Improved Enzymatic Production of the Fucosylated Human Milk Oligosaccharide LNFP II with GH29B α-1,3/4-l-Fucosidases

Abstract

Five GH29B α-1,3/4-l-fucosidases (EC 3.2.1.111) were investigated for their ability to catalyze the formation of the human milk oligosaccharide lacto-N-fucopentaose II (LNFP II) from lacto-N-tetraose (LNT) and 3-fucosyllactose (3FL) via transglycosylation. We studied the effect of pH on transfucosylation and hydrolysis and explored the impact of specific mutations using molecular dynamics simulations. LNFP II yields of 91 and 65% were obtained for the wild-type SpGH29^C and CpAfc2 enzymes, respectively, being the highest LNFP II transglycosylation yields reported to date. BbAfcB and BiAfcB are highly hydrolytic enzymes. The results indicate that the effects of pH and buffer systems are enzyme-dependent yet relevant to consider when designing transglycosylation reactions. Replacing Thr284 in BiAfcB with Val resulted in increased transglycosylation yields, while the opposite replacement of Val258 in SpGH29^C and Val289 CpAfc2 with Thr decreased the transfucosylation, confirming a role of Thr and Val in controlling the flexibility of the acid/base loop in the enzymes, which in turn affects transglycosylation. The substitution of an Ala residue with His almost abolished secondary hydrolysis in CpAfc2 and BbAfcB. The results are directly applicable in the enhancement of transglycosylation and may have significant implications for manufacturing of LNFP II as a new infant formula ingredient.

Introduction

Human milk oligosaccharides (HMOs) are lactose-based carbohydrate oligomers that represent the third most abundant component in human milk. HMOs play an important role in the health and development of breastfed infants in multiple ways.^1,2 To ensure that bottle-fed infants receive the same health benefits as breastfed infants, research into the synthesis of HMOs as a food additive in infant formula and as dietary supplements is gaining considerable attention.

Over 200 different HMO molecules have been identified, and more than 100 of these have been structurally studied.³ Briefly, the basic components for almost all natural HMOs are β-d-galactose (Gal), β-d-glucose (Glc), β-d-N-acetylglucosamine (GlcNAc), and α-l-fucose (Fuc), as well as the sialic acid α-d-N-acetylneuraminic acid (Neu5Ac). All HMOs have lactose (Gal-β-1,4-Glc) at their reducing end. Generally, more complicated HMOs have distinct fucose and/or sialic acid substitutions and/or are elongated by either β-N-acetyllactosamine (LacNAc; Gal-β-1,4-GlcNAc) or lacto-N-biose (Gal-β-1,3-GlcNAc) moieties, which in turn may also be fucosylated or sialylated.³ Fucosylated oligosaccharides make up more than half of HMOs, with, e.g., 2′-fucosyllactose (2′-FL), 3-fucosyllactose (3FL), and lacto-N-fucopentaose (LNFP) I/II/III being abundant. Fucosylated HMOs have attracted attention because of their physiological effects as prebiotics, antiadhesive antimicrobial agents, and immune regulators on breastfed infants.^1,4,5

Chemical strategies⁶ and direct microbial production by Escherichia coli fermentation have been implemented for the synthesis of certain HMOs.⁷⁻¹⁰ Specifically, the large-scale manufacture of 2′-FL, 3FL, and other neutral nonfucosylated HMOs such as lacto-N-tetraose (LNT) and lacto-N-neotetraose (LNnT) are commercially produced by fermentation of metabolically engineered E. coli.⁸ Furthermore, production of more complex fucosylated HMOs such as LNFP I can be achieved in engineered E. coli(11) and a recent publication reported the in vivo synthesis of LNFP III by using a metabolically engineered strain.¹² However, the use of microbial cell factories for production of such complex fucosylated HMOs suffers from low titers, side reactions, and intracellular product formation^7,13 and cell factories for production of LNFP II have not yet been reported. Thus, direct enzymatic synthesis by transfucosylation using regiospecific α-l-fucosidases is a promising option for direct in vitro synthesis of complex fucosylated oligosaccharides.¹⁴⁻¹⁷ Recent studies have reported that LNFP II was successfully synthesized in this way using LNT as acceptor and 3FL as fucosyl donor¹⁷⁻¹⁹ (Figure 1).

Figure 1.

Transglycosylation of LNT using 3FL as donor and GH29B α-l-fucosidases. The reaction scheme is shown according to the classical Koshland double-displacement mechanism. A fucosyl-enzyme intermediate is formed in the first step of the reaction, which is shared by the two reactions that can result from it, namely, hydrolysis and transglycosylation. The formation of lacto-N-fucopentaose II (LNFP II) from 3-fucosyllactose (3FL) and lacto-N-tetraose (LNT) occurs if the LNT attacks the fucosyl-enzyme intermediate, whereas the 3FL is hydrolyzed to lactose and fucose when the acceptor is a water molecule. The LNFP II transfucosylation product can also be a substrate for the secondary hydrolysis or function as a donor substrate.

α-l-Fucosidases catalyze the hydrolysis and/or transfer the nonreducing α-l-fucosyl residues of carbohydrates or glycoconjugates and are currently categorized in four GH families in the CAZy database (http://www.cazy.org/²⁰), i.e., GH29, GH95, GH141, and GH151. GH95 constitutes α-l-fucosidases as well as α-1,2-l-fucosidases (EC 3.2.1.63) with an inverting reaction mechanism, which makes them unsuitable for transglycosylation except that they can be transformed into fucosynthases.²¹ GH141 and GH151 are new families, including only a few poorly characterized members. The GH29 family includes α-l-fucosidases (EC 3.2.1.51), α-1,2-l-fucosidases (EC 3.2.1.63), α-1,3/4-l-fucosidases (EC 3.2.1.111), and α-1,6-l-fucosidases (EC 3.2.1.127). These enzymes all employ a retaining double-displacement reaction mechanism involving a catalytic nucleophile and a catalytic acid/base, both of which are amino acids with carboxylic acid side chains. They catalyze both hydrolysis and to a certain extent transglycosylation—resulting in the competition between the two pathways—which is usually expressed in terms of the ratio T/H²² (Figure 1). Several GH29 members have been used for transfucosylation forming relatively complicated HMO molecules^{18,19,23−25} as well as glycoproteins.²⁶ Commonly—based on sequence homology, substrate specificities, and phylogenetic relationships—GH29 α-l-fucosidases can be further divided into two subfamilies GH29A and GH29B. The former has relatively loose substrate specificities, while the latter is more regiospecific and mainly functions on α-1,3/4 fucosidic linkages with a branched Gal residue.²⁷

Efficient transglycosylation can be achieved via protein engineering, optimized reaction conditions, control of reaction time and by modulating temperature and/or pH.²⁸ The pH is important, because catalysis directly involves ionizable amino acid residues. Consequently, the pH-dependent activity of an enzyme is determined primarily by the pK_a values of one or more key ionizable groups around its active site cleft.²⁹ However, the effect of pH on modulating the T/H balance has only been scarcely investigated and never among the GH29B α-1,3/4-l-fucosidases, which are important synthetic tools due to their high regioselectivity. To address this paucity, one aim of this study was to give new insights into the pH-dependent transfucosylation and hydrolytic activities of five GH29B fucosidases, CloFuc, CpAfc2, BbAfcB, SpGH29^C, and BiAfcB, enabling an improved formation of LNFP II from LNT and 3FL (Figure 1). Four of these enzymes have previously been employed in transglycosylation, and two have crystal structures enabling reliable studies of structure–function relationships. Single mutant enzyme variants for each wild-type fucosidase were also investigated. Molecular dynamics (MD) simulations were performed to provide a reasonable strategy for improving the transglycosylation efficiency of GH29B fucosidases via protein engineering.

Materials and Methods

Chemicals

The fucosylated oligosaccharides 3FL and LNT were kindly provided by DSM-Firmenich (Hørsholm, Denmark). The external standard for analysis of lacto-N-fucopentaose (LNFP II) was purchased from Carbosynth (Compton, UK), while the standard for lacto-N-fucopentaose V (LNFP V) was purchased from Elicityl (Crolles, France). l-Fucose and all other chemicals were purchased from Sigma-Aldrich (Steinheim, Germany).

Enzymes and Gene Constructs

For SpGH29^C30 and BiAfcB, the amino acid sequence was retrieved from GenBank. The online tools SignalP 6.0³¹ and DeepTMHMM version 1.0.10³² were used to predict signal peptide and transmembrane region of SpGH29^C and BiAfcB, respectively. Signal peptides of BbAfcB, CpAfc2, and CloFuc were removed from the sequence, and the resulting genes were designed to harbor a C-terminal 6× histidine tag, codon-optimized for expression in E. coli, and synthesized and inserted into the vector pET22b(+) by GenScript (Piscataway, New Jersey) (Table S1). E. coli DH5α was used for plasmid propagation. E. coli BL21(DE3) was transformed with the resulting plasmids (and later similarly with mutant genes) and successful transformants selected via ampicillin resistance (Table S1). Fucosidases CpAfc2 as well as BbAfcB and CloFuc were expressed in a similar way as described previously.^23,33

For expression, transformed cells were inoculated in LB supplemented with 100 μg/mL ampicillin at 37 °C. At an OD₆₀₀ of 0.4–0.6, the temperature was reduced to 20 °C; expression was induced with 1 mM IPTG, and cells were grown with light shaking for 12–18 h before being harvested. CloFuc variants R364S and A260H were induced with 0.1 mM IPTG and cells were grown at 18 °C overnight to increase the expression yield. All expressed proteins were purified by IMAC purification, as described earlier.²³ Briefly, cells were harvested by centrifugation at 5000g for 30 min and 4 °C and the pellet was resuspended in binding buffer (20 mM sodium phosphate buffer, 500 mM NaCl, 20 mM imidazole, pH 7.5) before being disrupted in a Q500 Sonicator (Qsonica, Newtown, Germany). Cell debris was pelleted by centrifugation (20,000g, 20 min at 4 °C). The supernatant was then filtered through a 0.45 μm filter and applied to a 5 mL Ni²⁺ Sepharose HisTrap HP column (GE Healthcare, Uppsala, Sweden). Variants W369H (CloFuc), S290 K (CpAfc2), K809S, and A705H (BbAfcB) were subjected to further purification using gel filtration (Superdex 75 Increase 10/300 GL, GE Healthcare, Uppsala, Sweden) in 20 mM sodium phosphate at pH 6.0, and 100 mM NaCl, followed by desalting using an ultrafiltration spin column (GE Healthcare). Protein purity was confirmed by SDS-PAGE, and concentration was measured at 280 nm on a NanoDrop ND-1000 (Thermo Scientific, Waltham, Massachusetts, USA).

MD Simulations

MD simulations were performed using the available crystal structures of SpGH29^C (PDB IDs 6OR4, 6ORG, 6ORF, and 6ORH(30) and BiAfcB (PDB IDs 3UES and 3UET).³⁴ For further structural comparisons with the other enzymes, AlphaFold models of CloFuc, CpAfc2, and BbAfcB were built via ColabFold.³⁵ Active site residues were modeled by AlphaFold with very high confidence (pLDDT), and in a few cases 90 > pLDDT > 70 (high confidence). Flexible active site loops also had a high pDLLT score (75–80), except for the nucleophile-carrying loop of CpAfc2, which had a score of 66. MD simulations were carried out with the Groningen Machine for Chemical Simulations (GROMACS) 2021.2 software package³⁶ in combination with the CHARMM36m force field³⁷⁻³⁹ and TIP3P water model⁴⁰⁻⁴² using standard GROMACS settings (see Supplementary Methods). The apparent pK_a values of titratable residues were calculated with PROPKA3,⁴³ H++,⁴⁴ and constant pH as implemented in NAMD.⁴⁵

Site-Directed Mutagenesis

All mutagenic primers for site-directed mutagenesis of five fucosidases (CpAfc2, BbAfcB, CloFuc, SpGH29^C, and BiAfcB) were designed on Serial Cloner 2.6.1 with the assistance of the ExPASy ProtParam server and codon usage in E. coli (Table S2).

Around 10 nmol of mutagenic primers (both forward and reverse primers) was mixed with 100 ng of plasmid vector for mutagenesis of five WT fucosidases. The reaction conditions were 12.5 μL of CloneAmp HiFi PCR Premix containing CloneAmp HiFi Polymerase, dNTPs, and optimized buffer (Takara), and Milli-Q water replenishing to a 25 μL PCR. The reaction was thermocycled as follows: 98 °C for 30 s, and then 20 cycles at 98 °C for 10 s, 60–62 °C for 10 s (60 °C for mutation of CpAfc2 and SpGH29^C, 61 °C for CloFuc, and 62 °C for BbAfcB and BiAfcB) and 72 °C for 3–4 min (3 min for mutation CloFuc, SpGH29^C, and BiAfcB, and 4 min for CpAfc2 and BbAfcB), and a final extension at 72 °C for 10 min. PCR products were digested at 37 °C overnight with 1 μL of DpnI restriction enzymes (Thermo Scientific, Waltham, Massachusetts, USA). The resulting plasmids were treated and recycled according to the protocol of GFX PCR DNA and Gel Band Purification Kit (GE Healthcare, Uppsala, Sweden) and were transformed into E. coli DH5α to propagate, and then the supposed mutation plasmids were extracted by using GeneJET Plasmid Miniprep Kit (Thermo Scientific, USA). The achievement of site-directed mutagenesis for each mutant enzyme was verified via sequencing in the forward or reverse direction (Macrogen Europe, Amsterdam, The Netherlands).

Transglycosylation Activity

In investigating the hydrolytic and transglycosylation activities of GH29B fucosidases, CpAfc2, BbAfcB, CloFuc, SpGH29^C, and BiAfcB, two different buffer systems, namely, single buffer (SB) and universal buffer 2 (UB),⁴⁶ were used to compare the effect caused by buffers. Both buffers were prepared to varying pH values adjusted at room temperature, i.e., setup pH 4.5, pH 6.0, pH 7.5, and pH 9.0. In detail, 80 mM SB contains acetate buffer pH 4.5, phosphate buffer pH 6.0 and pH 7.5, and glycine–NaOH buffer pH 9.0, whereas 80 mM UB consists of Tris–HCl, Bis–Tris, and sodium acetate. Besides, the actual pH value in each reaction was measured (at 40 °C) to assess the true pH value experienced by the enzyme (Table S3). For the variants of the five GH29B fucosidases prepared by site-directed mutagenesis, a single optimal pH (as measured in the reaction at 40 °C) was used: UB at pH 7.2 for CpAfc2 and CloFuc, UB at pH 4.5 for BbAfcB, SB at pH 5.9 for SpGH29^C, and SB at pH 7.9 for BiAfcB.

100 mM LNT as acceptor and 10 mM 3FL as donor substrate, i.e., an acceptor-to-donor ratio (A:D) of 10, were applied for reactions catalyzed by each of the five GH29B fucosidases and corresponding variants. The enzyme concentration was 0.5 μM, and each reaction took place at 40 °C for up to 120 min in each type of buffer. At each time point, i.e., 1, 5, 10, 30, 60, and 120 min, 20 μL of reaction sample was transferred into 380 μL of preheated water for 20 times dilution and terminated by heating at 95 °C for 10 min. Samples were then transferred to HPLC vials for analysis.

HPAEC-PAD Analysis

To analyze products synthesized by the fucosidases, high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD) was carried out on an ICS3000 system (Dionex Corp., Sunnyvale, California, USA) using a CarboPac PA1 (4 mm × 250 mm) analytical column equipped with a CarboPac PA1 (4 mm × 50 mm) guard column (Dionex Corp., Sunnyvale, California, USA) and a flow rate of 1 mL/min at 25 °C. The eluent system comprised Milli-Q water (A), 500 mM NaOH (B), and 500 mM NaOAc with 0.02% (w/v) NaN₃ (C). Meanwhile, to be able to quantify both hydrolyzed substrate and transglycosylation product, external standards including fucose, 3FL, LNFP II, and lactose of 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, and 1 mM were prepared for calibration curves. The reaction products were quantified by isocratic elution at 90:10:0 (A/B/C, %) for 14 min, and the column was washed at 5:10:85 (A/B/C, %) for 4 min followed by re-equilibration of the column at 90:10:0 (A/B/C, %) for 10 min.

Transfucosylation yields were calculated as molar yields based on the donor substrate 3FL, i.e., the percentage of LNFP II produced in reaction compared to the initial 3FL concentration. Initial transfucosylation (hydrolysis) rates were defined as μM LNFP II (fucose) formed per min of reaction over the time where the curve of product formation remained linear.

LC-ESI-MS Analysis

Identification of transglycosylation product regioisomers and quantitation of potentially regioisomeric mixtures were done with liquid chromatography coupled to electrospray ionization-mass spectrometry (LC-ESI-MS) using samples from the reaction conditions and time giving rise to the highest product yield for each enzyme, namely, UB pH 7.2 for CloFuc, UB pH 8.3 for CpAfc2, UB pH 4.5 for BbAfcB, SB pH 5.9 for SpGH29^C, and SB pH 7.9 for BiAfcB. The LC-ESI-MS analysis was performed on an amaZon SL ion trap (Bruker Daltonics, Bremen, Germany) coupled to an UltiMate 3000 UHPLC from Dionex (Sunnyvale, California, USA) equipped with a porous graphitized carbon column (Hypercarb PGC, 150 mm × 2.1 mm, 3 μm; Thermo Fisher Scientific, Waltham, Massachusetts, USA) as described previously, allowing separation of regioisomers,^23,47 using a target mass of 850 m/z. Quantitation was performed in Compass TASQ 2.2 (Bruker Daltonics) by using LNFP II and LNFP V as external calibration standards. The diagnostic fragments of the different LNFP structures were confirmed in accordance with Pfenninger et al.⁴⁸

Statistics

One-way ANOVA for determination of statistical significance was performed with JMP, version 16.0.0 (SAS Institute Inc., Cary, North Carolina, USA). Statistical significance was established at p < 0.05. All data plots were fitted with Origin, version 9.8.0.200 (OriginLab, USA).

Sequence Analysis

For selected mutations, the amino acid conservation at a particular mutation site was analyzed. To do this, all sequences of Conserved Unique Peptide Pattern (CUPP) groups GH29:3.1, GH29:3.2, GH29:3.3, and GH29:3.4, which correspond to subfamily GH29B as described previously,³³ were downloaded from the CUPP Web server (www.cupp.info⁴⁹) and a multiple-sequence alignment of the resulting 4166 sequences was prepared in CLC Main Workbench version 23.0.5 (Qiagen Digital Insights, Redwood City, California, USA) using default settings. The columns of aligned residues corresponding to Thr284 and Ser285 of BiAfcB, respectively, were extracted, and the occurrence of each residue was calculated.

Results

Expression of Wild-Type and Mutant α-l-Fucosidases

CloFuc, CpAfc2, BbAfcB, SpGH29^C, and BiAfcB were successfully expressed in E. coli BL21(DE3) cells, reaching 46, 8, 48, 16, and 188 mg of purified enzyme per liter of culture, respectively. Expression levels of the mutant enzymes were lower than those of the WT for CloFuc and BbAfcB variants, similar for the CpAfc2 variants, and increased for SpGH29^C and BiAfcB variants (Table S4). The purity of the WT enzymes as well as the 23 enzyme mutants was satisfactory as assessed by SDS-PAGE, and the proteins obtained agreed with the expected molecular sizes (Figure S1).

Transglycosylation Activity of WT Fucosidases at Different pH Values

Hydrolytic and transfucosylation activity of the five GH29B fucosidases in the pH range from pH 4.5 to pH 8.5 was monitored over 2 h using 10 mM 3FL as fucosyl donor and 100 mM LNT as acceptor substrate, in both SB and UB buffer systems. UB is a mixed biological buffer crossing a broad pH range, is suitable for investigating enzyme reactions because of negligible metal-binding affinity, and can avoid interferences caused by changing the small molecule composition of buffers when pH is used as an experimental variable.⁴⁶ Meanwhile, SB solution was used as a reference in this study in case the enzyme is sensitive to the UB system. The actual pH values in reactions (Table S3 and legends of Figure 2a–e) indicated that the UB system reached true pH values at 40 °C closer to those of the setup pH, especially at the highest pH values. For SpGH29^C, CpAfc2, and BiAfcB, LNFP II was the only identified transglycosylation product detected, emphasizing the high regioselectivity of these GH29B enzymes (Figure S2). For BbAfcB, a low amount (9%) of LNFP V was detected; the low percentage of this other HMO, which is α-1,3-fucosylated on the reducing end Glc moiety, indicates that BbAfcB also has an adequately high regioselectivity. In contrast, the LC-ESI-MS analysis revealed that the CloFuc transglycosylation product at optimal conditions (UB at pH 7.2; see below) was a mixture of 68% LNFP II and 32% LNFP V (Figure S2).

Figure 2.

Transfucosylation and hydrolysis catalyzed by current GH29B α-l-fucosidases. Molar transfucosylation yields ([LNFP II]/[3FL]₀; left) and hydrolyzed fucose concentration (right) obtained in transglycosylation catalyzed by CloFuc (a), CpAfc2 (b), BbAfcB (c), SpGH29^C (d), and BiAfcB (e) using 3FL as fucosyl donor substrate and LNT as acceptor substrate (n = 2) at pH 4.5 (filled squares), pH 5.7–6.0 (filled circles), pH 6.9–7.2 (filled regular triangles), and pH 8.1–8.5 (filled inverted triangles) in universal buffer UB (solid lines), and at pH 4.4–4.5 (open squares), pH 5.9–6.1 (open circles), pH 7.3–7.4 (open regular triangles), and pH 7.8–7.9 (open inverted triangles) in SB (dashed lines). The actual pH values in reactions corresponding to setup pH are listed in Table S3. CloFuc and BbAfcB produced mixtures of LNFP II and LNFP V (68/32% at the maximum yield of WT CloFuc after 60 min of reaction in UB pH 7.2, and 91/9% at the maximum yield of WT BbAfcB after 5 min of reaction in UB pH 4.5), while all other enzymes produced LNFP II only (Figure S2).

For CloFuc and CpAfc2, important differences were observed between the two buffer systems (Figure 2a,b and Tables 1 and 2). CloFuc exhibited the highest initial transfucosylation rate and transglycosylation rate-to-hydrolysis rate ratio (T/H) at low pH 4.5 in both buffer systems (UB and SB), leading to transfucosylation yields of 16 and 27%, respectively, but low activities at high pH values of 7.8–8.4 (Figure 2a and Table 1). At neutral and slightly acidic conditions, the hydrolytic rates were high, resulting in highly transient LNFP maxima (Figure 2a and Table 1). The highest LNFP product yield of 38% was obtained around 60 min at pH 7.2 in UB solution (Figure 2a and Tables 1 and 2). Interestingly, the performance was much poorer in the phosphate buffer (SB) at pH 7.4 and a similar trend was observed at pH 6.0 in the SB system, indicating that CloFuc is sensitive to phosphate buffer. For CpAfc2, the hydrolytic rates were generally higher in the SB systems and lower in the UB systems, directly leading to the general tendency that higher LNFP II yields were obtained in the UB systems than in the SB systems (Figure 2b and Table 1). This effect was particularly dramatic at the highest pH, where the maximum LNFP II yield reached around 65% at pH 8.3 in the UB system, while it was only 4.3% in SB pH 7.9 coupled with the largest hydrolysis rate. This might be an effect of the difference in actual pH values (pH 8.3 vs pH 7.9) but could certainly also be caused by the difference between the UB and the glycine-NaOH buffer (SB). Similarly, the molar yield of LNFP II at the lowest pH 4.5 in UB reached 52.6%, which was twice the yield in SB pH 4.5 (Table 2). Furthermore, hydrolysis was negligible in UB pH 4.5, leading to the highest T/H value observed (328; Table 1). The molar LNFP II yields on 3FL at both the highest pH 8.3 and lowest pH 4.5 in UB were higher than the previously observed transient maxima of 39%²³ and 50%,¹⁷ where reactions took place at pH 7.0 in phosphate buffer. Importantly, all T/H values in UB solutions were higher than those at the same pH in SB systems (Table 1). Thus, CpAfc2-catalyzed transglycosylation is promoted in the UB system. Importantly, both the highest alkaline pH and the lowest acidic pH in UB play crucial roles in increasing the transfucosylation efficiency.

Table 1. Initial Transfucosylation rate (r_T), Initial Hydrolysis Rate (r_H), and the Ratio between Initial Transfucosylation Rate and Initial Hydrolysis Rate (T/H) in Transglycosylation Catalyzed by the GH29B Fucosidases Using 10 mM 3FL as Donor Substrate and 100 mM LNT as Acceptor Substrate (n = 2).

pH*	CloFuc	CpAfc2	BbAfcB	SpGH29C	BiAfcB
Initial transfucosylation rate rT (μM·min–1)
UB pH 4.5	149 ± 10cd	131 ± 1bc	143 ± 17b	174 ± 7e	291 ± 27b
SB pH 4.4–4.5	257 ± 10a	126 ± 1bc	140 ± 24b	356 ± 4d	358 ± 49a
UB pH 5.7–6.0	201 ± 1b	112 ± 9bc	625 ± 16a	256 ± 1e	113 ± 21de
SB pH 5.9–6.1	138 ± 12d	105 ± 1bc	41.1 ± 8.8c	766 ± 24a	221 ± 31c
UB pH 6.9–7.2	232 ± 26a	226 ± 70b	156 ± 8b	227 ± 2e	149 ± 4d
SB pH 7.3–7.4	168 ± 6c	133 ± 24bc	ND	604 ± 8b	230 ± 14c
UB pH 8.1–8.5	0.850 ± 0.0f	513 ± 167a	38.2 ± 11c	437 ± 107cd	64.6 ± 5.7e
SB pH 7.8–7.9	47.3 ± 0.3e	2.55 ± 0.9c	129 ± 18b	481 ± 0.0c	233 ± 27bc
Initial hydrolysis rate rH (μM·min–1)
UB pH 4.5	ND	0.400 ± 0.0d	1350 ± 90a	1.40 ± 0.0f	186 ± 30c
SB pH 4.4–4.5	29.2 ± 12.9d	3.45 ± 0.64d	493 ± 58b	9.30 ± 0.6cd	196 ± 11c
UB pH 5.7–6.0	74.2 ± 3.7b	7.65 ± 0.0cd	659 ± 87b	11.6 ± 0.0bc	370 ± 103b
SB pH 5.9–6.1	168 ± 7a	43.4 ± 3.8a	6320 ± 270a	18.9 ± 1.0a	573 ± 8a
UB pH 6.9–7.2	57.5 ± 3.1c	16.8 ± 2.0c	1200 ± 60a	10.0 ± 2.0cd	138 ± 39c
SB pH 7.3–7.4	23.3 ± 3.7de	33.0 ± 0.8b	3180 ± 0.0a	7.35 ± 0.6de	467 ± 39b
UB pH 8.1–8.5	ND	7.30 ± 0.28d	6510 ± 590a	4.71 ± 3.7ef	31.2 ± 3.5d
SB pH 7.8–7.9	12.6 ± 2.5e	49.5 ± 10.5a	8030 ± 30a	14.0 ± 0.1b	193 ± 20c
Initial transfucosylation rate/initial hydrolysis rate (T/H)
UB pH 4.5	ND	328	0.106	124	1.56
SB pH 4.4–4.5	8.80	36.5	0.284	38.3	1.83
UB pH 5.7–6.0	2.71	14.6	0.948	22.1	0.305
SB pH 5.9–6.1	0.821	2.42	0.00650	40.5	0.386
UB pH 6.9–7.2	4.03	13.5	0.130	22.7	1.08
SB pH 7.3–7.4	7.21	4.03	ND	82.2	0.493
UB pH 8.1–8.5	ND	70.3	0.00587	92.8	2.07
SB pH 7.8–7.9	3.75	0.0515	0.0161	34.4	1.21

^*The pH values indicated in the first column differ slightly between enzymes and buffer systems; please refer to Table S3 and the text for actual values measured in the enzyme reaction.

ND: no data due to the lack of transfucosylation or hydrolysis activity.

^a–fSignificant difference (p < 0.05) between values of each fucosidase across all pH and buffer systems.

Table 2. Maximum Product Yield ([LNFP II]) at Corresponding Time Points (min) Catalyzed by CloFuc, CpAfc2, BbAfcB, SpGH29^C, and BiAfcB Using 10 mM 3FL as Donor Substrate and 100 mM LNT as Acceptor Substrate (n = 2).

pH	CloFuc*		CpAfc2		BbAfcB*		SpGH29C		BiAfcB
	max LNFP (mM)	time (min)	max LNFP II (mM)	time (min)	max LNFP II (mM)	time (min)	max LNFP II (mM)	time (min)	max LNFP II (mM)	time (min)
UB pH 4.5	1.67 ± 0.07d	60	5.26 ± 0.43b	120	0.835 ± 0.10a	5	6.92 ± 0.07b	120	1.48 ± 0.16bc	5
SB pH 4.4–4.5	2.78 ± 0.00b	120	2.65 ± 0.10d	120	0.831 ± 0.13a	5	5.10 ± 0.49c	120	1.84 ± 0.22ab	5
UB pH 5.7–6.0	2.12 ± 0.01c	10	3.71 ± 0.04c	120	0.625 ± 0.02b	1	8.08 ± 0.08ab	60	1.12 ± 0.16c	10
SB pH 5.9–6.1	0.842 ± 0.05f	5	1.56 ± 0.10e	60	0.257 ± 0.05c	5	9.11 ± 0.53a	30	1.18 ± 0.15c	5
UB pH 6.9–7.2	3.84 ± 0.22a	60	3.68 ± 0.09c	60	0.156 ± 0.01c	1	8.39 ± 0.25ab	60	1.47 ± 0.01bc	10
SB pH 7.3–7.4	1.96 ± 0.21c	60	3.40 ± 0.02c	60	ND		9.00 ± 0.73a	60	2.24 ± 0.19a	10
UB pH 8.1–8.5	0.154 ± 0.00g	120	6.48 ± 0.01a	120	0.233 ± 0.04c	5	7.14 ± 0.37b	120	2.10 ± 0.51a	60
SB pH 7.8–7.9	1.26 ± 0.07e	120	0.433 ± 0.08f	60	0.784 ± 0.09ab	5	8.39 ± 1.43ab	60	2.26 ± 0.26a	10

^*CloFuc and BbAfcB produced mixtures of LNFP II and LNFP V (68/32% at the maximum yield of WT CloFuc after 60 min of reaction in UB pH 7.2, and 91/9% at the maximum yield of WT BbAfcB after 5 min of reaction in UB pH 4.5), while all other enzymes produced LNFP II only (Figure S2).

^a–gSignificant difference (p < 0.05) between values of each fucosidase across all pH and buffer systems.

ND: no data detected.

SpGH29^C was by far the best enzyme for transglycosylation among the five GH29B fucosidases. The molar LNFP II yields at almost all pH values of both buffer systems showed similar trends; i.e., the maximum yields of LNFP II consistently arrived at 70–90% within 30–60 min and either remained stable or decreased slightly to around 70–80% over the next hour, except that the LNFP II yield at pH 4.4 in SB solution reached 51% only (Figure 2d and Table 2). Interestingly, higher transglycosylation yields were obtained at all pH values in the SB systems, indicating a slight sensitivity of SpGH29^C to the UB system containing Tris–HCl, Bis–Tris, and sodium acetate. Moreover, the hydrolysis level of SpGH29^C was the lowest compared with the other four fucosidases (Figure 2d and Table 1). It cannot be overlooked that although a very high T/H was obtained at low pH (UB pH 4.5), a comparatively poorer performance was also observed in terms of the initial transfucosylation rate, which was remarkably better at high pH (SB pH 5.9, SB pH 7.3, UB pH 8.1, and SB pH 7.8), where there seemed to be a rate shift around pH 6–7 (Table 1). Considering the molar LNFP II yield and low hydrolysis, SpGH29^C is the most potent enzyme. Indeed, the LNFP II yields obtained with SpGH29^C and CpAfc2 in the current work are the highest ones reported to date; the best yield of LNFP II previously reported from enzymatic transfucosylation was 51% obtained with the BiAfcB mutant Q244K using 200 mM 3′FL and 200 mM LNT.¹⁹ Thus, SpGH29^C can be considered as the most promising template used for protein engineering to improve transfucosylation performance of other GH29B fucosidases, similar to the approaches previously explored for GH33 sialidases.⁵⁰

BiAfcB and BbAfcB both catalyzed transfucosylation with transient maximum LNFP II yields within 5–10 min, followed by substantial yield decreases during the 2 h of reaction (Figure 2c,e and Table 2). The maximum yield produced by BbAfcB never exceeded 10%, while BiAfcB reached maximum yields of more than 10% at all pH values. The LNFP II yields of BiAfcB at higher pH values (UB pH 8.5 and SB pH 7.9) exceed 20% (Table 2). Fucose concentrations increased continuously for BiAfcB, and all the corresponding hydrolysis rates were clearly higher than those of CloFuc, CpAfc2, and SpGH29^C, but lower than those of BbAfcB. T/H values were higher for BiAfcB than for BbAfcB at almost all pH levels, although both enzymes had lower T/H values than the other enzymes (Table 1). Importantly, there seemed to be a tendency toward a higher T/H at the high pH values (SB pH 7.9 and UB pH 8.4 for BbAfcB and UB pH 8.5 for BiAfcB) as LNFP II yields remained at a moderate level and the hydrolysis decreased at extended reaction times. This is similar to the elimination of hydrolytic activity observed for a GH1 β-glucosidase at high pH,⁵¹ indicating that an increase in T/H might be achieved by the pH modulation for BiAfcB and BbAfcB.

Transglycosylation Activity of Mutant Fucosidases

To understand the role of the active site residues and in turn rationally improve the predisposition for transglycosylation of the GH29B fucosidases, MD simulations were performed on the available crystal structures of SpGH29^C and BiAfcB, both with the Lewis a trisaccharide ligand, which resembles the nonreducing end of the LNFP II reaction product. A key point in the investigation was the new insight that SpGH29^C is superior in transglycosylation while BiAfcB is the opposite. Structural conservation of amino acids across all five fucosidases was confirmed by inspecting AlphaFold2-modeled structures of CloFuc, CpAfc2, and BbAfcB.

In the early part of our modeling work, it became apparent that we needed to select the protonation of the titratable residues of these enzymes with care (Supplementary Methods, Figure S3). Although it should be sufficient, especially when safeguarded by visual inspection of the H-bonding network, the automated process in GROMACS could not meaningfully reproduce the ligand-bound model from crystallography. In all of the MD simulations that were run, the nucleophilic amino acid residue (Nu) and its “shepherding” Tyr (SpGH29^C Asp171 and Tyr123, respectively) flipped from a catalytically relevant state to a seemingly more stable state in which the nucleophile is pointed away from the active site (Figure S4). This change happened quickly (within 30 ns) in most simulations, indicating a shortcoming of the initial model setup. As the only amino acid that could be ambiguous at the modeled pH, the His residues were investigated further. Furthermore, PROPKA3, H++, and constant pH were utilized but could only give inspiration the protonation states of the individual His residues. Therefore, we decided to test several models and select the ones that fit the X-ray data of these enzymes best. For this work, the so-called triad of His turned out to be the most impactful in this regard, and so all combinations of protonation states of these residues were tested. Out of all of the combinations tested, the combination where only the His with the highest residue number of the His triad (SpGH29^C His78) was ε protonated while the other two were δ protonated (SpGH29^C His28 and His77) resulted in the highest population of states that fit the X-ray crystallography data.

A series of enzyme variants were designed with the objective of studying the effect of the structural flexibility of the acid/base residue on transfucosylation. The flexible loop 6 containing the acid/base residue Glu215 in SpGH29^C ranges from Trp211 to Lys226 (Figure 3b and Figure S5). This included variants SpGH29^C V258T (Figures 3b and 4), CpAfc2 V289T, and BiAfcB T284V to investigate the role of Val258 in SpGH29^C and Val289 in CpAfc2, the two best transglycosylators in the study, which is Thr284 in BiAfcB, allowing formation of a hydrogen bond with a backbone carbonyl in the flexible loop containing the catalytic acid/base in BiAfcB, thereby reducing the loop flexibility. CloFuc and BbAfcB have a smaller alanine residue in this position. Furthermore, we wanted to investigate the role of Lys/Arg in possibly complexing the acid/base residue in CloFuc and BbAfcB, in a position where the other three fucosidases have a Ser residue, by the following variants: SpGH29^C S259K (Figures 3b and 4), CpAfc2 S290K, BiAfcB S285K, CloFuc R364S, and BbAfcB K809S.

Figure 3.

Zoom of the active site in the GH29B enzymes from Bifidobacterium longum subsp. infantis, BiAfcB (a, PDB 3UES), and from Streptococcus pneumoniae TIGR4, SpGH29^C (b, PDB 6ORG). (a) BiAfcB with important residues highlighted: the nucleophile Asp and its accompanying Tyr (cyan), the general acid/base Glu (orange), the glycan-stabilizing Trp “hands” (purple), the assisting Asp (pink), the His triad (the adjacent duo in lime green, the His at the end of strand β1 is lime), and the “shovel” Trp (black). The lowest energy docking of the Lewis A antigen glycan is shown, with the GlcNAc (subsite +1*³⁰) in blue, Gal (subsite +2′) in yellow, and Fuc (subsite −1) in red. (b) The nucleophile Asp (cyan) and the general acid/base Glu (orange), and the residues targeted by protein engineering in the current study, namely, residues Val258 (blue), Ser259 (teal), Trp264 (purple), Asp257 (light pink), Ala173 (salmon), and Trp127 (dark pink).

Figure 4.

Zoom of residue interactions (Thr/Val, Ser/Arg/Lys) in the active sites of GH29B fucosidases from the AlphaFold2-models, CloFuc (a), CpAfc2 (b), and BbAfcB (c), and crystal structures, SpGH29^C (d, PDB 6ORG) and BiAfcB (e, PDB 3UES). Residues Thr in BiAfcB and Val in CpAfc2 and SpGH29^C are colored in marine, Ser interacts with Asn in CpAfc2, SpGH29^C and BiAfcB are colored in teal and yellow, respectively, and Arg in CloFuc and Lys in BbAfcB are colored in limon. Hydrogen bonding within residues (or with the backbone of loop 7) is shown as dashed lines.

Previously, BiAfcB has undergone comprehensive mutation to improve its transfucosylation activity toward HMO synthesis. While a few variants were rationalized,²⁶ most remain presented without design rationale in the patent literature.⁵² Among these, the best single mutants were BiAfcB A174H and W135E.^19,52 In the study presented here, we investigated the transferability of these mutations by the following variants: SpGH29^C A173H and W127E (Figure 3b), CpAfc2 A205H and W164E, CloFuc A260H and W181E, and BbAfcB A705H. BbAfcB does not have a residue corresponding to the Trp.

Finally, we wanted to investigate the role of certain active site residues, namely, a Trp that interacts with the fucose moiety, with the fucosyl-enzyme intermediate, and with the acid/base, and an Asp that interacts with the Trp and with the galactose part of the substrate, and may also assist the catalytic acid/base residue, by creating the following variants: SpGH29^C W264F, W264H, D257L, D257N (Figure 3b), and CloFuc W369H, D362L, D362N, and D362A.

Transfucosylation and hydrolytic activity of the five wild-type fucosidases and their corresponding variants were monitored over 2 h using 10 mM 3FL as fucosyl donor and 100 mM LNT as acceptor substrate at certain pH values, i.e., UB pH 7.2 for CpAfc2 and CloFuc, UB pH 4.5 for BbAfcB, SB pH 5.9 for SpGH29C, and SB pH 7.9 for BiAfcB (Figure 5a–e and Tables 3 and 4). At these conditions, high transfucosylation yields and rates were accompanied by high rates of undesired hydrolysis. Comparing WT and mutants at these conditions should enable facile interpretation of the mutant effect, as appreciable rates of the competing hydrolysis and transglycosylation were observed in the WT enzymes.

Figure 5.

Transfucosylation and hydrolysis catalyzed by current GH29B α-l-fucosidase variants. Molar transfucosylation yields ([LNFP II]/[3FL]₀) and hydrolyzed fucose concentration obtained in transglycosylation catalyzed by the wild type and mutants: CloFuc (a), CpAfc2 (b), BbAfcB (c), SpGH29^C (d), and BiAfcB (e) using 3FL as fucosyl donor substrate and LNT as acceptor substrate (n = 2) at 40 °C and corresponding optimal pH (CloFuc and CpAfc2 at UB pH 7.2, BbAfcB at UB pH 4.5, SpGH29^C at SB pH 5.9, and BiAfcB at SB pH 7.9). The symbol coloring corresponds to the type of variant: wild type (red); Val/Thr variants affecting acid/base flexibility (olive); variants targeting Ser/Lys/Arg possibly affecting the acid/base (cyan); variants of substrate- and acid/base-interacting Trp (wine); variants targeting assisting Asp (magenta); and variants transferring the best BiAfcB variants Ala to His (navy) and Trp to Glu (orange). CloFuc and BbAfcB produced mixtures of LNFP II and LNFP V (68/32% at the maximum yield of WT CloFuc after 60 min of reaction in UB pH 7.2, and 91/9% at the maximum yield of WT BbAfcB after 5 min of reaction in UB pH 4.5), while all other enzymes produced LNFP II only (Figure S2).

Table 3. Initial Transfucosylation Rate (r_T), Initial Hydrolysis Rate (r_H), and the Ratio between Initial Transfucosylation Rate and Initial Hydrolysis Rate (T/H), Maximum Product Yield ([LNFP II]) at Corresponding Time Points (min) Catalyzed by the Wild Type and Corresponding Variants of CloFuc, CpAfc2, BbAfcB, SpGH29^C, and BiAfcB Targeting Acid/Base Flexibility and Acid/Base-Interacting Residues Using 10 mM 3FL as Donor Substrate and 100 mM LNT as Acceptor Substrate (n = 2).

enzymes	rT (μM·min–1)	rH (μM·min–1)	T/H	max LNFP II* (mM)	time (min)
	Wild type
CloFuc WT	232 ± 26a	57.5 ± 3.1b	4.03	3.84 ± 0.22a	60
CpAfc2 WT	226 ± 70a	16.8 ± 2.0b	13.5	3.68 ± 0.09a	60
BbAfcB WT	143 ± 17a	1350 ± 90a	0.106	0.835 ± 0.1b	5
SpGH29CWT	766 ± 24a	18.9 ± 1.0b	40.5	9.11 ± 0.53a	30
BiAfcB WT	233 ± 27a	193 ± 20a	1.21	2.26 ± 0.26b	10
	Val/Thr variants affecting acid/base flexibility
CpAfc2 V289T	97.6 ± 0.8b	258 ± 91a	0.378	1.30 ± 0.13b	30
SpGH29CV258T	329 ± 11b	69.7 ± 15.3a	4.72	3.84 ± 0.22c	30
BiAfcB T284V	294 ± 25a	56.8 ± 10.1b	5.18	4.50 ± 0.44a	30
	Variants targeting Ser/Lys/Arg possibly affecting the acid/base
CloFuc R364S	42.5 ± 8.5b	91.5 ± 4.2a	0.464	0.715 ± 0.22b	30
CpAfc2 S290K	3.00 ± 0.0c	0.600 ± 0.0b	5.00	0.345 ± 0.01c	120
BbAfcB K809S	55.7 ± 31.1b	961 ± 567ab	0.0580	0.320 ± 0.15c	5
SpGH29CS259K	6.30 ± 2.3de	1.90 ± 0.3c	3.32	0.745 ± 0.30d	120
BiAfcB S285K	8.20 ± 2.1c	5.75 ± 1.5c	1.43	0.350 ± 0.00c	120
	Variants of substrate- and acid/base-interacting Trp
CloFuc W369H	ND	0.150 ± 0.1e	ND	ND	ND
SpGH29CW264F	183 ± 46c	10.8 ± 1.2bc	16.9	5.84 ± 0.90b	60
SpGH29CW264H	1.10 ± 0.6e	0.750 ± 0.1c	1.47	0.125 ± 0.06d	120
	Variants targeting assisting Asp
CloFuc D362L	ND	0.550 ± 0.4e	ND	ND	ND
CloFuc D362N	4.00 ± 1.6c	9.80 ± 3.4d	0.408	0.490 ± 0.18bc	120
CloFuc D362A	0.100 ± 0.0c	1.75 ± 0.1de	0.0571	0.0100 ± 0.00c	120
SpGH29CD257L	1.25 ± 0.1e	0.900 ± 0.3c	1.39	0.120 ± 0.06d	120
SpGH29CD257N	25.7 ± 5.9de	8.20 ± 0.8bc	3.13	2.92 ± 1.03c	120

^*CloFuc and BbAfcB produced mixtures of LNFP II and LNFP V (68/32% at the maximum yield of WT CloFuc after 60 min of reaction in UB pH 7.2, and 91/9% at the maximum yield of WT BbAfcB after 5 min of reaction in UB pH 4.5), while all other enzymes produced LNFP II only (Figure S2).

^a–eSignificant difference (p < 0.05) between the WT and corresponding variants (including variants in Table 4) for each fucosidase.

ND: no data detected due to the lack of transfucosylation or hydrolysis activity.

Table 4. Initial Transfucosylation Rate (r_T), Initial Hydrolysis Rate (r_H), and Initial Transfucosylation Rate/Initial Hydrolysis Rate (T/H), Maximum Product Yield ([LNFP II]) at Corresponding Time Points (min) Catalyzed by the Wild-Type Enzymes and Corresponding Variants Designed by Transferring the Best BiAfcB Mutations (Ala to His and Trp to Glu) to CloFuc, CpAfc2, BbAfcB, and SpGH29^C Using 10 mM 3FL as Donor Substrate and 100 mM LNT as Acceptor Substrate (n = 2).

enzymes	rT (μM·min–1)	rH(μM·min–1)	T/H	max LNFP II* (mM)	time (min)
CloFuc WT	232 ± 26a	57.5 ± 3.1b	4.03	3.84 ± 0.22a	60
CpAfc2 WT	226 ± 70a	16.8 ± 2.0b	13.5	3.68 ± 0.09a	60
BbAfcB WT	143 ± 17a	1350 ± 90a	0.106	0.835 ± 0.1b	5
SpGH29CWT	766 ± 24a	18.9 ± 1.0b	40.5	9.11 ± 0.53a	30
CloFuc A260H	21.2 ± 0.4bc	36.0 ± 9.1c	0.589	1.10 ± 0.49b	120
CpAfc2 A205H	28.8 ± 6.2bc	ND	ND	3.40 ± 0.70a	120
BbAfcB A705H	24.2 ± 1.4b	3.30 ± 0.3b	7.33	2.74 ± 0.18a	120
SpGH29CA173H	43.4 ± 3.7d	12.8 ± 0.8bc	3.39	5.79 ± 0.99b	120
CloFuc W181E	ND	0.400 ± 0.3e	ND	ND	ND
CpAfc2 W164E	0.300 ± 0.0c	0.250 ± 0.2b	1.20	0.0400 ± 0.00c	120
SpGH29CW127E	3.70 ± 0.4de	2.65 ± 0.4c	1.40	0.424 ± 0.04d	120

^*CloFuc and BbAfcB produced mixtures of LNFP II and LNFP V (68/32% at the maximum yield of WT CloFuc after 60 min of reaction in UB pH 7.2, and 91/9% at the maximum yield of WT BbAfcB after 5 min of reaction in UB pH 4.5), while all other enzymes produced LNFP II only (Figure S2).

^a–eSignificant difference (p < 0.05) between the WT and corresponding variants (including variants in Table 3) for each fucosidase.

ND: no data detected due to the lack of transfucosylation or hydrolysis activity.

Among the five enzymes, WT SpGH29^C maintained the highest transglycosylation rate, T/H value, and maximum LNFP II yield compared with all variants (Figure 5d and Tables 3 and 4). As hypothesized, the V258T variant displayed an increased hydrolysis rate compared to the WT. The molar LNFP II yield obtained with SpGH29^C V258T arrived at the transient maximum 38% LNFP II after 30 min, followed by a sharp decrease due to the dramatic hydrolysis over the whole period. In contrast, the LNFP II produced by variant A173H of SpGH29^C steadily climbed within the 2 h of reaction. Although the maximum yield of 58% achieved was significantly lower than that of WT SpGH29^C (91%), this variant may have the potential to catalyze formation of more targeted fucosylated product within longer time due to the low hydrolytic rate. The variant most closely resembling the WT was W264F, although the substitution affected transglycosylation more negatively (fourfold decrease in rate) than hydrolysis (twofold rate decrease). Interestingly, substitution of the same Trp with His (W264H) dramatically decreased both the hydrolytic rate and, especially, the transglycosylation rate, resulting in a 27-fold decrease in T/H (Figure 5d and Table 3). As similar observations were made for the same variant in CloFuc (Figure 5a and Table 3), this indicates the high importance of the aromatic properties of Trp in this position. Similarly, the negative effect of substituting Asp257 was more pronounced for transglycosylation than for hydrolysis. As expected, the least dramatic change, D257N, had a smaller impact than D257L. Finally, neither the transglycosylation yield nor hydrolytic product was significantly observed in the reactions catalyzed by SpGH29^C variants S259K and W127E (Tables 3 and 4), resulting in around 12-fold and 29-fold lower T/H values compared with that of the WT, respectively.

For CloFuc (Figure 5a and Tables 3 and 4), the transglycosylation activities of all mutants were negatively affected and even practically inactive variants existed, namely, W369H, D362L, D362A, and W181E. Both D362N and A260H remained active in transglycosylation and had gradual increases in LNFP yield over 2 h, reaching 5 and 11%, respectively, coupled with lower hydrolysis rates than the WT, yet their T/H was below 1 due to the markedly reduced transglycosylation rate. The variant R364S preferred hydrolysis to transglycosylation and LNFP levels decreased after reaching the maximum yield of 7% after 30 min.

As observed for SpGH29^C, the transglycosylation yield of CpAfc2 V289T was much lower than that of WT CpAfc2 and even sharply reduced when the maximum LNFP II arrived due to the dramatic secondary hydrolysis, which was evidenced by a rapid increase in released fucose (Figure 5b and Table 3). Both S290K and W164E had low LNFP II yields (3.5 and 0.4%, respectively) and negligible hydrolysis, and both appeared to be crippled enzymes. Importantly, A205H was the best variant as it presented a linear growth in the LNFP II yield and finally exceeded the WT after 2 h of reaction, while the hydrolytic activity was undetectable (Table 4).

Similarly, the BbAfcB A705H variant showed significant improvement of the WT enzyme: Although the transglycosylation rate was six times lower, the hydrolytic rate of BbAfcB A705H was as much as 400 times lower than that of the WT, resulting in continuous production of LNFP, reaching a 27% yield after 2 h (8% after 5 min for WT) and a 70-fold higher T/H (Figure 5c and Tables 3 and 4). BbAfcB K809S showed a low transient maximum of LNFP (3%) at 5 min, which was similar to that of the WT enzyme.

The BiAfcB T284V variant was a significant improvement as compared with the WT (Figure 5e and Table 3). Although BiAfcB T284V retained a behavior similarly to that of the WT BiAfcB, namely, a rapidly achieved transient maximum of LNFP II followed by a fast decrease due to significant hydrolysis, BiAfcB T284V reached an LNFP II yield of 45%, which was twice as high as that of the WT (22.6%). This resulted in the transfucosylation rate being 1.3 times higher, the hydrolysis rate 3.4 times lower, and the resulting T/H thus more than four times higher than that of the WT. While the T/H value of variant S285K was close to that of WT BiAfcB, this mutation crippled the enzyme as also observed for CpAfc2 and SpGH29^C.

Discussion

Effect of pH on GH-Catalyzed Transglycosylation

Various strategies have been employed to improve transglycosylation efficiency and avoid undesirable primary and/or secondary hydrolysis in GHs including GH29 fucosidases, including optimization of the choice of donor and acceptor,⁵³ increased substrate concentrations and A:D ratio,^17,19,25 and protein engineering.^17,18,54 Studies of other GHs indicate that conformational changes in the active site region and ionization of a residue involved in hydrolysis could account for pH-dependent hydrolysis and pH-independent transglycosylation.^55,56 Recently, a GH29A α-l-fucosidase from Pedobacter sp. was reported to have the highest hydrolytic activity at pH 5.0, while maximum formation of fucosyllactose was obtained at pH 8.5.⁵⁷ Examples of pH control as a strategy for improving transglycosylation or reducing hydrolytic activity exist for enzymes from other GH families, e.g., α-amylase mutants showing increased transglycosidic activity at acidic pH⁵⁸ and an engineered β-glucosidase from Thermotoga neapolitana demonstrating eliminated hydrolysis at alkaline pH.⁵¹ Thus, in the current study, HMO core structures LNT and 3FL, were used as acceptor and fucosyl donor, respectively, in transfucosylation catalyzed by GH29B α-l-fucosidases CloFuc, CpAfc2, BbAfcB, SpGH29^C, and BiAfcB to investigate the effect of pH of two different buffer systems on transfucosylation and hydrolysis activity. The significance of pH and the buffer system were enzyme-dependent. SpGH29^C and CpAfc2 exhibited higher molar yields of LNFP II and lower hydrolytic activity compared to the other fucosidases. Furthermore, both fucosidases were more prone to less hydrolytic activity in the UB system in comparison with SB solutions. Although less potent than SpGH29^C and CpAfc2, CloFuc also reached appreciable transfucosylation yields. As previously observed, BbAfcB and BiAfcB were both characterized by high 3FL hydrolysis rates and short-lived transient maxima of LNFP II, although BiAfcB performed better in terms of the LNFP II yield. Importantly, the level of hydrolysis by BiAfcB was markedly reduced at high pH.

Active Site Topology in GH29 Fucosidases

The fucosidases in the GH29B subfamily have high specificity for terminal α-1,3/4-fucosidic linkages⁵⁹ requiring a galactose moiety in the +2′ subsite.³⁰ The high specificity and regioselectivity, as compared to fucosidases from the rest of GH29,³³ is reflected in the remarkable consistency of the structural arrangement of the active site amino acids and their role in tuning enzyme activity. The GH29B fucosidases adapt a widely conserved TIM barrel fold, (β/α)₈, and the active site sits on the inside of the C-terminal end of the inner β-barrel (Figure 3a). The active sites of the five fucosidases studied in this work all hold ensembles of certain residues that play the same role in all known members of GH29B. As retaining GHs, their active sites have a catalytic nucleophile, namely, Asp on strand β4 (Asp172 in BiAfcB and Asp171 in SpGH29^C), and a general acid/base Glu residue on the flexible loop 6 (at the end of strand β6; Glu217 in BiAfcB and Glu215 in SpGH29^C; Figure 3), on the opposite side of the active site relative to the nucleophile. The nucleophile is accompanied by a highly conserved Tyr (Tyr131 in BiAfcB)⁶⁰ on strand β3 (Figure 3a), which forms hydrogen bonds to the nucleophile and restricts it in space to subsite −1 as required for substrate binding and subsequent formation of the fucosyl-enzyme intermediate. In most of the MD trajectories, the heavy atom distance is indicative of H-bonding, and in the MD runs where that distance grows large, the nucleophilic Asp is far from subsite −1 and thus not in a catalytically relevant position (Figure S4c–f).

A triad of His residues plays a role in fucose binding, thereby defining the enzyme specificity. A duo of adjacent His residues are on loop 2 following strand β2, while the third His is on strand β1 (Figure 3a). Besides binding Fuc, the His duo also interacts with loop 1 to control its mobility. On loop 1, strand β1 is a mobile Trp residue, which seems to have the role of “shoveling” or otherwise blocking the substrate from exiting the active site (Figure 3a). The role of the His triad on T/H was recently investigated for GH29A members AlfB and AlfC from Lactobacillus casei.⁶¹ Each His in the duo was found to possess opposite but determinant roles in regulating the T/H ratio. Changing the upstream His to Phe increased transglycosylation yields and T/H significantly, whereas the same modification to the downstream His almost eliminated transglycosylation and severely decreased the T/H.⁶¹ Possibly, the upstream His of the duo has the secondary role of modulating the flexibility of the surrounding environment while the downstream His, which is close to the nucleophile-shepherding Tyr and to the O-2 of Fuc, could be important for the nucleophilic attack—that is, stabilizing conformations of Fuc that are more susceptible to nucleophilic attack, thus effectively lowering the barrier of the relevant transition state (TS).²² The other possibility would be that the downstream His is involved in “tuning” the fucosyl-enzyme intermediate, making it more stable and thus preventing the reaction back to the reactant state. This could result in a thermodynamic sink, which would not necessarily affect the hydrolysis and transglycosylation to the same extent, thus affecting T/H. Substituting the His on strand β1 with Phe completely killed enzyme activity in AlfB,⁶¹ pointing to a crucial role of this His that interacts with O-4 of Fuc (Figure 3a).

The active sites of GH29B also hold two Trp residues that act as glycan “hands” and stabilize the binding of the glycan moieties on either side of the scissile bond (Figure 3a). The Trp on strand β6 interacts with the Gal in subsite +2′, while the Trp on loop 7 interacts with the Fuc in subsite −1 (Figure 3). The Trp on loop 7 will also be important for the hydrolysis and transglycosylation transition states of the enzyme, as it interacts with the fucosyl-enzyme intermediate. The mutagenesis results confirm the importance of Trp on loop 7 in catalysis: CloFuc W369H (Figure 5a and Table 3) lost transglycosylation activity and had a negligible hydrolysis rate, whereas both W264F and W264H of SpGH29^C (Figure 5d, Table 3, and Figure 3b) lowered the molar LNFP II yield, the hydrolysis rate, and T/H as compared with WT.

Upstream of loop 7, Trp sits an Asp residue located on strand β7 (Figure 3), which interacts with the O-6 on the Gal moiety and with the “hand” Trp on strand β6 via hydrogen bonds. This Asp residue plays a role in the binding of glycans, but our MD simulations suggest that it also assists the acid/base residue during catalysis, for example via changes in the water network in the vicinity of the acid/base residue and/or by influencing the protonation event of the linkage oxygen of the scissile bond of the glycan. A hydrogen bond network is observed around this Asp residue, which involves both the Gal O-6, the acid/base residue, and Trp “hands” (Figure S5); this might explain the strict requirement for Gal branching on GH29B substrates. In turn, this relationship between the acid/base Glu and the assisting Asp implies that their relative pK_a values are important for enzyme activity, as the protonation state of the Asp could alter the hydrogen bond network and even Gal binding. The importance of the Asp residue is confirmed by the mutagenesis studies: Changing it to its structurally close relative Asn diminished the hydrolytic rate by an order of magnitude and resulted in 10-fold decreases in T/H for both SpGH29^C and CloFuc, whereas the more dramatic changes to Leu or Ala resulted in rate losses of another order of magnitude with the largest losses observed for the transglycosylation rates (Table 3). The steric effects of the Leu evidently abolish transglycosylation from CloFuc but not entirely from SpGH29^C, perhaps because SpGH29^C has a far more flexible active site. On the other hand, the Asn variants that retain H-bond forming capabilities while being unable to function in an amphoteric role seem to reveal a proton conveying role of assisting Asp. Since the Asp-to-Asn mutations affect transglycosylation more negatively than hydrolysis, this implies that retaining a proton-conveying residue in this position favors transglycosylation more than hydrolysis. Extending the lifetime of the fucosyl-enzyme intermediate stochastically favors the transglycosylation of the less abundant glycoside acceptor. It follows that the protonation of the glycosidic bond oxygen and subsequent abstraction of a proton from the acceptor would shorten the lifetime of said fucosyl-enzyme intermediate, i.e., a well-positioned general acid/base will favor hydrolysis. An intriguing possibility here is that Asp can abstract the proton from the glycoside acceptor instead of the acid/base residue, thus explaining its role in favoring transglycosylation, yet this option remains speculative.^62,63

Increased Acid/Base Loop Flexibility Fosters Transglycosylation

Recent studies on GH1 and GH51 enzymes have indicated that increased flexibility of the acid/base residue, and even increased active site flexibility in general, favors transglycosylation.^51,62,63 Hydrolysis requires the accurate positioning of a water molecule in the active site. In GHs, the catalytic base residue interacts with water to deprotonate it for reaction.⁶⁴ Increased residue flexibility could cause a spatial positioning of water, which is unfavorable for hydrolysis, and it could extend the lifetime of the glycosyl-enzyme intermediate to allow the larger carbohydrate acceptor time to enter the active site^51,63 By comparing the five GH29B structures under study, we observed from the MD simulations that Thr284 in BiAfcB can form a hydrogen bond to the backbone carbonyl of Glu217, which is adjacent to the catalytic acid/base (Figure 4e). This hydrogen bond can reduce the mobility of loop 6, which in addition to placing catalytic water for reaction may also lead to the enzyme populating closed states in which substrates and products have less access to and from the active site. This proposition is supported by our MD simulations of the wild-type vs the mutant variants (Figure S6). Increasing the lifetime of the closed states would favor hydrolysis when water is present in the active site before the aglycone leaves to make room for a larger nucleophile to form product oligosaccharides. In CpAfc2 and SpGH29^C, this Thr is replaced by a Val residue, which is structurally similar but does not have the ability of forming hydrogen bonds (Figure 4b,d). Thus, we hypothesized that replacing Thr with Val in BiAfcB would increase transglycosylation yields and T/H whereas the superior transglycosylation behavior of SpGH29^C and CpAfc2 would decrease when Val is replaced with Thr. Indeed, SpGH29^C V258T and CpAfc2 V289T had a transglycosylation rate, which was 2.3 times lower than that of the WT, whereas the hydrolytic rate increased 3.7 times for SpGH29^C V258T and as much as 15 times for CpAfc2 V289T, leading to 8-fold and 35-fold decreases in T/H, respectively (Table 3). In contrast, BiAfcB T284V exhibited a 4-fold increase in T/H compared to the WT as a result of a 3.4-fold decrease in hydrolytic rate and a slight 1.3-fold increase in transglycosylation rate (Figure 5e and Table 3). Increased flexibility of the acid/base-carrying loop by point mutation was also recently demonstrated to favor transglycosylation in the GH29A fucosidase AlfC.²⁶ The conservation of the residue corresponding to BiAfcB Thr284 reveals that Val or Thr is present in 92% of the sequences in GH29B while residue Ala, which is present in both CloFuc and BbAfcB, is quite rare (1.4% of the sequences; Table S5). We propose that this exchange of Thr with Val is a generic strategy for improving transglycosylation in the 20% of the GH29B enzymes that have Thr in this position, whereas the 72% with a naturally occurring Val may have a natural advantage in transglycosylation.

Another difference observed when comparing the five GH29B structures under study was that the AlphaFold2 models of CloFuc and BbAfcB display Arg and Lys residues, respectively, at the end of strand β7 (Figure 4a,c). The other three fucosidases have Ser in this position (Figure 4b,d,e), and Ser is by far the most common residue in GH29B (89.2% of the sequences; Table S5). Arg found in CloFuc is observed in 1.2% of the sequences, whereas Lys observed in BbAfcB is present in 0.6% of the sequences only (Table S5). In the crystal structures of SpGH29^C and BiAfcB, the role of this Ser seems to be interaction with an Asn at the end of strand β8 and/or with the backbone of loop 7 that holds one of the Trp “hands” (Figure 4d,e). In contrast, the Arg/Lys side chains point toward the active site, and the distance from their terminal side chain nitrogen atom to the acid/base carboxyl oxygen is <3 Å, suggesting interaction (Figure 4a,c). For CpAfc2 and SpGH29^C, replacement of the Ser with Lys had a more detrimental effect on transglycosylation than on hydrolysis (Table 3) whereas the replacement was approximately equally detrimental to both rates in BiAfcB (Table 3). Replacement of Arg/Lys with Ser in CloFuc and BbAfcB was also more severe to transglycosylation than to hydrolysis, resulting in lower T/H values (Table 3). In all cases, swapping residues between these two naturally occurring extremes (Ser and Arg/Lys) was detrimental to the enzyme activity, indicating important, yet different roles of these residues across the enzymes.

Transferability of Successful Mutations between GH29B Enzymes

Furthermore, we have demonstrated that the Trp-to-Glu mutation, which had a highly positive effect on transglycosylation in BiAfcB,^19,52 is not a transferable mutation, as it caused loss of both hydrolysis and transglycosylation in CloFuc, CpAfc2, and SpGH29^C (Figure 5a,b,d and Table 4).

In contrast, the substitution of an Ala residue on loop 4 with a His, which also had highly positive effect on BiAfcB transfucosylation,^19,52 was partially transferable to the other GH29B enzymes. While there was no positive effect on transglycosylation in CloFuc and SpGH29^C, this mutation abolished the hydrolytic activity in CpAfc2 and dramatically reduced it in BbAfcB, leading to increasing product yields throughout the reaction time, and the highest yield of LNFP II obtained with BbAfcB in this work (Table 4 and Figure 5c). The undetectable hydrolytic activity of CpAfc2 A205H removed the transient nature of LNFP II product formation (Figure 5b), which reduces the need for tight reaction time control. To elucidate the effects of this substitution on the active site, MD simulations with this Ala residue mutated to His were performed. As little can be known about the protonation state of this residue, both δ and ε protonations were tested for all the enzymes. In the case of SpGH29^C and BiAfcB, the results clearly indicated that the δ-protonated variant kept the Nu residue in a catalytically relevant position (Figure S4) and thus alleviates the aforementioned problem of a flipped Nu residue (Figure S4).

In conclusion, mutations based on MD simulations strongly indicate the importance of acid–base flexibility, the key role of certain conserved residues, and various interactions in the active site of GH29B α-1,3/4-l-fucosidases. As demonstrated, these enzymes can catalyze the regioselective formation of the important HMO LNFP II from simpler HMOs, which are currently available from cell factories.⁸ At the 91% yield obtained with SpGH29, the cost of substrates 3FL and LNT in the current reaction setup amounts to 21% of the cost of the LNFP II product, when comparing prices at Biosynth Carbosynth. At a yield of 50%, this increases to 38% of the cost. Optimization of A:D and substrate concentration may decrease these percentages. Prices of cell-factory-derived substrates are much lower, and unlike 3FL and LNT, the more complex LNFP II is currently not available from cell factories. Thus, the reaction is readily implementable as a downstream processing option in modern cell factory-based HMO production.

Glossary

Abbreviations used

HMO

human milk oligosaccharide

2′-FL

2′-fucosyllactose

3FL

3-fucosyllactose

LNT

lacto-N-tetraose

LNnT

lacto-N-neotetraose

LNFP II

lacto-N-fucopentaose II

LNFP V

lacto-N-fucopentaose V

Gal

β-d-galactose

Glc

β-d-glucose

GlcNAc

β-d-N-acetylglucosamine

Fuc

α-l-fucose

LacNAc

β-N-acetyllactosamine

MD

molecular dynamics

LB

lysogeny broth

IPTG

isopropyl β-d-1-thiogalactopyranoside

IMAC

immobilized metal affinity chromatography

SB

single buffer

UB

universal buffer 2

A:D

acceptor-to-donor ratio

HPAEC-PAD

high-performance anion exchange chromatography with pulsed amperometric detection

LC-ESI-MS

liquid chromatography-electrospray ionization-mass spectrometry

CUPP

conserved unique peptide patterns

r_T

initial transfucosylation rate

r_H

initial hydrolysis rate

T/H

the ratio between initial transfucosylation rate and initial hydrolysis rate

TS

transition state

Nu

nucleophile.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jafc.4c01547.

• (Figure S1) SDS-PAGE of purified enzymes; (Figure S2) LC-ESI-MS analysis for identification and quantitation of LNFP reaction products; (Figure S3) zoom of the loop 1 interaction with loop 7 in SpGH29^C; (Figure S4) evolution of certain active site differences in SpGH29^C during MD simulations (Figure S5) a closeup on the H-bonding network around Gal O-6 in SpGH29^C; (Figure S6) evolution of certain distances in MD runs showing the effect of mutation T284V in BiAfcB and V258T in SpGH29^C; (Table S1) microbial origin, accession number, and vectors used for fucosidase production; (Table S2) designed primers for site-directed mutagenesis; (Table S3) actual pH values experienced by the fucosidases in the transglycosylation reactions; (Table S4) expected sizes and expression yields for all fucosidases expressed in the current work; (Table S5) conservation of investigated residues (Val/Thr and Ser/Arg/Lys) in GH29B fucosidases; and molecular dynamics (MD) and constant pH MD including (Table S6) selected results from constant pH MD runs (PDF)

The authors declare no competing financial interest.