector and expressed in E. coli Turner or B834 cells as C-terminal His 6tagged recombinant proteins [40], their expression levels were not reported. We found that removing Bt3994 N-terminal 24 amino acid residues and Bt1769 N-terminal 18 amino acid residues significantly improved their soluble expression levels in E. coli BL21 (DE3) cells. Both ∆24Bt3994-His 6 and ∆18Bt1769-His 6 were expressed at a level of around 55 mg/L LB media (Table 1) as soluble and active proteins with the expected molecular weights of 82 and 83 kDa, respectively (Figure 2A,B).
Although Bacteroides thetaiotaomicron mannosidases Bt3994 and Bt1769 were cloned previously in pET21a vector and expressed in E. coli Turner or B834 cells as C-terminal His6-tagged recombinant proteins [40], their expression levels were not reported. We found that removing Bt3994 N-terminal 24 amino acid residues and Bt1769 N-terminal 18 amino acid residues significantly improved their soluble expression levels in E. coli BL21 (DE3) cells. Both Δ24Bt3994-His6 and Δ18Bt1769-His6 were expressed at a level of around 55 mg/L LB media (Table 1)   To obtain active and soluble recombinant hGnT-I and hGnT-II from E. coli, their Nterminal sequences containing the putative transmembrane domains were removed, and the truncated sequences were expressed as fusion proteins with an N-terminal maltose binding protein (MBP) and a C-terminal His6-tag. The resulting MBP-∆28hGnT-I-His6 and MBP-∆27hGnT-II-His6 were expressed at a level of 5 and 1 mg/L, respectively (Table 1), with the expected molecular weights of 91 and 92 kDa ( Figure 2C,D).
Bβ4GalT1 expressed in E. coli has been purified from inclusion bodies [42,44,48]. We previously cloned and expressed an N-terminal 128 amino acid-truncated Bβ4GalT1 in pET15b as an N-terminal His6-tagged fusion protein (His6-∆128Bβ4GalT1) in E. coli BL21 (DE3) cells, which had a relatively low soluble expression (<1 mg/L culture) [45]. We redesigned the construct to express a protein with both an N-terminal MBP fusion and Cterminal His6-tag. The resulting MBP-∆128Bβ4GalT1-His6 was expressed in a dramatically improved 60 mg/L yield (Table 1) with an expected molecular weight at 75 kDa ( Figure  2E). To obtain active and soluble recombinant hGnT-I and hGnT-II from E. coli, their Nterminal sequences containing the putative transmembrane domains were removed, and the truncated sequences were expressed as fusion proteins with an N-terminal maltose binding protein (MBP) and a C-terminal His 6 -tag. The resulting MBP-∆28hGnT-I-His 6 and MBP-∆27hGnT-II-His 6 were expressed at a level of 5 and 1 mg/L, respectively (Table 1), with the expected molecular weights of 91 and 92 kDa ( Figure 2C,D).
Bβ4GalT1 expressed in E. coli has been purified from inclusion bodies [42,44,48]. We previously cloned and expressed an N-terminal 128 amino acid-truncated Bβ4GalT1 in pET15b as an N-terminal His 6 -tagged fusion protein (His 6 -∆128Bβ4GalT1) in E. coli BL21 (DE3) cells, which had a relatively low soluble expression (<1 mg/L culture) [45]. We redesigned the construct to express a protein with both an N-terminal MBP fusion and Cterminal His 6 -tag. The resulting MBP-∆128Bβ4GalT1-His 6 was expressed in a dramatically improved 60 mg/L yield (Table 1) with an expected molecular weight at 75 kDa ( Figure 2E).
Similar to other eukaryotic sialyltransferases, hST6GAL-I is a CAZy GT29 enzyme [56]. It has been shown to selectively α2-6-sialylate glycoprotein N-glycans [57] and has been successfully expressed in E. coli as a soluble and active fusion protein with MBP at its N-terminus [31]. The soluble expression of the N-terminal MBP-fused and C-terminal His 6tagged N-terminal truncated hST6GAL-I (MBP-∆89hST6GAL-I-His 6 ) that we constructed reached 30 mg/L LB culture (Table 1) with an expected molecular weight at 80 kDa ( Figure 2F).
Campylobacter jejuni CjCst-I is an α2-3-sialyltransferase belonging to CAZy GT42 family [58,59]. It has been shown to utilize Galβ1-3/4OR as acceptors. The 145 residues at the C terminus of CjCst-I were removed and MBP was fused at the N terminus to form MBP-CjCst-I∆145-His 6 . It was successfully expressed in E. coli as a soluble and active fusion protein at a level of 60 mg/L LB culture with an expected molecular weight at 77 kDa ( Figure 2G).

Step-Wise Reactions and Enzyme Activty Determination Using Glycoprotein Substrates
With all enzymes in hand, their activities and applications for glycoprotein in vitro N-glycan processing were tested using step-wise enzymatic reactions and the product formation was monitored by matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectrometry (MS) analysis. Reaction conditions were optimized by varying the types of the buffers used, pH, temperature, ion strength, incubation time, etc.
As shown in Figure 3, treating RNase B (5 mg/mL) ( Figure 3A) with 3% (w/w) EfMan-I-His 6 at 30 • C for 2 h, the high-mannose-type N-glycans Man 5-9 GlcNAc 2 on RNase B were completely trimmed down to Man 5 GlcNAc 2 ( Figure 3B). Treating the resulting RNase B sample with 3% (w/w) MBP-∆28hGnT-I in the presence of MnCl 2 (2 mM) and UDP-GlcNAc (1 mM) at 30 • C for 2 h completed the reaction for the formation of RNase B with a homogeneous GlcNAc 1 Man 5 GlcNAc 2 N-glycan ( Figure 3C). Incubation of the resulting RNase B with 4% (w/w) ∆24Bt3994-His 6 and 3% (w/w) ∆18Bt1769-His 6 in the presence of 2 mM CaCl 2 at 30 • C for 2 h completed the cleavage of the terminal α1-6and α1-3-linked mannose residues to form RNase B with a homogeneous GlcNAc 1 Man 3 GlcNAc 2 N-glycan, which was not cleaved further in the presence of both ∆24Bt3994-His 6 and ∆18Bt1769-His 6 ( Figure 3D). The resulting reaction mixture was incubated with MBP-∆27hGnT-II-His 6 (10% w/w) in the presence of 2 mM MnCl 2 and 1 mM UDP-GlcNAc at 30 • C overnight to form RNase B containing homogeneous GlcNAc 2 Man 3 GlcNAc 2 N-glycan ( Figure 3E). MBP-∆128Bβ4GalT1-His 6 (3% w/w) was then used to process the N-glycan in the resulting RNase B to form RNase B containing homogeneous Gal 2 GlcNAc 2 Man 3 GlcNAc 2 N-glycan by incubating at 30 • C in 2 h in the presence of 5 mM MnCl 2 and 2 mM UDP-Gal ( Figure 3F). Notably, all enzymes used for RNase B N-glycan processing were active in the presence of Tris-HCl (100 mM, pH 7.5).

Multi-Step OPME N-Glycan Processing
Due to the high specificity of the acceptor substrate preference of mammalian glycosyltransferases and the high efficiency of the bacterial mannosidases used in the glycoprotein in vitro N-glycan processing described above, we hypothesized that the step-by-step process was not necessary, and one-pot approaches were possible and could be more efficient. To test this hypothesis, a series of one-pot multienzyme (OPME) reactions were Figure 3. MALDI-TOF MS analysis results for RNase B with heterogeneous high-mannose N-glycans (A) and RNase B sequentially treated with EfMan-I-His 6 (B), MBP-∆28hGnT-I-His 6 (C), ∆24Bt3994-His 6 and ∆18Bt1769-His 6 (D), MBP-∆27hGnT-II-His 6 (E), and MBP-∆128Bβ4GalT1-His 6 (F) in stepwise reactions. The peak marked with an asterisk (*) in each figure is for RNase A, the nonglycosylated ribonuclease that is present in the commercially obtained RNase B sample, which was used as the internal standard. The schematic illustrations of the corresponding reactions are shown. The number above each peak represents the m/z value of the expected product + 2Na species.

Multi-Step OPME N-Glycan Processing
Due to the high specificity of the acceptor substrate preference of mammalian glycosyltransferases and the high efficiency of the bacterial mannosidases used in the glycoprotein in vitro N-glycan processing described above, we hypothesized that the step-by-step process was not necessary, and one-pot approaches were possible and could be more efficient. To test this hypothesis, a series of one-pot multienzyme (OPME) reactions were carried out and the resulting RNase B samples were analyzed directly by MALDI-TOF MS assays after dialysis.
The N-glycan of the RNase B product of the OP5E reaction was released by peptide:Nglycosidase F (PNGase F) and analyzed by MALDI-TOF MS assays. The results ( Figure 5) were consistent with those obtained with the intact RNase B samples. Only the target Nglycan GlcNAc 2 Man 3 GlcNAc 2 (+Na, m/z = 1339.737) was observed without the presence of other N-glycans, confirming the efficiency of the OP5E reactions and the N-glycan homogeneity of the RNase B glycoprotein obtained.
The N-glycan of the RNase B product of the OP5E reaction was released by peptide:N-glycosidase F (PNGase F) and analyzed by MALDI-TOF MS assays. The results ( Figure 5) were consistent with those obtained with the intact RNase B samples. Only the target N-glycan GlcNAc2Man3GlcNAc2 (+Na, m/z = 1339.737) was observed without the presence of other N-glycans, confirming the efficiency of the OP5E reactions and the Nglycan homogeneity of the RNase B glycoprotein obtained. Incubating the reaction mixture of the OP5E reaction with UDP-Gal (2 mM) and MBP-∆128Bβ4GalT1-His6 (3% w/w) at 30 °C for 2 h completed the formation of RNase B with Gal2GlcNAc2Man3GlcNAc2 N-glycan ( Figure 4E).
The formation of RNase B with disialylated biantennary complex N-glycan was accomplished by incubating with MBP-∆89hST6GAL-I-His6 (3% w/w) or MBP-CjCst-I∆145-His6 (3% w/w) in the presence of cytidine 5′-monophosphate-Neu5Ac (CMP-Neu5Ac, 5 mM) for the formation of α2-6 or α2-3-sialyl linkage, respectively. It was observed that sialic acids were cleaved from sialosides during MALDI-TOF analysis. Therefore, highresolution mass spectrometry (HRMS) analysis was used to analyze the N-glycans released from RNase B sialylation products by PNGase F digestion and purified by Cotton HILIC SPE microtips [60]. We found that in situ generation of CMP-Neu5Ac from Neu5Ac and CTP by Neisseria meningitides CMP-sialic acid synthetase (NmCSS) [61] during enzymatic sialylation process in a one-pot two-enzyme system [62] further improved the efficiency of sialylation. As shown in Figure 6, the ionized species of disialyl biantennary Nglycans released from RNase B sialyation products (Neu5Ac2Gal2GlcNAc2Man3GlcNAc2, Incubating the reaction mixture of the OP5E reaction with UDP-Gal (2 mM) and MBP-∆128Bβ4GalT1-His 6 (3% w/w) at 30 • C for 2 h completed the formation of RNase B with Gal 2 GlcNAc 2 Man 3 GlcNAc 2 N-glycan ( Figure 4E).
The formation of RNase B with disialylated biantennary complex N-glycan was accomplished by incubating with MBP-∆89hST6GAL-I-His 6 (3% w/w) or MBP-CjCst-I∆145-His 6 (3% w/w) in the presence of cytidine 5 -monophosphate-Neu5Ac (CMP-Neu5Ac, 5 mM) for the formation of α2-6 or α2-3-sialyl linkage, respectively. It was observed that sialic acids were cleaved from sialosides during MALDI-TOF analysis. Therefore, high-resolution mass spectrometry (HRMS) analysis was used to analyze the N-glycans released from RNase B sialylation products by PNGase F digestion and purified by Cotton HILIC SPE microtips [60]. We found that in situ generation of CMP-Neu5Ac from Neu5Ac and CTP by Neisseria meningitides CMP-sialic acid synthetase (NmCSS) [61] during enzymatic sialylation process in a one-pot two-enzyme system [62] further improved the efficiency of sialylation. As shown in Figure 6, the ionized species of disialyl biantennary N-glycans released from RNase B sialyation products (Neu5Ac 2 Gal 2 GlcNAc 2 Man 3 GlcNAc 2 , m/z value expected 1110.3842; m/z values observed were 1110.3849 and 1110.3848 for the double-charged α2-6 and α2-3-linked species, respectively) were clearly observed. Ionized species for monosialylated N-glycan (Neu5Ac 1 Gal 2 GlcNAc 2 Man 3 GlcNAc 2 , m/z value expected: 1930.6803) and N-glycan released from the RNase B substrate for sialylation reactions (Gal 2 GlcNAc 2 Man 3 GlcNAc 2 , m/z value expected 1663.5819, or 1675.5610 for its Cl adduct) were not observed. This indicated that the di-sialylation reactions went to completion.
double-charged α2-6 and α2-3-linked species, respectively) were clearly observed. Ionized species for monosialylated N-glycan (Neu5Ac1Gal2GlcNAc2Man3GlcNAc2, m/z value expected: 1930.6803) and N-glycan released from the RNase B substrate for sialylation reactions (Gal2GlcNAc2Man3GlcNAc2, m/z value expected 1663.5819, or 1675.5610 for its Cl adduct) were not observed. This indicated that the di-sialylation reactions went to completion.

Conclusions
Using RNase B as a model glycoprotein substrate, we have successfully established a glycoprotein in vitro N-glycan processing platform for the production of glycoproteins containing homogeneous α2-6-or α2-3-linked disialylated biantennary complex N-glycans using enzymes expressed in E. coli. Several mammalian glycoprotein N-glycan processing glycosyltransferases including hGnT-I, hGnT-II, Bβ4GalT1, and hST6GAL-I have been successfully expressed in E. coli Origami B (DE3) cells as N-terminal MBP-fused and C-terminal His6-tagged fusion proteins in a soluble and active form. In addition, bacterial A B Figure 6. HRMS (negative mode) assay results for the N-glycans released from the RNase B products produced by a one-pot two-enzyme (OP2E) sialylation reaction containing Neisseria meningitidis CMP-sialic acid synthetase (NmCSS) and MBP-∆89ST6GAL-I-His 6 (A) or MBP-CjCst-I∆145-His 6 (B). The m/z value expected for the ionized disialyl N-glycan (Neu5Ac 2 Gal 2 GlcNAc 2 Man 3 GlcNAc 2 ) was 1110.3842. The ionized substrate (m/z value expected 1663.5819), substrate + Cl (m/z value expected 1675.5610), or monosialylated N-glycan (m/z value expected 1930.6803) was not observed.

Conclusions
Using RNase B as a model glycoprotein substrate, we have successfully established a glycoprotein in vitro N-glycan processing platform for the production of glycoproteins containing homogeneous α2-6or α2-3-linked disialylated biantennary complex N-glycans using enzymes expressed in E. coli. Several mammalian glycoprotein N-glycan processing glycosyltransferases including hGnT-I, hGnT-II, Bβ4GalT1, and hST6GAL-I have been successfully expressed in E. coli Origami B (DE3) cells as N-terminal MBP-fused and C-terminal His 6 -tagged fusion proteins in a soluble and active form. In addition, bacterial mannosidases including EfManI, Bt3994, and Bt1769, as well as a bacterial sialyltransferase CjCst-I have been successfully expressed as His 6 -tagged proteins. These enzymes can be easily purified using a single Ni 2+ -column. The in vitro processing of the high-mannose-type N-glycans in glycoprotein RNase B has been successfully achieved with the combination of these enzymes used in sequential step-wise reactions or in one-pot reactions. The platform developed can find broad applications for producing glycoproteins with homogeneous N-glycoforms.

Cloning
All polymerase chain reactions (PCRs) were carried out with Phusion ® HF DNA polymerase by following the standard protocol provided by the manufactory unless noted. Briefly, the reaction was performed in a reaction mixture (50 µL) containing a template (10 ng plasmid or synthetic DNA, or 1 µg of genomic DNA), 10 × Phusion ® HF buffer (5 µL), dNTP mixture (1 mM each), and 5 U (1 µL) of Phusion ® HF DNA polymerase, forward and reverse primers (1 µM each). The reaction mixture was subjected to 30 cycles of amplification. The primers and the annealing temperature (Ta) used for each PCR reaction are listed in Table 2. The PCR products were purified by GeneJET Gel Extraction Kit and digested with two restriction enzymes at 37 • C for 2 h. The digested products were purified by GeneJET Gel Extraction Kit and ligated with vector plasmid pre-digested with the same restriction enzymes and similarly gel extraction purified. The ligation was carried out at 16 • C overnight using T4 DNA ligase. The ligated product was transformed into the chemical competent E. coli DH5α cells. Plasmids were purified using GeneJET Plasmid Miniprep Kit and sequences were confirmed by DNA sequencing (See Supplementary Materials). Positive plasmids were selected and transferred to chemically competent E. coli BL21 (DE3) or Origami B (DE3) cells for expression. MBP-∆28hGnT-I-His 6 : A synthetic gene encoding an N-terminal 28 amino acid truncated human GnT-I (hGnT-I, GenBank accession number: NP_001108089.1) (∆28hGnT-I) with codon optimized for E. coli expression was cloned in pET15b vector to construct plasmid pET15b-∆28hGnT-I. PCR was performed with 5 U (1 µL) of Herculase-enhanced DNA polymerase in Herculase buffer and other conditions were the same as described above. The resulting plasmid was used as the PCR template to construct pMAL-c2X-∆28hGnT-I in pMAL-c2X vector for expressing MBP-∆28hGnT-I-His 6 .

Enzyme Expression and Purification
To express enzymes in E. coli BL21 (DE3) expression system, cells harboring the plasmid of interest were cultured in LB media (10 g/L tryptone, 5 g/L yeast extract, and 10 g/L NaCl) supplemented with ampicillin (100 µg/mL). When the OD 600 nm of the culture reached 0.6-0.8, isopropyl-1-thio-D-galactopyranoside (IPTG, 0.1 mM) was added and the culture was incubated at 20 • C for 20 h. To express enzymes in E. coli Origami B (DE3) harboring pGro7 and the plasmid of interest, cells were cultured in LB media supplemented with ampicillin (50 mg/mL), tetracycline (5 mg/L), chloramphenicol (17 mg/L), kanamycin (25 mg/L), and L-arabinose (1 g/L, for chaperon expression). When the OD 600 nm of the culture reached 0.6-0.8, IPTG (0.1 mM) was added and the culture was incubated at 16 • C for 48 h.
After the expression was completed, cells were harvested by centrifugation (6000× g) at 4 • C for 30 min and re-suspended in lysis buffer (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 0.1% Triton X-100). The cell resuspension was subjected to the homogenizer (EmulsiFlex-C3) to break the cells. The cell lysate was obtained as the supernatant after centrifugation (9016× g) at 4 • C for 60 min and purified by a Ni 2+ -NTA affinity column, such as a mini Nuvia IMAC cartridge (5 mL), on a Bio-Rad NGC system. The column was pre-equilibrated with 6 column volumes of binding buffer containing Tris-HCl buffer (50 mM, pH 8.0), NaCl (300 mM). It was washed with 10 column volumes of binding buffer, followed by washing with 10 column volumes of 10% elute buffer, and 10 column volumes of 20% elute buffer, and then eluted with elute buffer containing Tris-HCl (50 mM, pH 8.0), NaCl (300 mM), imidazole (250 mM). The fractions containing the purified protein were combined for dialysis against a dialysis buffer (Tris-HCl, 50 mM, pH 7.5, 250 mM NaCl) or for concentration using a protein concentrator (10 KDa cut-off). Finally, 20% glycerol (for MBP-∆128Bβ4GalT1-His 6 ) or 10% glycerol (for other enzymes) was added before storing the samples at −20 • C.

Stepwise Enzymatic Reaction Using RNase B as The Substrate
RNase B (5 mg/mL, 330 µM) was incubated with EfMan-I-His 6 (150 µg/mL, 1.8 µM) in Tris-HCl (100 mM, pH 7.5) at 30 • C for 2 h. The resulting mixture was then dialyzed against ddH 2 O and subjected to MALDI-TOF MS analysis. It was also used for the assays described below.
The dialyzed EfMal-I-His 6 -treated RNase B (~330 µM) was used as the acceptor substrate. It was incubated with UDP-GlcNAc (1 mM) and MBP-∆27hGnT-I-His 6 (1.6 µM) in Tris-HCl (100 mM, pH 7.5) containing MnCl 2 (2 mM) at 30 • C for 2 h. The weight ratio of MBP-∆27hGnT-I-His 6 to RNase B was 3% (w/w). The resulting mixture was then dialyzed against ddH 2 O and subjected to MALDI-TOF MS analysis. It was also used for the assays described below.
The RNase B obtained above was incubated with UDP-GlcNAc (2 mM) and MBP-∆27hGnT-II-His 6 (3.2 µM) in Tris-HCl (100 mM, pH 7.5) containing MnCl 2 (2 mM) at 30 • C for 2 h. The weight ratio of ∆27hGnT-II to RNase B was 10% (w/w). The resulting mixture was then dialyzed against ddH 2 O and subjected to MALDI-TOF MS analysis. It was also used for the assays described below.
Dialysis was not necessary for step-wise reactions, but was preferred before MALDI-TOF MS analysis of the samples.
One-pot two-enzyme (OP5E) reactions were carried out similarly to OP4E reactions except for the fact that one more enzyme MBP-∆27hGnT-II-His 6 (3.2 µM, 10% w/w) was added and the reaction was continued overnight.

MALDI-TOF MS Analyses of RNase B samples and the Released N-Glycans
Fresh solutions of 2,5-DHB (15 mg/mL) dissolved in ddH 2 O and sinapinic acid (SA) (20 mg/mL) dissolved in ACN/0.1%TFA (7:3) were prepared. A mixed solution of these two with a 1:1 (v/v) ratio was used as the matrix for MALDI-TOF MS analysis of RNase B samples (~1 mg/mL) dialyzed against ddH 2 O using Slide-A-Lyzer™ MINI Dialysis (10 k MWCO) devices.
To release N-glycans from RNase B samples, RNase B (1 mg) was denatured by adding 0.5% SDS and dithiothreitol (DTT) (40 mM) in 200 µL followed by incubation at 98 • C for 10 min and then at room temperature for 5 min. PNGase F (100 ng) was then added and the glycans were released by incubation at 37 • C for 2 h. The deglycosylated proteins were precipitated by adding three volumes of pre-chilled ethanol followed by incubation on ice for 20 min. The mixtures were then centrifuged at 16,200× g for 5 min and the supernatants containing the glycans were purified with graphitized carbon cartridge and dried in a speed vacuum. They were dissolved in ddH 2 O and used for MALDI-TOF MS analysis using 2,5-DHB dissolved in ACN/0.1%TFA (7:3) (25 mg/mL) as the matrix.

HRMS Analysis of Sialylated N-Glycans
RNase B (100 µg) samples were denatured and the N-glycans were released by treating with PNGase F (10 ng) similar to the conditions described above. The samples were cleaned using a homemade cotton tip via hydrophilic interaction liquid chromatography-solid phase extraction (HILIC-SPE) [60]. Briefly, samples were mixed with acetonitrile (ACN, 50% v/v) centrifuged (16,200× g) at 4 • C for 5 min. The supernatant was transferred to a clean tube, pipetting up-and-down for a total of 20 times in a 10 µL-tip packed with a small volume of cotton. The cotton tip was washed 3 times with 20 µL of 85% ACN with 1% TFA, followed by 3 times with 20 µL of 85% ACN, and eluted with ddH 2 O (10 µL). The eluant was used for HRMS analysis.