Improvement of the Transglycosylation Efficiency of a Lacto-N-Biosidase from Bifidobacterium bifidum by Protein Engineering

Abstract

The lacto-N-biosidase LnbB from Bifidobacterium bifidum JCM 1254 was engineered to improve its negligible transglycosylation efficiency with the purpose of enzymatically synthesizing lacto-N-tetraose (LNT; Gal-β1,3-GlcNAc-β1,3-Gal-β1,4-Glc) in one enzymatic step. LNT is a prebiotic human milk oligosaccharide in itself and constitutes the structural core of a range of more complex human milk oligosaccharides as well. Thirteen different LnbB variants were expressed and screened for transglycosylation activity by monitoring transglycosylation product formation using lacto-N-biose 1,2-oxazoline as donor substrate and lactose as acceptor substrate. LNT was the major reaction product, yet careful reaction analysis revealed the formation of three additional LNT isomers, which we identified to have a β1,2-linkage, a β1,6-linkage, and a 1,1-linkage, respectively, between lacto-N-biose (Gal-β1,3-GlcNAc) and lactose. Considering both maximal transglycosylation yield and regioselectivity as well as minimal product hydrolysis, the best variant was LnbB W394H, closely followed by W465H and Y419N. A high transglycosylation yield was also obtained with W394F, yet the substitution of W394 and W465 of the subsite −1 hydrophobic platform in the enzyme with His dramatically impaired the undesirable product hydrolysis as compared to substitution with Phe; the effect was most pronounced for W465. Using p-nitrophenyl-β-lacto-N-bioside as donor substrate manifested W394 as an important target position. The optimization of the substrate concentrations confirmed that high initial substrate concentration and high acceptor-to-donor ratio both favor transglycosylation.

1. Introduction

Lacto-N-biosidases (LNBases; EC 3.2.1.140) specifically catalyze the hydrolytic release of the disaccharide lacto-N-biose (LNB; Gal-β1,3-GlcNAc) from the nonreducing end of glycans, e.g., from the abundant type I human milk oligosaccharide (HMO) lacto-N-tetraose (LNT; Gal-β1,3-GlcNAc-β1,3-Gal-β1,4-Glc) [1,2,3]. In the CAZy database of carbohydrate-active enzymes (CAZy) [4], LNBases are found in glycoside hydrolase (GH) families 20 (GH20) and 136 (GH136). The LNBases of GH136 employ a classical double-displacement mechanism possibly including water-mediated proton transfer [5], whereas the GH20 LNBases employ a substrate-assisted mechanism, where the 2-acetamido group of the substrate (typically of a GlcNAc residue in the substrate molecules) acts as a nucleophile [6,7,8,9]. This mechanism is usually (but not always) employed in enzymes active on GlcNAc- or GalNAc-containing substrates and operates through an oxazoline or oxazolinium ion intermediate, which—unlike the case for the retaining GHs that work via a classical double-displacement mechanism—is not covalently linked to the enzyme [6,7,8,9,10,11].

The ability of LNBases to release and transfer disaccharide moieties in a single reaction step is an attractive trait for targeted transglycosylation reactions, e.g., for the synthesis of HMOs, which are crucial for infant health and development, but absent from infant formula [12]. LNT is the most important HMO core structure as well as being among the most abundant HMOs in itself [13]. Two examples of the use of LNBases for transglycosylation exist in the literature, although only low transglycosylation yields were achieved: In the first example from 1999, a GH20 LNBase isolated from Aureobacterium sp. L-101 was used for catalyzing the formation of LNT from p-nitrophenyl-β-lacto-N-bioside (pNP-LNB) and lactose [14] (Figure 1B). In the other example, the GH20 LNBase LnbB from Bifidobacterium bifidum JCM 1254 was shown to catalyze transglycosylation with pNP-LNB and lactose to form LNT at low yields [3] (Figure 1B); LnbB also catalyzed the transfer of a lacto-N-bioside unit from LNT to 1-alkanols (C₁–C₄) as well as a condensation reaction with LNB and lactose to form LNT and an isomer [3].

The fact that GH20 enzymes employ an oxazoline intermediate has recently been exploited for the synthesis of LNT [15] (Figure 1A) as well as for the synthesis of the HMO precursor lacto-N-triose II (LNT2; GlcNAc-β1,3-Gal-β1,4-Glc) [16,17,18]. To ensure oxazoline stability, these reactions require enzymes able to work at or preferably above neutral pH [8].

The ability to easily synthesize and use the reaction intermediate as donor substrate facilitates the use of glycosynthases, i.e., GHs where the catalytic nucleophile is replaced by a nonfunctional residue to suppress the inherent hydrolytic activity of the enzyme and thus favor transglycosylation [8,17,19]. To construct glycosynthase variants of LnbB, two different substrate-interacting residues were targeted previously [15]. The polarizing Asp residue of the catalytic Asp–Glu pair (D320-E321), which apparently stabilizes the positive charge development in the oxazolinium ion transition state [7,20], was substituted by Glu or Ala. Alternatively, a Tyr residue (Y419), which stabilizes the reaction intermediate and/or binds water or an alternative acceptor substrate through hydrogen bonding [7,21,22], was substituted by Phe to remove its hydrogen bonding capacity. This mutation was first shown to improve transglycosylation for chito-oligosaccharide synthesis in a GH20 β-N-acetylhexosaminidase from Talaromyces flavus; the substitution to Asn inspired by the closely related GH84 β-N-acetylglucosaminidases also improved transglycosylation performance [22]. In the reaction with pNP-LNB and lactose, all mutants of LnbB led to slightly increased LNT yields (10–13%) compared to the WT (8%). However, in the reaction with lacto-N-biose 1,2-oxazoline (LNB-oxazoline) and lactose, the highest LNT yields were obtained with the WT (67%) [15].

The combined use of glycosynthase variants and oxazoline substrates has proven particularly useful for the synthesis of LNT2 from N-acetyl-D-glucosamine 1,2-oxazoline and lactose [17,23], whereas LNT synthesis suffered from enzymatic as well as spontaneous hydrolysis of LNB-oxazoline substrate when using the LNBase LnbB from Bifidobacterium bifidum JCM 1254 [15]. Obviously, the use of the reaction intermediate—the oxazoline substrate derivative—as donor substrate is not limited to glycosynthase variants, as also shown in the very first report on the use of an oxazoline donor substrate in transglycosylation [24]. Indeed, successful transglycosylation with oxazoline donor substrates has been reported for wild-type GH20 enzymes as well as for variants subjected to other transglycosylation-improving engineering strategies such as loop engineering [15,16,17,18]. In the current work, with the purpose of maximizing transglycosylation efficiency of LnbB (GenBank ABZ78855.1), we designed a set of 12 single amino acid mutations and one deletion mutant of LnbB following several different protein engineering strategies (Figure 2). The 13 LnbB variants were individually expressed heterologously in Escherichia coli BL21 (DE3), and their ability to catalyze LNT formation through transglycosylation was compared in a reaction system of LNB-oxazoline and lactose. For the best-performing variants, the formation of transglycosylation products was studied in more detail in a more classical reaction system comprising pNP-LNB as donor substrate and lactose as acceptor substrate.

Castejón-Vilatersana et al. [9] recently presented a detailed study of transglycosylation by 23 different LnbB variants targeting engineering of substrate-interacting residues. In this work, we present nine additional variants and show further improvement by alternative amino acid substitutions. We also compare LNB-oxazoline and pNP-LNB as donor substrates for the best LnbB variants. Finally, we question the previous statements [9,15] that LnbB-catalyzed transglycosylation is regioselective.

2. Materials and Methods

2.1. Chemicals

2-Chloro-1,3-dimethylimidazolinium chloride, p-nitrophenyl-β-lacto-N-bioside (pNP-LNB), and lacto-N-triose (LNT2) were purchased from Carbosynth (Compton, UK). The lacto-N-biose (LNB) and the lacto-N-tetraose (LNT) standards were kindly provided by Glycom A/S (Hørsholm, Denmark). Triethylamine (≥99.5%) was purchased from Carl Roth (Karlsruhe, Germany). Lactose and all other chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA).

2.2. In Silico Analyses

All characterized GH20 sequences available in the CAZy database (www.cazy.org [4] 126 in total, accessed on 17 June 2019) were aligned by MUSCLE 3.8 [29] to identify conserved residues. The CUPP webserver (www.cupp.info [30], accessed on 25 June 2021) was used to group the GH20 sequences which are predicted to be functionally similar to the two known LNBases from B. bifidum [3] and Streptomyces sp. [1,2]; the 46 sequences in CUPP group GH20:7.1 were then aligned with MUSCLE 3.8 to identify conserved residues among the LNBases of the Actinobacteria class.

The two-dimensional LNB–LnbB interaction diagram has been generated by LigPlot+ version 2.1 [31] using the PDB entry 4H04 of LnbB in complex with LNB. Three-dimensional structures of LnbB were visualized in The PyMOL Molecular Graphics System, version 2.3.3 (Schrödinger, LLC, New York, NY, USA) using crystal structures of LnbB in complex with LNB (PDB 4H04) and LNB-thiazoline (PDB 4JAW) [7], respectively.

2.3. Protein Engineering and Recombinant Expression of LnbB in E. coli

The amino acid sequence for LnbB (GenBank ABZ78855.1) was checked for presence of a signal peptide using the SignalP 5.0 webserver [32] (signal peptide predicted for 1–35). In addition, the sequence was checked for presence of transmembrane helices using the TMHMM 2.0 webserver [33] (C-terminal membrane anchor predicted at 1087-1106). As a result, only the sequence 37-1064 was selected for expression, which is in accordance with the LnbB originally expressed and characterized (35-1064) [3]. The sequence (37-1064) was codon-optimized for E. coli expression and inserted into a pET-24b(+) expression vector by GenScript (Piscataway, NJ, USA) using restriction sites NdeI and XhoI to add a C-terminal His₆-tag.

All mutants were constructed using CloneAmp^TM polymerase (Takara, Kusatsu, Japan), a set of mutagenic primers listed in Table S1 (Supplementary Materials), and the pET-24b(+)/LnbB (37-1064) plasmid as template. The plasmid template was then digested with DpnI at 37 °C overnight, and the PCR products were purified using the Illustra GFX PCR DNA and Gel Band Purification Kit (GE Healthcare Life Sciences, Chicago, IL, USA). Competent E. coli DH5α cells were then transformed with the purified PCR products and plated on LB agar plates supplemented with kanamycin (50 µg/mL). Positive transformants were selected, and corresponding plasmids were extracted using GeneJET Plasmid Miniprep Kit (ThermoFisher Scientific, Waltham, MA, USA). All constructs were checked by sequencing (Macrogen Europe, Amsterdam, The Netherlands).

E. coli BL21 (DE3) cells were used for protein production: competent E. coli BL21 (DE3) cells were transformed with the plasmids and selected for kanamycin resistance. Overnight precultures of transformed cells grown at 37 °C in LB medium containing 50 μg/mL kanamycin were used to inoculate 500 mL LB medium supplemented with 50 μg/mL kanamycin and incubated at 37 °C. When the OD₆₀₀ of each culture reached a value of 0.6–0.8, gene expression was induced using 1 mM IPTG for 20 h at 25 °C. Cells were harvested by centrifugation (5000× g, 15 min), and the pellet was resuspended in binding buffer (20 mM sodium phosphate buffer, 500 mM NaCl, 20 mM imidazole, pH 7.4) before sonication using an UP 400S Ultrasonic processor (Hielscher Ultrasonics, Teltow, Germany). Cell debris was removed by centrifugation (20,000× g, 30 min at 4 °C).

Proteins were purified on Ni²⁺ Sepharose columns (GE Healthcare, Chicago, IL, USA), previously equilibrated with binding buffer. His₆-tagged proteins were eluted using 250 mM of imidazole in binding buffer. Imidazole was then removed using PD-10 columns (GE Healthcare) according to the supplier’s recommendations; the resulting protein buffer was 20 mM sodium phosphate with 100 mM NaCl at pH 7.4. Proteins were visualized on SDS-PAGE (Biorad, Hercules, CA, USA). Protein concentrations were determined at an absorbance of 280 nm using an LnbB extinction coefficient of 183,370 M⁻¹ cm⁻¹ and molecular weight of 111,068 g/mol, as predicted by the Expasy web server (https://web.expasy.org/protparam/ accessed on 1 November 2021).

2.4. Synthesis of LNT2 Isomers

Please refer to Section S1 of the Supplementary Materials for a detailed description. The LNT2 isomer 2′GlcNAc-Lac (GlcNAc-β1,2-Gal-β1,4-Glc; 2′LNT2) was chemically synthesized from 6′O-[4-chlorobenzoyl]-protected lactose acetonide and peracetylated N-trichloroacetyl-D-glucosaminyl thiophenyl donor using N-bromosuccinimide (NBS), trifluoromethanesulfonic acid (TfOH), and dichloromethane (DCM), followed by Zemplen-deprotection, N-acetylation, and acidic deprotection [34,35] (Scheme S1). The structure was verified by ¹H and ¹³C NMR (Section S1); all NMR spectra were recorded on an 800 MHz Bruker Avance III (Bruker Daltonics GmbH, Bremen, Germany) essentially as described previously [18]. The LNT2 isomer 4′GlcNAc-Lac (GlcNAc-β1,4-Gal-β1,4-Glc; 4′LNT2) was chemically synthesized as a 1:1 mixture with LNT2 (GlcNAc-β1,3-Gal-β1,4-Glc) from a p-chlorobenzoyl protected benzyl lactoside acceptor with free 3′-OH and 4′-OH and peracetylated N-trichloroacetyl-D-glucosaminyl thiophenyl donor using the same procedure as for 2′GlcNAc-Lac except for the final hydrogenation step (Pd/C) [34,35] (Scheme S2). ¹H and ¹³C NMR confirmed the structure of the mixture LNT2 and 4′GlcNAc-Lac (Section S1).

The LNT2 isomer 6′GlcNAc-Lac (GlcNAc-β1,6-Gal-β1,4-Glc; 6′LNT2) was chemically synthesized from a pivaloyl-protected benzyl lactoside acceptor with free 4′OH and 6′OH and an oxazoline donor using trimethylsilyl trifluoromethanesulfonate (TMSOTf) and 1,2-dichloroethane (DCE) followed by Zemplen-deprotection and hydrogenation [34,36,37,38] (Scheme S3). The structure was verified by ¹H and ¹³C NMR (Section S1).

2.5. Synthesis of LNB-Oxazoline

Lacto-N-biose 1,2-oxazoline (LNB-oxazoline) was prepared as described previously [16] utilizing 2-chloro-1,2-dimethylimidazolinium chloride as reactant and triethylamine as HCl scavenger as originally suggested by Shoda and co-workers [39]: a total of 96 mg lacto-N-biose (LNB) and 312 μL triethylamine was solubilized in 688 μL dH₂O in a round-bottom flask. The solution was kept at 0–4 °C under magnetic stirring. The synthesis was initiated by addition of 127 mg 2-chloro-1,2-dimethylimidazolinium chloride while stirring continuously for 15 min. The solution was pH-adjusted to approx. pH 8.0 by drop-wise addition of HCl (2 M). The prepared LNB-oxazoline was used for enzymatic reaction immediately after preparation. The presence of LNB-oxazoline was confirmed by LC-ESI-MS.

2.6. Screening of LnbB Variants by Enzymatic Transglycosylation with LNB-Oxazoline and Lactose

The enzymatic reaction was performed with 50 μL of resulting, neutralized LNB-oxazoline reaction solution, which was added to 42.8 mg of lactose in a 2 mL reaction tube kept at 37 °C in a thermomixer (950 rpm). The reaction was started by addition of 150 μL 20 mM sodium phosphate buffer (pH 7.4; above optimal pH [3] to increase oxazoline stability) containing 1.33 μM LnbB (resulting concentration 1 μM). Duplicate reactions were monitored for up to 24 h by inactivating reaction samples by heating at 90 °C for 10 min after 15 min and 3 and 24 h of reaction. The reaction samples were analyzed by LC-ESI-MS.

2.7. Enzymatic Transglycosylation with pNP-LNB and Lactose for Selected LnbB Variants

Transglycosylation reactions with LnbB WT and five selected mutants (W394F, W394H, Y419N, W465H and W465F) were performed using 2.5 mM pNP-LNB as donor and 200 mM lactose acceptor at pH 4.5 and 30 °C (optimal pH for pNP-LNB [3] and the same conditions as used by Castejón-Vilatersana et al. [9]). For optimization of the reaction conditions with pNP-LNB donor and lactose acceptor for LnbB W394H, transglycosylation reactions were performed using varying concentrations of pNP-LNB as donor (from 2.5 to 10 mM) and varying concentrations of lactose acceptor (200 or 400 mM). Duplicate reactions were monitored for up to 26 h by inactivating reaction samples by heating at 90 °C for 10 min after 5 and 15 min and 2, 24, or 26 h of reaction. The reaction samples were analyzed by LC-ESI-MS.

2.8. Determination of Melting Temperature T_m

The thermal stabilities of LnbB WT and five selected mutants (W394F, W394H, Y419N, W465F and W465H) were described by the thermal unfolding transition midpoint, T_m in °C, at which half of the protein population is unfolded. The T_m of the proteins (prepared at a final concentration of 3 µM) were determined from the maximum of the first derivative of the F_330nm/F_350nm curve (ratio of Trp emission at 330 and 350 nm) using a Prometheus NanoTemper Penta instrument (NanoTemper Technologies, Munich, Germany).

2.9. Analysis of Transglycosylation Reactions

Identification and quantitation of LNT transglycosylation products obtained from LNB-oxazoline and pNP-LNB reactions were performed by LC-ESI-MS on an Amazon SL iontrap (Bruker Daltonics, Bremen, Germany) coupled to an UltiMate 3000 UHPLC from Dionex (Sunnyvale, CA, USA) equipped with a porous graphitized carbon column (Hypercarb PGC, 150 mm × 2.1 mm, 3 μm; Thermo Fisher Scientific, Waltham, MA, USA) as described previously [40], using a target mass of 550 m/z. Quantitation was performed in Compass TASQ 2.2 (Bruker Daltonics) using LNT as external calibration standard. The concentrations of LNT isomers were expressed as LNT equivalents.

In addition, the analytical method used by Castejón-Vilatersana et al. [9] was also applied: 3 µL of a reaction sample in isopropanol:water (1:1) was injected on a BEH Amide column (Waters; 130 Å, 1.7 μm, 2.1 × 100 mm) with precolumn using isocratic elution in acetonitrile/water (65:35), 1% formic acid at 40 °C at a flow rate of 0.2 mL/min. Detection of the reaction products was performed as described above.

2.10. Preparative Purification of LNT Isomers

Isolation of individual LNT isomers was performed according to the analysis procedure with an injection volume of 50 µL and manual fractionation while bypassing the MS detector. To obtain a quasiequimolar composition of the isomers, two reactions were combined in a 1:1 ratio (i.e., samples after 2 and 24 h of reaction for W394H and W465H, respectively), resulting in concentrations of 12–39 µM of each isomer. Fractions were dried down at 40 °C under a stream of N₂. Each fraction was redissolved in 50 µL MilliQ water.

2.11. Structural Elucidation of LNT Isomers

The GH35 β1,3-galactosidase BgaC from B. circulans ATCC 31382 (GenBank BAA21669.1) [41,42] was expressed and purified essentially as described for LnbB above (Figure S3). The full-length gene was codon-optimized, synthesized, and inserted into a pJ411 vector by DNA 2.0 (Menlo Park, CA, USA) to result in an enzyme with an N-terminal His₆-tag followed by a thrombin cleavage site. BgaC is specific for β1,3-linked Gal [42] and was employed to degrade the LNT isomers as part of the structural elucidation: By BgaC-catalyzed removal of the terminal Gal moiety in LNB, the LNT isomers became LNT2 isomers (Figure S2), which are simpler to synthesize chemically for use as external identification standards than LNT isomers. GH35-catalyzed degradation of the LNT isomers was performed using a final concentration of 4.3 µM enzyme, 8–28 µM isomer, and 10 mM acetate buffer pH 5 and incubated at 30 °C for 15 h. The digested samples were diluted 1:1 in MilliQ water and reanalyzed by LC-ESI-MS as described above. The resulting product peaks were compared to those of LNT2 and chemically synthesized 2′-GlcNAc-Lac, 4′-GlcNAc-Lac, and 6′-GlcNAc-Lac external standards for structural elucidation.

2.12. Reduction

In order to investigate the possible presence of LNB-1,1-Lac, selected reaction samples were reduced according to Bao et al. [43] with modifications: The isomers were reduced to their alditols by adding 10 µL freshly made 0.25 M aqueous NaBH₄ to 10 µL of reaction containing the isomers, followed by incubation at room temperature for 15 h. The reduction was terminated by adding 15 µL 0.2 M acetic acid. No further sample preparation prior to analysis.

2.13. Statistics

One-way ANOVA for determination of statistical significance was carried out in JMP Pro 14.1.0 (SAS Institute Inc., Cary, NC, USA). Statistical significance was established at p < 0.05.

3. Results and Discussion

3.1. Screening of LnbB Variants

Thirteen LnbB variants were designed and successfully expressed recombinantly in E. coli BL21 (DE3) (Figure S3): N259T, D320E, D320T, W373F, W373H, W394F, W394H, G398R, ΔG398, Y419N, W465F, W465H, and D467N (Figure 2 and Figure 3). Of these, D320E, W373F, W394F, and W465F have been studied elsewhere [9,15], while the remaining nine variants are new and provide a foundation for comparing the effect of choosing different amino acid substitutions at N259, D320, W373, W394, Y419, W465, and D467. N259, Q190, and D467 essentially define subsite −2 (the Gal binding site), D320 interacts with the GlcNAc residue on one side of subsite −1, while Y419 and the three Trp residues are situated around subsite −1 (Figure 3). G398 is approximately 13 Å away from the ligand; this distance reduces to approximately 10 Å if mutated to Arg in PyMOL (Figure S3).

The transglycosylation abilities of each of the variants and the WT enzyme were screened using LNB-oxazoline as donor substrate, lactose as acceptor substrate, and the same enzyme dosage for all variants (1 µM). The formation of LNT and isomers was monitored over 24 h (

Table 1. Maximum yields of LNT and of LNT plus isomers (listed as total transglycosylation products; TG) from transglycosylation reactions by LnbB WT and variants, using LNB-oxazoline as donor and lactose as acceptor and 1 µM of enzyme at pH 7.4 and 37 °C. Superscript letters (^a–i) indicate statistically significant difference between the values within the column (p < 0.05). The time at which the maximum transglycosylation yields are reached is indicated (samples taken after 0.25, 3, and 24 h of reaction). Secondary hydrolysis is indicated as the percentage of LNT, which was hydrolyzed after 24 h of reaction (calculated as (max. LNT yield − LNT yield at 24 h]/[max. LNT yield)). An asterisk (*) indicates that the LNT concentration after 24 h is not significantly different from the maximum yield (secondary hydrolysis not statistically significant; p < 0.05). Regioselectivity is indicated by the percentage of total transglycosylation products (TG), which is identified as LNT. For each column, the best results are marked in bold, whereas the second best results are underlined.

Variant	Max. LNT [μM]	Max. TG [μM]	Time of Max. Yield [h]	% LNT Hydrolyzed after 24 h	% LNT of Total TG
WT	495 ± 14 e	854 ± 7 g	0.25	89 ± 0.2%	58 ± 1% e
N259T	754 ± 32 cd	1226 ± 26 e	3	66 ± 1%	62 ± 1% d
D320E	185 ± 9 f	685 ± 51 h	24	-	27 ± 1% i
D320T	303 ± 5 f	798 ± 2 gh	24	-	38 ± 1% h
W373F	839 ± 14 c	1192 ± 14 e	0.25	92 ± 2%	70 ± 0.3% ab
W373H	696 ± 30 d	1020 ± 42 f	0.25	90 ± 2%	68 ± 0.1% b
W394F	1279 ± 17 a	1865 ± 39 b	3	26 ± 2%	69 ± 1% b
W394H	1189 ± 6 a	1655 ± 9 c	3	6 ± 6% *	72 ± 0.01% a
G398R	750 ± 106 cd	1183 ± 145 e	0.25	91 ± 3%	63 ± 1% cd
ΔG398	720 ± 122 d	1169 ± 134 ef	0.25	91 ± 2%	61 ± 3% d
Y419N	1285 ± 5 a	2388 ± 12 a	3	14 ± 6% *	54 ± 0.1% f
W465F	1007 ± 9 b	1467 ± 8 d	0.25	93 ± 2%	69 ± 0.3% b
W465H	1235 ± 7 a	1930 ± 24 b	3	8 ± 4% *	64 ± 0.4% c
D467N	536 ± 3 e	1125 ± 1 ef	3	15 ± 0.4%	48 ± 0.2% g

; Figure 4 and Figure S4). In all reactions, LNT was formed as the major product along with three different isomers (see Section 3.2). The different mutants showed maximum LNT yields ranging from 185 to 1285 µM, whereas the maximum yield of total transglycosylation products (LNT plus isomers) ranged from 685 to 2388 µM (Table 1). For WT LnbB, a maximum LNT yield of 495 µM was measured after only 15 min of reaction. However, the formed product was quickly hydrolyzed by secondary hydrolysis, and after 24 h of reaction, only 56 µM LNT remained (Table 1). The following sections examine the variants based on design strategy.

3.1.1. Glycosynthase Variants: D320E, D320T, and Y419N

The synthesis of LNT was significantly reduced for the variants D320E and D320T, i.e., mutations targeting the assisting catalytic Asp residue (Figure 2). Both mutations thus also resulted in very low secondary hydrolysis of the formed LNT (Figure S4). D320T was designed based on recent work on the β-N-acetylhexosaminidase BbhI from B. bifidum JCM 1254 [27]. In that work, BbhI was subjected to directed evolution, which revealed the polarizing Asp residue (D746 in BbhI) and W805 as important mutation targets to improve transglycosylation. Site-saturation mutagenesis of these two residues indicated D746T and W805R as the best variants with D746T exhibiting a transglycosylation yield of 85% vs. 45% for the WT and low secondary hydrolysis [27]. However, to reach this yield, the authors used high enzyme concentration of 41 µM for the variant. Similarly, the mutation of the assisting Asp residue to Glu has previously had a large, positive effect on BbhI using LNB-oxazoline as donor substrate [17], but this effect has not been transferable to LnbB due to both enzymatic and spontaneous LNB-oxazoline hydrolysis [15]. When using pNP-LNB as donor substrate, only the moderate improvement of transglycosylation was previously observed for D320 variants [9,15].

By contrast, the variant Y419N targeting the oxazoline-stabilizing Tyr residue showed the highest LNT yield (1285 µM; obtained after 3 h of reaction) as well as the highest total transglycosylation yield (2388 µM). The secondary hydrolysis of the LNT was negligible, and only 14% of the maximum LNT yield was hydrolyzed after 24 h of reaction (Table 1; Figure 4). However, regioselectivity was remarkably low for Y419N compared to the other best variants (below), as only 54% of the formed product was LNT (Table 1). Contrary to our findings with Y419N, the mutant Y419F previously showed high secondary hydrolysis of the formed LNT in the reaction with LNB-oxazoline and lactose and lower yield than the WT [15]. Again, the choice of amino acid substitution may play an important role, especially for secondary hydrolysis. Mutation to Phe removes the hydrogen bonding capacity but maintains the rest of the residues structure. The Asn amine group has hydrogen bonding capacity, yet Asn is smaller; thus the distance to the substrate increases (approximately 4.5 Å to the LNB-thiazoline sulfur atom in PDB 4JAW), making hydrogen bonding less likely. Possibly, the increased distance destabilizes the transition state sufficiently to impair hydrolysis, while the hydrogen bonding capacity retains enough of the original role of the Tyr residue to foster transglycosylation.

3.1.2. Conserved Residues: W373F, W373H, W394F, W394H, W465F, W465H, and D467N

LNT formation was not significantly improved for the D467N variant (Table 1; Figure S4), which is in line with the observations on the equivalent BbhI variant D884N [26]. The total transglycosylation yield did improve compared to the WT, but regioselectivity was low (only half of the formed product was LNT; Table 1). Substitution of D467 in LnbB with Ala or Glu significantly reduced the hydrolytic activity, but did not improve transglycosylation dramatically [9]. D467 is conserved across GH20, and was selected for conservative mutation, i.e. substitution with structurally similar residues, to decrease transition state stabilization following the recently elaborated semi-generic sequence-based engineering approach [26].

Transglycosylation activity was significantly improved for the variants W394F, W394H, W465F, and W465H (Table 1). These Trp variants also exhibited preference for the formation of LNT (68–72% of the formed product was LNT). The variant W465F gave a maximum LNT yield of 1007 µM after 15 min of reaction. Unfortunately, the product was rapidly hydrolyzed, and only 7% of the formed LNT remained after 24 h of reaction (Table 1; Figure 4). The other variant of this position, W465H, showed a maximum LNT yield of 1235 µM after 3 h of reaction. Here, the secondary hydrolysis was statistically insignificant, and only 8% of the formed LNT was hydrolyzed after 24 h of reaction. The equivalent mutation in BbhI, W882H, dramatically improved the transglycosylation yield to 66% (compared to 16% for the WT), yet the BbhI W882H mutant suffered from a significant loss of activity and was dosed nine times higher than the WT in that study [26].

Recently, W394F was determined as the best among 23 LnbB variants with a maximum LNT yield of 32% using 2.5 mM pNP-LNB as donor and 200 mM lactose as acceptor [9]. Interestingly, the mutation of the equivalent position in BbhI (W801H) did not affect the transglycosylation yield [26]. In the current work, the mutants W394F and W394H reached maximum LNT yields of 1279 and 1189 µM, respectively, both after 3 h of reaction. Considering both transglycosylation yield, secondary hydrolysis, and regioselectivity, W394H was the best-performing mutant in the variant screening on LNB-oxazoline and lactose (Table 1).

Introducing conservative mutations of Trp residues involves a dilemma: should one aim to maintain the aromatic nature (mutation to Phe) or the hydrogen bonding capacity (mutation to His)? Some studies follow the Trp-to-Phe strategy [9,44], while others suggest Trp-to-His [26]. We included both substitutions. For both W394 and W465, the mutation of the Trp residue to a His appears to significantly reduce the LNT secondary hydrolysis (Table 1). For W465, the mutation to Phe retains the hydrophobic stacking with the GlcNAc residue, whereas the mutation to His enables hydrogen bonding to the hydroxyl groups on C4 and C6 or the ring oxygen. W394 interacts with the oxazoline ring, and the mutation in this position therefore impacts the interaction with this intermediate in the transition state. In BbhI, W882H (corresponding to W465H in LnbB) gave the highest transglycosylation yield, but in this case, the secondary hydrolysis was still pronounced [26]. BbhI W882F has not been investigated. While the substitution of Trp to His appears superior in the case of W364 and W465 in LnbB acting on LNB-oxazoline and lactose, more data are required to confirm whether this is a general trend. Indeed, for W373F and W373H, where the improvement of LNT yield was more moderate (839 and 696 µM, respectively) and secondary hydrolysis just as pronounced as for WT LnbB, the substitution of Trp to Phe caused a slight, yet significant, increase in transglycosylation yield compared to substitution to His. Presumably, stacking interactions are important in this position. The less pronounced effect of mutation of W373 compared to that of W394 and W465 is in line with the study of Castejón-Vilatersana et al. [9].

GH20 β-N-acetylhexosaminidases and GH84 O-GlcNAcases are known to exhibit tight hydrophobic pockets around subsite −1, where several Trp residues accurately position the substrate’s acetamido group near the anomeric carbon and possibly also act to stabilize the transition state oxazoline charge development through orbital interactions [45,46]. Indeed, the three conserved Trp residues targeted for engineering line subsite −1 of LnbB (Figure 3). Conserved Trp residues also ensure substrate binding in GH18 chitinases [47,48]. While the number of conserved Trp residues in the active site may be particularly high in GH20s and GH84s due to their role in oxazolinium ion stabilization [45], Trp residues are common at GH binding sites due to their hydrophobic nature and carbohydrate stacking abilities. Hence, targeting substrate-interacting Trp residues appears to be a viable strategy for improving transglycosylation across numerous GH families. In a recent study, the mutation of conserved Trp residues at subsite −1 to His improved transglycosylation yields in a GH2 β-mannosidase, two GH29 α-L-fucosidases, a GH10 β-endo-xylanase, and in BbhI of GH20 [26]. The interference with substrate binding invites the use of increased enzyme dosages to compensate for the loss of catalytic efficiency [26]. However, the results on LnbB (Table 1 and [9]) underline that this type of mutants can outcompete the WT and other variants even at equimolar dosages.

3.1.3. Residues Conserved in LNBases Only: N259T

Selecting conserved LNBase residues for mutation from alignment across the entire GH20 is complicated by the fact that LNBases are a subclass with a highly defined subsite −2 not found in the β-N-acetylhexosaminidases [7], which make up the majority of the glycoside hydrolases belonging to GH20 [4]. Identifying amino acids conserved specifically among the LNBases is furthermore hampered by the fact that only two LNBases have been characterized. To aid the identification of conserved residues in LNBases, the CUPP online platform (cupp.info, accessed on 25 June 2021) [30] was used to identify proteins predicted to be similar in function to the known LNBases by peptide-based functional annotation. Thus, a multiple sequence alignment was prepared of all sequences in CUPP group GH20:7.1. This group comprises the two known LNBases and several other protein sequences all originating from the Actinobacteria class. This alignment identified >60 fully conserved residues. Zooming in on those close to the active site, the following interesting residues were identified around subsite −2: E257, N259, H263 (fully conserved), E215, P466, and L574 (conserved across all sequences except one). Of these, N259 and E216 were closest to the LNB-thiazoline ligand of PDB 4JAW [7] (≤3.0 Å) (Figure 3). N259 was selected for conservative mutation to Thr. LnbB N259T showed a maximum LNT yield of 754 µM, which is a significant improvement compared to the WT, but the level of secondary hydrolysis was high (66% of the LNT hydrolyzed after 24 h; Table 1). Recently, other mutations of N259 (N259A and N259Q) dramatically improved the transglycosylation ability of LnbB in the reaction with pNP-LNB and lactose [9], again emphasizing the importance of amino acid selection in rational enzyme design. In the WT, N259 hydrogen bonds with O2 and O3 of Gal in subsite −2 (Figure 3). The mutation to Thr somewhat retains this capacity, but the distance to the ligand increases from approximately 3 to approximately 4 Å, thereby likely decreasing the transition state stabilization. The mutation to Ala completely removes the interaction, whereas mutation to the much larger Gln may either sterically disturb or improve the enzyme–ligand interaction in this position. In the recently published study on LnbB engineering [9], several subsite −2 mutants (H263A, H263R, N259Q, and Q190L) exhibited markedly improved transglycosylation yields (15–21%) as a result of adequately reduced hydrolytic activity, yet the best variant (32% transglycosylation yield) had the subsite −1 targeting mutation W394F.

3.1.4. The Arg Residue at the Edge of the Active Site: G398R and ΔG398

In the study on BbhI combining directed evolution and site-directed mutagenesis, the mutant W805R gave an 82% yield vs. 45% obtained with the WT [27]. However, unlike the polarizing Asp—the other superior mutant of that study—W805 of BbhI is not conserved across GH20s. Considering the effect of W805R in BbhI [27] and the repositioning of an Arg residue in a similar location caused by the introduction of a five amino acid loop directly upstream of R355 in soil metagenome HEX1 [28], we recently proposed that the presence of an Arg in this position could be a new strategy for improving GH20 transglycosylation [8]. LnbB has a Gly residue (G398) in the position corresponding to W805 in the multiple sequence alignment (Figure 2), followed by an Arg residue (R399). To investigate the role of an Arg residue in this position in LnbB, we assessed two different variants: G398R and ΔG398. Both mutants exhibited a similarly mild improvement in LNT formation as N259T, W373F, and W373H (Table 1), indicating that an introduction of Arg in this position approximately 10 Å from the active site has a positive effect on transglycosylation (Figure S4).

3.2. Regioselectivity in Transglycosylation

Previous studies on LnbB-catalyzed transglycosylation reported regioselective LNT formation detected by UV [15] or MS [9] using HILIC separation. However, this is quite in contrast to the data obtained in the current study: during the screening of LnbB variants, four different transglycosylation products were observed with masses corresponding to hex₃hexNAc₁ (m/z 706 corresponding to the deprotonated [M−H]⁻) using a porous graphitized column and negative mode as detection (Figure 5). The same isomers were observed in the reactions with pNP-LNB as donor substrate.

When the same sample containing four different LNT isomers was analyzed using a chromatographic setup as described by Castejón-Vilatersana et al. [9] and detection in positive mode, we observed only one peak with an apparent LNT concentration of approximately the sum of the four isomers in our regular chromatography (Figure 6). The large irregular peak at 4 min in positive mode (Figure 6A) is likely due to the isomeric pattern of lactose in dimeric form (m/z 707) since it is also present in the blank sample (Figure 6C). As a control, the samples were analyzed using the same setup but in negative mode. This still resulted in a single peak yet without the noise observed in positive mode (Figure 6E). These results suggest that the previously reported regioselectivity of LnbB cannot necessarily be claimed.

Low regioselectivity in transglycosylation was previously reported for GH20 β-N-acetylhexosaminidases from Aspergillus oryzae [49,50], Paraglaciecola hydrolytica [18], Talaromyces flavus [51], Nocardia orientalis [52], the Chamelea gallina mollusk [53], and soil metagenome HEX1 [28]. Observed products include β1,3-, β1,4-, β1,6-, and unidentified linkages [8] as well as β1,1-linked nonreducing oligosaccharides [18,49,51]. Numerous cases of regioselective transglycosylation with GH20 enzymes have also been reported [8], yet comparison of the current work with previous reports on LnbB transglycosylation suggests that great care must be taken in product analysis in order to confirm the true regioselectivity of an enzyme.

Reaction Product Analysis

During the quantification of the reaction products formed by LnbB variants, four different products were observed with masses corresponding to hex₃hexNAc₁ (m/z 706 and m/z 752, corresponding to the deprotonated [M−H]⁻ and formic acid adduct [M+FA]⁻, respectively; Figure 1 and Figure 7). Based on retention time and fragmentation pattern, the dominant isomer was identified as LNT (LNB-β1,3-Lac). In order to compare the additional isomers to the synthesized GlcNAc-β1,2/3/4/6-Lac (LNT2 isomers) standards, the isomers were isolated by preparative HPLC and digested with the GH35 β1,3-specific β-galactosidase BgaC. The hydrolysis products of two isomers could be identified as 2′LNT2 (GlcNAc-β1,2-Lac) and 6′LNT2 (GlcNAc-β1,6-Lac) by the retention time (Figure 7E,I) and fragmentation pattern (Figure S5A–D), hence corresponding to LNB-β1,2-Lac and LNB-β1,6-Lac. However, the purified fraction of the identified LNB-β1,6-Lac (Figure 7H) appears to contain impurities in a separate peak. Since none of the compounds eluted at the same time as 4′LNT2, it can be concluded that no LNB-β1,4-Lac was formed by LnbB, and it can be assumed that the impurities could be LNB substitutions on the Glc moiety of lactose (Figure S2). Finally, the identification of LNT2 (GlcNAc-β1,3-Lac) verified the presence of LNT as a major product (Figure 7K and Figure S5E,F).

The β-galatosidase treatment of the last isomer did not result in any LNT2-like structure (Figure 7G), although the isolated fraction clearly contained an LNT-like structure (Figure 7F). It was previously reported that enzymatic transglycosylation can yield products with 1,1 linkages, i.e. a linkage between two reducing ends, both for GH20 β-N-acetylhexosaminidases with the formation of β-Gal-1,4-β-Glc-1,1-β-GlcNAc (Lac-1,1-GlcNAc; Figure S2) [18] as well as for β-galactosidase-catalyzed formation of galacto-oligosaccharides (GOS) used for infant nutrition such as β-Gal-1,1-β-Glc [54]. To investigate the presence of 1,1-linked compounds in the LnbB reaction mixture, selected reactions were reduced by NaBH₄. The comparison of the extracted ion chromatograms of native LNT isomers (m/z 706) with those of the reduced reactions (expected LNT alditols m/z 708), it was evident that one compound (15.5 min in Figure 8) could not be reduced, whereas all other compounds were reduced, now appearing as m/z 708 (Figure 8). It can be assumed this compound lacks the reducing end and is most likely LNB-1,1-Lac.

3.3. Significance of the Donor Substrate

The five best LnbB variants W394F, W394H, Y419N, W465F, and W465H were selected according to their higher LNT production and low secondary hydrolysis in the case of W394H, Y419N, and W465H. The variants were further characterized by monitoring transglycosylation reactions over 24 h using pNP-LNB as donor and lactose as acceptor and 1 μM of enzyme at pH 4.5 and 30 °C (

Table 2. Maximum yields of LNT and of LNT plus isomers (total transglycosylation products; TG) from transglycosylation reactions by LnbB WT and variants, using 2.5 mM pNP-LNB as donor, 100 mM lactose as acceptor, and 1 µM of enzyme at pH 4.5 and 30 °C. Superscript letters (^a–f) indicate statistically significant difference between the values within the column (p < 0.05). The time at which the maximum transglycosylation yields are reached is indicated (samples taken after 0.08, 0.25, 2, and 24 h of reaction). Secondary hydrolysis is indicated as the percentage of LNT, which has been hydrolyzed after 24 h of reaction (calculated as (max. LNT yield − LNT yield at 24 h]/[max. LNT yield)). An asterisk (*) indicates that the LNT concentration after 24 h is not significantly different from the maximum yield (secondary hydrolysis not statistically significant; p < 0.05). For each column, the best results are marked in bold, whereas the second-best results are underlined.

		LNT Formation			Total LNT Isomer (TG) Formation
Variant	Time of Max. Yield [h]	Max. LNT [µM]	Max. LNT Yield	% LNT Hydrolyzed after 24 h	Max. TG [µM]	Max. TG Yield	% Total TG Hydrolyzed after 24 h
WT	0.08	7 ± 0.3 f	0.3% f	55 ± 8%	29 ± 3 d	1.2% d	52 ± 7%
W394F	2	164 ± 0.05a	6.6% a	97 ± 0.1%	285 ± 5 a	11% a	91 ± 0.3%
W394H	2	103 ± 3 b	4.1% b	6 ± 18% *	286 ± 5 a	11% a	43 ± 10%
Y419N	0.25	76 ± 3 d	3.0% d	95 ± 1%	171 ± 3 b	6.8% b	86 ± 2%
W465F	0.08	46 ± 0.04 e	1.8% e	93 ± 0.5%	94 ± 0.7 c	3.7% c	86 ± 2%
W465H	24	92 ± 0.7 c	3.7% c	-	175 ± 3 b	7.0% b	-

; Figure 9).

Compared to the activity screening with LNB-oxazoline as donor substrate (Figure 4; Table 1), regioselectivity was lower when using pNP-LNB as donor substrate (Figure 9; Table 2). The percentage of LNT produced of the total transglycosylation products varied from 47% to 59%.

As expected, all variants showed higher maximum LNT and total transglycosylation product yields compared to the WT. LnbB WT produced a maximum of 7 µM LNT in the early stage of the reaction (15 min), representing 0.3% of the maximum theoretical molar yield based on the donor substrate. The five mutants showed maximum LNT yields ranging from 46 to 164 µM representing 1.8% to 6.6% of the maximum theoretical LNT yield, whereas the maximum yield of total transglycosylation products (LNT and LNT isomers) ranged from 94 to 286 µM, corresponding to 3.7% to 11% of the maximum theoretical yield (Table 2; Figure 9).

In the previously presented reactions using LNB-oxazoline as donor, all five variants exhibited similar maximum LNT yields with no statistically significant differences between them—except for W465F, where the production of LNT was a slightly lower (Table 1). Interestingly, this is not the case when using pNP-LNB as donor, underlining the importance of the choice of donor substrate for transglycosylation reactions. With pNP-LNB as donor substrate, W394 of the subsite −1 hydrophobic platform (Figure 3) was the best engineering target for improving transglycosylation. W394F showed the highest maximum LNT yield with 164 µM of LNT produced, whereas W394H showed a maximum LNT yield of 103 µM (after 2 h of reaction). Looking at the sum of LNT isomers, no difference was observed between the transglycosylation performances of W394F and W394H; both reached a maximum transglycosylation yield of 11% after 2 h of reaction. Thus, with pNP-LNB as donor substrate, regioselectivity was higher for W394F than for W394, but as also observed in the reactions with LNB-oxazoline as donor substrate, substituting Trp with His rather than Phe had a positive effect on the undesirable product hydrolysis (Figure 9). The same trend was observed for W465 (Figure 9), where the W465H did not reach the end of the reaction after 24 h (92 µM LNT formed). For Y419N, secondary hydrolysis was more severe with pNP-LNB as donor substrate compared to LNB-oxazoline (Figure 4 and Figure 9). Recently, W394F was also determined as the best variant for improved transglycosylation with a maximum transglycosylation yield of 32% also using 2.5 mM pNP-LNB as donor and 200 mM lactose as acceptor; W394H was not included in that study, but several other variants were investigated, including the substitution of Trp with Ala, Lys, Glu, and Gln [9].

Furthermore, the substrate concentrations of pNP-LNB and lactose were optimized for LnbB W394H (

Table 3. Optimization of substrate concentrations for the transglycosylation reaction with LnbB W394H using pNP-LNB as donor substrate (D) and lactose as acceptor substrate (A) at pH 4.5 and 30 °C with 1 µM of enzyme. Product concentrations and molar yield (calculated on donor substrate) of LNT and of LNT plus isomers (total transglycosylation products; TG) are given, with the time at which the maximum yield is reached given in brackets (monitored over 24 h). Superscript letters (^a–e) indicate statistically significant difference between the values within the column (p < 0.05).

Reaction Conditions			LNT Formation		Total LNT Isomer (TG) Formation
pNP-LNB [mM]	Lactose [mM]	A/D Ratio	Max. LNT [µM]	Max. LNT Yield	Max. TG [µM]	Max. TG Yield
2.5	200	80	107 ± 14 e (2 h)	4% b	225 ± 26 e (2 h)	9% c
2.5	400	160	188 ± 22 de (2 h)	8% a	396 ± 25 d (2 h)	16% a
5	200	40	247 ± 11 cd (2 h)	5% b	539 ± 25 c (2 h)	11% b
5	400	80	342 ± 12 bc (2 h)	7% a	765 ± 28 b (2 h)	15% a
10	200	20	389 ± 41 b (24 h)	4% b	759 ± 23 b (2 h)	8% c
10	400	40	746 ± 81 a (24 h)	7% a	1224 ± 65 a (2 h)	12% b

, Figure 10), which was one of the best variants for both donor substrates, especially in terms of low secondary hydrolysis (Table 1 and Table 2). Reaction conditions can be optimized to favor transglycosylation over hydrolysis, and it is generally observed that an increase in initial substrate concentrations and a high acceptor-to-donor (A/D) ratio tend to give improved transglycosylation yields [55]. In this study, the concentrations of pNP-LNB donor substrate ranged from 2.5 to 10 mM, whereas lactose acceptor concentrations were either 200 or 400 mM, with A/D ratio ranging from 20 to 160 at pH 4.5, 30 °C, and using 1 µM of LnbB W394H.

As expected, the maximum LNT yield of 746 µM and a total LNT isomer yield of 1224 µM were reached when using the highest substrate initial concentrations of both donor (10 mM pNP-LNB) and acceptor (400 mM lactose). The best percentage yields on donor substrate of LNT (8%) and total LNT isomers (16%) were obtained in the reaction using the highest A/D ratio of 160 (Table 3; Figure 10). For the same initial donor concentration, maximum transglycosylation yields were significantly improved when doubling the acceptor substrate concentration, i.e., when doubling the acceptor-to-donor ratio (Table 3).

3.4. Thermal Stability of Best Variants

All variants had significantly lower melting temperature (T_m) values than the WT enzyme (64.4 °C;

Table 4. Thermal stability of LnbB WT and selected variants determined as the thermal unfolding transition midpoint, T_m in °C, at which half of the protein population is unfolded—often referred to as the melting temperature. Superscript letters (^a–e) indicate significant difference in the T_m value between enzyme variants (p < 0.05).

Enzyme	Tm [°C]
WT	64.4 ± 0.01 a
W394F	58.0 ± 0.06 e
W394H	59.4 ± 0.12 d
Y419N	60.6 ± 0.03 c
W465F	63.0 ± 0.04 b
W465H	62.9 ± 0.01 b

). For the W465 variants, the decrease in T_m was not dramatic (63 °C). For Y419N, the T_m decreased by approximately 2 °C, whereas the mutation of W394 had the largest effect on T_m. (58.0–59.4 °C). It appears that the decrease in T_m could be linked to solvent exposure: W465 is placed deep in the active site ‘below’ the ligand, whereas Y419 and especially W394 are on the outside, more exposed to the solvent. The T_m values are in the same range as previously reported for recombinantly produced LnbB WT and variants (61–62.5 °C) [9] yet with slightly more variation.

4. Conclusions

LnbB represents the archetype of LNBases, and this study confirms how LnbB, which possesses negligible transglycosylation activity in itself, can be engineered by single-point mutations to catalyze formation of transglycosylation products, here exemplified by the formation of its best known substrate: the human milk oligosaccharide LNT. The results emphasize how both the choice of donor substrate as well as the careful consideration of amino acid substitutions in rational design affect the resulting transglycosylation capacity. The optimization of the substrate concentrations confirms the theory that high initial substrate concentration and high A/D ratio both favor transglycosylation. Finally, we demonstrated the importance of good separation in product analysis in transglycosylation, where regioselectivity is desired yet often not obtained. High resolution in product analysis can in turn pave the way for the identification of novel products: LnbB catalyzes the formation of four different linkages between LNB and lactose in both reaction systems, with LNT being the major product. In previous works [9,15] as well as in the current one, LnbB is able to degrade all formed products. This suggest that although highly selective for the internal β1,3-linkage in LNB (between subsites −1 and −2) [3], LnbB does not exhibit a similar high specificity toward the linkage at the cleavage point.