β- N -Acetylhexosaminidases for Carbohydrate Synthesis via Trans-Glycosylation

: β- N -acetylhexosaminidases (EC 3.2.1.52) are retaining hydrolases of glycoside hydrolase family 20 (GH20). These enzymes catalyze hydrolysis of terminal, non-reducing N -acetylhexosamine residues, notably N -acetylglucosamine or N -acetylgalactosamine, in N -acetyl-β-D-hexosaminides. In nature, bacterial β- N -acetylhexosaminidases are mainly involved in cell wall peptidoglycan synthesis, analogously, fungal β- N -acetylhexosaminidases act on cell wall chitin. The enzymes work via a distinct substrate-assisted mechanism that utilizes the 2-acetamido group as nucleophile. Curiously, the β- N -acetylhexosaminidases possess an inherent trans-glycosylation ability which is potentially useful for biocatalytic synthesis of functional carbohydrates, including biomimetic synthesis of human milk oligosaccharides and other glycan-functionalized compounds. In this review, we summarize the reaction engineering approaches (donor substrate activation, additives, and reaction conditions) that have proven useful for enhancing trans-glycosylation activity of GH20 β- N -acetylhexosaminidases. We provide comprehensive overviews of reported synthesis reactions with GH20 enzymes, including tables that list the specific enzyme used, donor and acceptor substrates, reaction conditions, and details of the products and yields obtained. We also describe the active site traits and mutations that appear to favor trans-glycosylation activity of GH20 β- N -acetylhexosaminidases. Finally, we discuss novel protein engineering strategies and suggest potential “hotspots” for mutations to promote trans-glycosylation activity in GH20 for efficient synthesis of specific functional carbohydrates and other glyco-engineered products.


Introduction
N-acetylhexosamines are important constituents in several biological and biochemically significant structures. The most abundant representative is the 2-acetamido-2-deoxy derivative of glucose, namely N-acetylglucosamine (GlcNAc), which is the main constituent of chitin-the second most abundant biopolymer on earth. Chitin is found in the exoskeletons of arthropods (e.g., shrimp, crabs, and insects) and in fungal cell walls [1,2]. Furthermore, both GlcNAc and its epimer N-acetylgalactosamine (GalNAc) are essential constituents of protein glycosylation structures in eukaryotes as they form part of the glycan core structures.
β-N-acetylhexosaminidases (EC 3.2.1.52) of glycoside hydrolase family 20 (GH20) catalyze the hydrolytic removal of N-acetylhexosamines from the non-reducing end of N-acetyl-β-D-hexosaminides and may act on either N-acetylglucosides or N-acetylgalactosides (including chains of mixed glucosides as long as either GlcNAc or GalNAc is in the non-reducing end) [3].

GH20 β-N-Acetylhexosaminidases
High-value GlcNAc-containing oligosaccharides or glycosylated molecules can be synthesized using GH20 β-N-acetylhexosaminidases. These enzymes can be found throughout the whole tree of life. However, in the CAZy database [25] there are 16 times more bacterial GH20 β-N-acetylhexosaminidase sequences than from any other domain of life. In contrast, there are significantly more studies using fungal β-N-acetyl hexosaminidases than of any other origin, which is due to the eminent contribution to this field by Vladimír Křen and colleagues.
Furthermore, GH20 β-N-acetylhexosaminidases are usually well-expressed in a recombinant host (either cytosolic or secreted) or can be easily purified from wild-type cultivations since these enzymes are often secreted by their natural host organism. This facilitates simple purification of the enzymes for biocatalytic purposes, which was applied in >95% of the literature studied for this article.
Mechanistically, enzymes from family GH20 are special. In contrast to most other glycoside hydrolases (GHs), the hydrolytic cleavage of the glycosidic linkage to the GlcNAc residues is catalyzed via a so-called substrate-assisted mechanism ( Figure 2) [26][27][28][29][30]. In this case, the reaction intermediate is not bound to the enzyme as is the case for the classical Koshland mechanism [31]. The

GH20 β-N-Acetylhexosaminidases
High-value GlcNAc-containing oligosaccharides or glycosylated molecules can be synthesized using GH20 β-N-acetylhexosaminidases. These enzymes can be found throughout the whole tree of life. However, in the CAZy database [25] there are 16 times more bacterial GH20 β-N-acetylhexos-aminidase sequences than from any other domain of life. In contrast, there are significantly more studies using fungal β-N-acetyl hexosaminidases than of any other origin, which is due to the eminent contribution to this field by Vladimír Křen and colleagues.
Furthermore, GH20 β-N-acetylhexosaminidases are usually well-expressed in a recombinant host (either cytosolic or secreted) or can be easily purified from wild-type cultivations since these enzymes are often secreted by their natural host organism. This facilitates simple purification of the enzymes for biocatalytic purposes, which was applied in >95% of the literature studied for this article.
Mechanistically, enzymes from family GH20 are special. In contrast to most other glycoside hydrolases (GHs), the hydrolytic cleavage of the glycosidic linkage to the GlcNAc residues is catalyzed via a so-called substrate-assisted mechanism ( Figure 2) [26][27][28][29][30]. In this case, the reaction intermediate is not bound to the enzyme as is the case for the classical Koshland mechanism [31]. The reactive  Figure 2). Whether the intermediate in the GH20 catalysis is actually the oxazolinium ion as proposed for GH84 O-GlcNAcases [32] or the uncharged glucoxazoline (Glc-oxa) as proposed for GH18 chitinases [33] is not fully clarified yet. However, it is tempting to speculate that due to the close resemblance of the catalytic mechanisms and motifs of enzymes from families GH20 and GH84 (adjacent Asp-Glu pair), and their proposed substrate-assisted catalytic mechanism, the formation of a glucoxazolinium ion is likely. Nucleophilic attack of water then leads to release of GlcNAc as a reaction product (hydrolysis, Figure 2: R 1 H, R 2 = H). During this attack the water molecule is stabilized in the active site by a conserved Tyr residue [34]. However, it is a general feature of many retaining GHs that other nucleophiles such as carbohydrates or alcohols are also accepted, which then leads to a glycosylated product (trans-glycosylation, Figure 2: R 1 H, R 2 H) [35]. Most of the CAZymes acting on GlcNAc residues follow this mechanism (e.g., chitinases (GH18), O-GlcNAcases (GH84), endo-β-N-acetylglucosaminidases (GH85)) [3,[36][37][38]. Only hexosaminidases from family GH3 [39] and the recently discovered GH136 lacto-N-biosidase LnbX from Bifidobacterium longum subsp. longum [40] were shown to catalyze cleavage of their substrates utilizing the classical double-displacement mechanism involving an enzyme-coupled intermediate.  Figure 2). Whether the intermediate in the GH20 catalysis is actually the oxazolinium ion as proposed for GH84 O-GlcNAcases [32] or the uncharged glucoxazoline (Glc-oxa) as proposed for GH18 chitinases [33] is not fully clarified yet. However, it is tempting to speculate that due to the close resemblance of the catalytic mechanisms and motifs of enzymes from families GH20 and GH84 (adjacent Asp-Glu pair), and their proposed substrate-assisted catalytic mechanism, the formation of a glucoxazolinium ion is likely. Nucleophilic attack of water then leads to release of GlcNAc as a reaction product (hydrolysis, Figure 2: R 1 ≠ H, R 2 = H). During this attack the water molecule is stabilized in the active site by a conserved Tyr residue [34]. However, it is a general feature of many retaining GHs that other nucleophiles such as carbohydrates or alcohols are also accepted, which then leads to a glycosylated product (trans-glycosylation, Figure 2: R 1 ≠ H, R 2 ≠ H) [35]. Most of the CAZymes acting on GlcNAc residues follow this mechanism (e.g., chitinases (GH18), O-GlcNAcases (GH84), endo-β-N-acetylglucosaminidases (GH85)) [3,[36][37][38]. Only hexosaminidases from family GH3 [39] and the recently discovered GH136 lacto-N-biosidase LnbX from Bifidobacterium longum subsp. longum [40] were shown to catalyze cleavage of their substrates utilizing the classical doubledisplacement mechanism involving an enzyme-coupled intermediate.

Figure 2.
Proposed substrate-assisted mechanism of GH20 β-N-acetylhexosaminidases [29]. Indication of the intermediate as a Glc-oxazolinium ion is based on recent computational studies on a GH84 O-GlcNAcase probably utilizing the same mechanism [32].
In nature, glycans, oligo-and polysaccharides as well as other glycosylated molecules are synthesized by glycosyltransferases (GTs), which strictly require a nucleotide-activated derivative as substrate. Therefore, these enzymes are not very attractive for biocatalytic reactions, because the prices for nucleotide-activated sugars are exceptionally high (1,220,000 €/kg (795,000 €/mol) for UDP-GlcNAc at Carbosynth) [41]. Furthermore, functional expression of GTs can be challenging. In contrast, GHs with trans-glycosylase activity are more attractive for large-scale in vitro synthesis of such valuable products. GHs have a broader substrate acceptance and can use non-natural activated carbohydrates or natural disaccharides as donor molecules. However, challenges are also encountered with GH trans-glycosylases. Firstly, the regioselectivity of GH-catalyzed transglycosylation reactions can be rather low, which leads to a mixture of the desired product and undesired side products. Secondly, GH trans-glycosylases are in fact hydrolytic enzymes or are derived from such. They are usually still able to hydrolyze both the substrate and the desired product (secondary hydrolysis) [35,42]. However, as outlined in the following, several strategies may be employed to increase trans-glycosylation activity and/or diminish hydrolytic activity in general. Proposed substrate-assisted mechanism of GH20 β-N-acetylhexosaminidases [29]. Indication of the intermediate as a Glc-oxazolinium ion is based on recent computational studies on a GH84 O-GlcNAcase probably utilizing the same mechanism [32].
In nature, glycans, oligo-and polysaccharides as well as other glycosylated molecules are synthesized by glycosyltransferases (GTs), which strictly require a nucleotide-activated derivative as substrate. Therefore, these enzymes are not very attractive for biocatalytic reactions, because the prices for nucleotide-activated sugars are exceptionally high (1,220,000 €/kg (795,000 €/mol) for UDP-GlcNAc at Carbosynth) [41]. Furthermore, functional expression of GTs can be challenging. In contrast, GHs with trans-glycosylase activity are more attractive for large-scale in vitro synthesis of such valuable products. GHs have a broader substrate acceptance and can use non-natural activated carbohydrates or natural disaccharides as donor molecules. However, challenges are also encountered with GH trans-glycosylases. Firstly, the regioselectivity of GH-catalyzed trans-glycosylation reactions can be rather low, which leads to a mixture of the desired product and undesired side products. Secondly, GH trans-glycosylases are in fact hydrolytic enzymes or are derived from such. They are usually still able to hydrolyze both the substrate and the desired product (secondary hydrolysis) [35,42]. However, as outlined in the following, several strategies may be employed to increase trans-glycosylation activity and/or diminish hydrolytic activity in general.

Increased Trans-Glycosylation Activity by Reaction Engineering
Since synthesis using GHs can be compromised by secondary hydrolysis and low regioselectivity, a number of reaction engineering strategies, notably including various donor molecule activation approaches, have been reported to help increase yields of GH20-catalyzed trans-glycosylation reactions. Surprisingly, immobilization or continuous product removal approaches, which have been successfully applied for other GH-catalyzed trans-glycosylation reactions [42], have not yet been used to optimize GH20 catalyzed trans-glycosylation reactions.

Reverse Hydrolysis VS. Trans-Glycosylation
In general, GH-catalyzed synthesis of oligosaccharides or other glycosylated products can be carried out in two different modes of reaction: reverse hydrolysis or trans-glycosylation.
In a reverse hydrolysis reaction, the GH catalyzes a condensation reaction of two carbohydrates (where the activated donor cannot be distinguished among the two reactants) to yield specific disaccharides (Table 1, Figure 2: R 1 = H and R 2 H). The formation of glycosylated amino acids and alkyl glycosides by enzyme-catalyzed reverse hydrolysis have also been described (Table 1). To the best of our knowledge, the term reverse hydrolysis was first used in a review from 1986 [43] and later in the context of glucose-disaccharide formation by β-glucosidase from almond [44], where the authors demonstrated that a high substrate load is required and that a high reaction temperature favors product formation. Ten years later, the first reports on GH20-catalyzed reactions involving the hexosaminidases from Bacillus circulans (BcHex) and Aspergillus oryzae (AoHex) via reverse hydrolysis setup were published [45,46]. Clear disadvantages of such reverse hydrolysis reactions are the long reaction times (72-360 h, Table 1) and the low yields, which rarely exceed 15% isolated yield [23] though higher yields (up to 46%, Table 1) have been reported for non-isolated product outcomes [47].

p-Nitrophenyl Activated Donors
The p-nitrophenyl (pNP) derivates of hexosaminides ( Figure 3: pNP-GlcNAc) are by far the most popular donor molecules for trans-glycosylation reactions (Tables 3-6). The pNP-glycosides were developed as colorimetric substrates to study hydrolytic GH reactions, because the released pNP can be detected photometrically at λ = 405 nm. However, due to their low price (47,250 €/kg (16,000 €/mol) for pNP-GlcNAc at Carbosynth) [68] and not least the fact that pNP is a good leaving group, they have become attractive as donor molecules for trans-glycosylation reactions. Furthermore, the pNPhexosaminides are accepted by a wide range of GH20 enzymes as substrate or donor molecule, including fungal (Tables 3 and 4), bacterial (Table 5), and enzymes of other origin (Table 6). Indeed, the first attempt at synthesizing a GlcNAc-containing HMO (lacto-N-tetraose (LNT)) utilized a pNPactivated lacto-N-biose as donor (Table 5) [69]. However, despite their popularity, the pNPhexosaminides are not suitable as substrates for synthesis of food products (e.g., HMOs) or in other highly regulated fields, due to the toxicity of the released pNP [70].

Other Activated Donors
In addition to the phenyl-activated donors [48,65,71], other synthetic hexosaminide donor derivatives (Tables 2 and 7) have been used. These have mainly been employed to increase solubility of the donor (e.g., the 2-hydroxy-3-nitro-pyridyl activated GlcNAc ( Figure 3: GlcNAc-NPy; Table 7)) [72] or to increase yield and regioselectivity (e.g., the o-nitrophenyl (oNP) derivate of α-GlcNAc ( Figure 3: oNP-GlcNAc; Table 7)) [73]. Moreover, the glycosyl azide of GlcNAc ( Figure 3: GlcNAc-N3) has been demonstrated to be a superior donor giving higher yields compared to the conventional pNP-GlcNAc, due to its higher solubility (Tables 2 and 7) [66]. However, since the released byproducts from these donor molecules may be irritant (e.g., 2-hydroxy-3-nitro-pyridine) [74], pose an environmental hazard (e.g., oNP) [75], or be acutely toxic (e.g., phenol from Ph-activated [76] and azide from N3-activated donors [77]), none of these activated donor molecules are feasible for use in synthesis of food components or in similarly regulated applications.  (Tables 3-6). The pNP-glycosides were developed as colorimetric substrates to study hydrolytic GH reactions, because the released pNP can be detected photometrically at λ = 405 nm. However, due to their low price (47,250 €/kg (16,000 €/mol) for pNP-GlcNAc at Carbosynth) [68] and not least the fact that pNP is a good leaving group, they have become attractive as donor molecules for trans-glycosylation reactions. Furthermore, the pNP-hexosaminides are accepted by a wide range of GH20 enzymes as substrate or donor molecule, including fungal (Tables 3 and 4), bacterial (Table 5), and enzymes of other origin (Table 6). Indeed, the first attempt at synthesizing a GlcNAc-containing HMO (lacto-N-tetraose (LNT)) utilized a pNP-activated lacto-N-biose as donor (Table 5) [69]. However, despite their popularity, the pNP-hexosaminides are not suitable as substrates for synthesis of food products (e.g., HMOs) or in other highly regulated fields, due to the toxicity of the released pNP [70].

Oxazoline-Derivates as Activated Donors
Recently, the oxazoline derivative of GlcNAc (Figure 3: Glc-oxa) has been demonstrated to work as donor molecule for GH20-catalyzed synthesis of the HMO-precursor molecule lacto-N-triose II (LNT II, Figure 1) using a mutant of the enzyme Hex1 [108], which was previously discovered in a metagenomic library [109]. Glc-oxa is the natural intermediate of a GH20 reaction and can easily be synthesized using high temperature and alkali [110]. However, oxazolines are unstable at acidic conditions and less useful, eg, in fungal hexosaminidase reactions having acidic pH optima [64]. Nevertheless, the ability of bacterial hexosaminidases with neutral or alkaline pH optima to utilize Glc-oxa as donor for trans-glycosylation reactions as demonstrated (Table 9) [24,101,105,108,111] together with its comparably good price per mole (1,710,000 €/kg (347,000 €/mol) for Glc-oxa at Carbosynth [112]) may increase the popularity of Glc-oxa as donor molecule for such enzymatic synthesis reactions in the future. The fact that there are no byproducts from trans-glycosylation reactions using Glc-oxa and the option of conducting Glc-oxa synthesis under mild conditions [113,114] with enzymatic trans-glycosylation in a cascade fashion [108] might lead to industrialization of GH20-catalyzed trans-glycosylation with Glc-oxa donors.

Acceptor:Donor Ratio
Another general strategy to increase trans-glycosylation yields is to increase the acceptor:donor ratio (A:D ratio). This approach has proven successful for many trans-glycosylation reactions involving different enzymes as reviewed in [42]. The rationale of this strategy is the direct dependency of the ratio between trans-glycosylation rate and hydrolysis rate on the acceptor concentration [42]. The method has also been widely applied for GH20-catalyzed trans-glycosylation reactions (Tables 1 and 3-10). For carbohydrate acceptor substrates, the A:D ratio is somewhat limited by substrate solubility, but A:D ratios of up to 100:1 have been reported (Tables 3 and 5) [51,83]. Higher ratios of up to 910:1 have been achieved with non-carbohydrate acceptors like alcohols (Table 10) [71], which are usually highly soluble. However, apart from the auto-condensation reactions, for which the A:D ratio is always considered 1:1 (Table 2), a few examples of GH20 trans-glycosylation reactions using an A:D ratio of 1:1 also exist (Tables 1, 3, 4, 9 and 10) [19,23,24,46,52,53,64,88,90,101,115,116]. Most noteworthy in this context is the recently reported synthesis of LNT II where an isolated yield of 86% (281 g/L) was achieved in a reaction using Lac and Glc-oxa in a ratio of 1:1 at a concentration of 600 mM each (Table 9) [100].      [51] NoHex (Nocardia orientalis   [24] A few examples of successful reactions with A:D ratios <1 (as low as 1:10) have been reported (Tables 4, 5 and 8) [20,24,61,85,88,100,[118][119][120] using carbohydrates as acceptors. The most extreme A:D ratio of 1:5000 was used in a thioglycoligase reaction carried out with a mutant of the hexosaminidase from Streptomyces plicatus, which was able to S-glycosylate small peptides and proteins [24]. Such a low A:D ratio could only be achieved because of the different type of reaction.

pH and Temperature Modifying Trans-Glycosylation Activity
In contrast to other trans-glycosylases [123], there are no specific reports of increased trans-glycosylation activity for GH20s by modifying the pH. However, two cases of independent studies using varying conditions for the same enzyme in similar reactions exist. Three independent studies investigated the auto-condensation of pNP-GlcNAc catalyzed by AoHex at different pH and temperature conditions [53,54,56], resulting in varying regioselectivity ( Table 2). Whereas the reaction at neutral pH (pH 7.0) and slightly elevated temperature (50 • C) led to the β-1,3-linked product only [53], the main product of reactions at slightly acidic pH (pH 5.5-6.0) and moderate temperatures (35-37 • C) was the β-1,4-linked product with the β-1,6-linked dimer as side product [54,56]. The higher trans-glycosylation yields obtained in the study using slightly acidic pH [56] are probably due to the higher substrate concentration and solvent addition.
Four independent studies have reported synthesis of LNT II from pNP-GlcNAc and Lac (Figure 4) using the enzyme BbhI from Bifidobacterium bifidum (Table 5) [101][102][103][104]. Two of the studies used identical conditions for the reaction (pH 5.8, 55 • C, A:D = 20:1, 20% (v/v) DMSO) and reported similar yields: Both, Chen et al. [103] and Schmölzer et al. [101] thus reported a yield of approximately 45%. In another study, the synthesis of LNT II by BbhI was carried out at neutral pH (pH 7.0) and 37 • C, with an A:D ratio of 4:1 and without DMSO, leading to a yield of 16% (based on HPLC analysis using an internal standard) after 0.6 h [102]. In the other two studies, the D746A and D746E mutants of BbhI, which have been used for LNT II synthesis at pH 5.8 and at two different temperatures of 37 and 55 • C [101,104]. For the mutant D746A, the reaction at the lower temperature (37 • C) resulted in a higher yield of 58% [104] compared to 40% in the other study, where the reaction has been performed at 55 • C [101]. In the case of D746E the opposite trend was observed. The reaction at 37 • C gave a slightly lower yield of 63% [104] compared to 71% [101]. Overall, the addition of DMSO and a lower pH seem to favor trans-glycosylation activity for this specific reaction. A few examples of successful reactions with A:D ratios <1 (as low as 1:10) have been reported (Tables 4, 5 and 8) [20,24,61,85,88,100,[118][119][120] using carbohydrates as acceptors. The most extreme A:D ratio of 1:5000 was used in a thioglycoligase reaction carried out with a mutant of the hexosaminidase from Streptomyces plicatus, which was able to S-glycosylate small peptides and proteins [24]. Such a low A:D ratio could only be achieved because of the different type of reaction.

pH and Temperature Modifying Trans-Glycosylation Activity
In contrast to other trans-glycosylases [123], there are no specific reports of increased transglycosylation activity for GH20s by modifying the pH. However, two cases of independent studies using varying conditions for the same enzyme in similar reactions exist. Three independent studies investigated the auto-condensation of pNP-GlcNAc catalyzed by AoHex at different pH and temperature conditions [53,54,56], resulting in varying regioselectivity ( Table 2). Whereas the reaction at neutral pH (pH 7.0) and slightly elevated temperature (50 °C) led to the β-1,3-linked product only [53], the main product of reactions at slightly acidic pH (pH 5.5-6.0) and moderate temperatures (35-37 °C) was the β-1,4-linked product with the β-1,6-linked dimer as side product [54,56]. The higher trans-glycosylation yields obtained in the study using slightly acidic pH [56] are probably due to the higher substrate concentration and solvent addition.
Four independent studies have reported synthesis of LNT II from pNP-GlcNAc and Lac ( Figure  4) using the enzyme BbhI from Bifidobacterium bifidum (Table 5) [101][102][103][104]. Two of the studies used identical conditions for the reaction (pH 5.8, 55 °C, A:D = 20:1, 20% (v/v) DMSO) and reported similar yields: Both, Chen et al. [103] and Schmölzer et al. [101] thus reported a yield of approximately 45%. In another study, the synthesis of LNT II by BbhI was carried out at neutral pH (pH 7.0) and 37 °C, with an A:D ratio of 4:1 and without DMSO, leading to a yield of 16% (based on HPLC analysis using an internal standard) after 0.6 h [102]. In the other two studies, the D746A and D746E mutants of BbhI, which have been used for LNT II synthesis at pH 5.8 and at two different temperatures of 37 and 55 °C [101,104]. For the mutant D746A, the reaction at the lower temperature (37 °C) resulted in a higher yield of 58% [104] compared to 40% in the other study, where the reaction has been performed at 55 °C [101]. In the case of D746E the opposite trend was observed. The reaction at 37 °C gave a slightly lower yield of 63% [104] compared to 71% [101]. Overall, the addition of DMSO and a lower pH seem to favor trans-glycosylation activity for this specific reaction.

Increased Trans-Glycosylation Activity by Enzyme Engineering
As an alternative to reaction engineering, trans-glycosylation yields can be increased by enzyme engineering. In contrast to the amount of available data for natural GH20 trans-glycosylases, there are only a few studies on engineering these enzymes. Nevertheless, based on the described mutations it is possible to deduce some general mutation guidelines for GH20 β-N-acetylhexosaminidase engineering to increase their trans-glycosylation activity. For an overview of hexosaminidase enzyme engineering efforts to improve other protein characteristics (e.g., thermal stability) please refer to the review by Slámová and Bojarová [137].

Mutation of the Water-Stabilizing Tyr
The first engineering study on a GH20 enzyme was carried out on Tf Hex from Talaromyces flavus only five years ago [64]. In this work, the authors were inspired by a previous mutational study on a GH85 endo-β-N-acetylglucosaminidase, which revealed that mutation of the water stabilizing conserved Tyr residue in the active site to Phe led to increased trans-glycosylation activity and diminished hydrolytic activity [34]. Mutating Y470 in Tf Hex ( Figure 5A) in a similar manner to Phe led to one of the highest yields (41%) reported for GH20 β-N-acetylhexosaminidase catalyzed auto-condensation reactions ( Table 2) [64]. Introduction of the hetero-aromatic His residue in this position also increased the trans-glycosylation yield, though to a somewhat lower degree (Table 2). Interestingly, both mutations also shifted the product spectrum towards longer chito-oligosaccharides ((GlcNAc) 3 and (GlcNAc) 4 ). The Y470N mutation, which was inspired by the natural presence of Asn in the same position of closely related GH84 β-N-acetylglucosaminidases, even led to synthesis of insoluble oligomers ((GlcNAc) 7 and longer, Table 2). However, transferring a similar mutation (Tyr to Phe) to the bacterial enzymes BbhI and LnbB (both from Bifidobacterium bifidum) turned out to be less beneficial than in Tf Hex (Table 5) [101,105]. In the case of BbhI, the yield of the trans-glycosylation product was increased only to a minor extent when using the pNP-activated donor substrate due to the persistence of secondary hydrolysis. However, when using Glc-oxa as donor at least the synthesis of LNT II by BbhI-Y827F showed a 1.4-fold higher yield (80%) compared to the wild-type (WT) ( Table 9) [101]. In contrast, the LnbB-Y419F mutant showed an opposite trend: Whereas the formation of LNT from Lac and LNB-oxa by LnbB-Y419F resulted in a 45% lower yield of the desired product compared to the WT (Table 9), the use of pNP-LNB as donor led to a 1.6-fold higher yield (Table 5) [105]. In summary, mutating the water-stabilizing Tyr to a Phe residue can be a successful strategy to increase trans-glycosylation activity in GH20 enzymes depending on the substrates and enzymes used.
Catalysts 2020, 10, 365 25 of 37 led to one of the highest yields (41%) reported for GH20 β-N-acetylhexosaminidase catalyzed autocondensation reactions (Table 2) [64]. Introduction of the hetero-aromatic His residue in this position also increased the trans-glycosylation yield, though to a somewhat lower degree (Table 2). Interestingly, both mutations also shifted the product spectrum towards longer chitooligosaccharides ((GlcNAc)3 and (GlcNAc)4). The Y470N mutation, which was inspired by the natural presence of Asn in the same position of closely related GH84 β-N-acetylglucosaminidases, even led to synthesis of insoluble oligomers ((GlcNAc)7 and longer, Table 2). However, transferring a similar mutation (Tyr to Phe) to the bacterial enzymes BbhI and LnbB (both from Bifidobacterium bifidum) turned out to be less beneficial than in TfHex (Table 5) [101,105]. In the case of BbhI, the yield of the trans-glycosylation product was increased only to a minor extent when using the pNP-activated donor substrate due to the persistence of secondary hydrolysis. However, when using Glc-oxa as donor at least the synthesis of LNT II by BbhI-Y827F showed a 1.4-fold higher yield (80%) compared to the wild-type (WT) ( Table 9) [101]. In contrast, the LnbB-Y419F mutant showed an opposite trend: Whereas the formation of LNT from Lac and LNB-oxa by LnbB-Y419F resulted in a 45% lower yield of the desired product compared to the WT (Table 9), the use of pNP-LNB as donor led to a 1.6-fold higher yield (Table 5) [105]. In summary, mutating the water-stabilizing Tyr to a Phe residue can be a successful strategy to increase trans-glycosylation activity in GH20 enzymes depending on the substrates and enzymes used.

Mutation of the Aglycone Binding Site
Another possibility to increase trans-glycosylation activity of fungal enzymes was presented at the 13 th Carbohydrate Biotechnology Meeting in Toulouse, France, in 2019 by researchers from the Křen group: Mutation of the aglycone binding residues V306 and F453 in AoHex ( Figure 5B) to Trp led to an increased trans-glycosylation activity and reduced hydrolytic activity [140]. Due to the close relatedness (78% similarity) the same mutation should be possible in TfHex ( Figure 5A) and other fungal β-N-hexosaminidases. However, it seems that this strategy is not applicable to bacterial GH20 enzymes since these have a slightly different aglycone binding site topology, as discussed below.

Mutation of the Aglycone Binding Site
Another possibility to increase trans-glycosylation activity of fungal enzymes was presented at the 13th Carbohydrate Biotechnology Meeting in Toulouse, France, in 2019 by researchers from the Křen group: Mutation of the aglycone binding residues V306 and F453 in AoHex ( Figure 5B) to Trp led to an increased trans-glycosylation activity and reduced hydrolytic activity [140]. Due to the close relatedness (78% similarity) the same mutation should be possible in Tf Hex ( Figure 5A) and other fungal β-N-hexosaminidases. However, it seems that this strategy is not applicable to bacterial GH20 enzymes since these have a slightly different aglycone binding site topology, as discussed below.

Mutation of the Catalytic Asp-Glu Pair
A general and popular strategy to create so called glycosynthases was introduced by Stephen Withers and colleagues over 20 years ago. They demonstrated that a crippled CAZyme, in which the catalytic nucleophile was mutated to a non-functional Ala residue, is still able to catalyze trans-glycosylation when using a properly activated donor molecule [141]. Due to their relatively simple way of preparation, the glycosyl fluorides became popular as donor molecules for this approach. However, the fluoride approach is not applicable in industrial applications, and was moreover never followed for GH20 enzymes because of the significantly different reaction mechanism. While Glc-oxa should in theory be a properly activated donor molecule, its low stability at low pH hinders the creation of GH20 glycosynthases from fungal β-N-acetylhexosaminidases since these require a low pH for optimal activity (Tables 3 and 4). The discovery of bacterial GH20 enzymes with pH optimum ≥7.0 (e.g., Hex1, Table 8) [109] paved the way for the glycosynthase approach in GH20 by mutating the catalytic Asp. The power of such a GH20 glycosynthase for LNT II synthesis ( Figure 6) from Lac and Glc-oxa (A:D = 1:1, concentration of both = 600 mM) was recently convincingly demonstrated using the BbhI-D746E mutant, which resulted in 86% isolated yield (281 g/L) in a 30 min reaction at pH 7.5 ( Table 9) [101]. This result was a 1.5-fold yield increase compared to the WT (58%; Table 9). Alternative mutations of the same residue (D746A and D746Q) led to real glycosynthases with completely abolished hydrolytic activity, but also to significantly slower enzymes, which was reflected in the lower yield after a longer reaction time (Table 9). Surprisingly, transfer of these mutations to LnbB was not as successful. Both mutants (D320E and D320A, Table 9) were significantly slower in synthesis of LNT from Lac and LNB-oxa ( Figure 6) leading to less than 50% of the wild-type yield in more than ten times longer reactions [105]. Finally, a similar 1.5-fold increase in yield compared to the WT was obtained with BbhI-D746E using a pNP-donor molecule, but the obtained yield was much lower (18%; Table 5) [101]. Recently, a new mutational study on BbhI combining directed evolution and site saturation mutagenesis (SSM) showed that the BbhI-D746T led to an almost doubled yield of LNT II when using Lac and pNP-GlcNAc as substrates. The maximum reported yield was 85% (Table 5) [104]. Furthermore, we can conclude from the SSM of position D746 that introduction of any residue bigger than Glu lead to complete inactivation of the enzyme, probably due to steric hindrance of substrate binding, (Table 5). Similarly, introduction of the two amino acid amides Asn and Gln in this position also lead to enzyme inactivation, although varying results were obtained in two independent studies on BbhI-D746Q (Table 5).

Mutation of the Catalytic Asp-Glu Pair
A general and popular strategy to create so called glycosynthases was introduced by Stephen Withers and colleagues over 20 years ago. They demonstrated that a crippled CAZyme, in which the catalytic nucleophile was mutated to a non-functional Ala residue, is still able to catalyze transglycosylation when using a properly activated donor molecule [141]. Due to their relatively simple way of preparation, the glycosyl fluorides became popular as donor molecules for this approach. However, the fluoride approach is not applicable in industrial applications, and was moreover never followed for GH20 enzymes because of the significantly different reaction mechanism. While Glc-oxa should in theory be a properly activated donor molecule, its low stability at low pH hinders the creation of GH20 glycosynthases from fungal β-N-acetylhexosaminidases since these require a low pH for optimal activity (Tables 3 and 4). The discovery of bacterial GH20 enzymes with pH optimum ≥7.0 (e.g., Hex1, Table 8) [109] paved the way for the glycosynthase approach in GH20 by mutating the catalytic Asp. The power of such a GH20 glycosynthase for LNT II synthesis ( Figure 6) from Lac and Glc-oxa (A:D = 1:1, concentration of both = 600 mM) was recently convincingly demonstrated using the BbhI-D746E mutant, which resulted in 86% isolated yield (281 g/L) in a 30 min reaction at pH 7.5 (Table 9) [101]. This result was a 1.5-fold yield increase compared to the WT (58%; Table 9). Alternative mutations of the same residue (D746A and D746Q) led to real glycosynthases with completely abolished hydrolytic activity, but also to significantly slower enzymes, which was reflected in the lower yield after a longer reaction time (Table 9). Surprisingly, transfer of these mutations to LnbB was not as successful. Both mutants (D320E and D320A, Table 9) were significantly slower in synthesis of LNT from Lac and LNB-oxa ( Figure 6) leading to less than 50% of the wild-type yield in more than ten times longer reactions [105]. Finally, a similar 1.5-fold increase in yield compared to the WT was obtained with BbhI-D746E using a pNP-donor molecule, but the obtained yield was much lower (18%; Table 5) [101]. Recently, a new mutational study on BbhI combining directed evolution and site saturation mutagenesis (SSM) showed that the BbhI-D746T led to an almost doubled yield of LNT II when using Lac and pNP-GlcNAc as substrates. The maximum reported yield was 85% (Table 5) [104]. Furthermore, we can conclude from the SSM of position D746 that introduction of any residue bigger than Glu lead to complete inactivation of the enzyme, probably due to steric hindrance of substrate binding, (Table 5). Similarly, introduction of the two amino acid amides Asn and Gln in this position also lead to enzyme inactivation, although varying results were obtained in two independent studies on BbhI-D746Q (Table 5).  (Tables 5, 9, and 10) [24]. Normally, the catalytic Glu residue is required for protonation of the glycosidic bond, which in turn leads to release of the leaving group. Thus, a mutation in this position would be expected to lead to enzyme inactivation. However, the Withers group demonstrated almost 20 years ago that the activity of such a mutated GH can actually be rescued by the use of nucleophiles with a low pKa value such as thiols, which leads to the formation of thioglycosides [142]. The SpHex-E314A mutant was not only able to catalyze synthesis of thioglycosides from pNP-GlcNAc and Glc-oxa as donor substrates and the respective thio-sugars as acceptors, but could also catalyze the transfer of a GlcNAc moiety to the free amino acid Cys as Mutation of the catalytic Glu residue in SpHex to a non-functional Ala as recently described by Tegl et al. enabled a thioglycoligase reaction (Tables 5, 9 and 10) [24]. Normally, the catalytic Glu residue is required for protonation of the glycosidic bond, which in turn leads to release of the leaving group. Thus, a mutation in this position would be expected to lead to enzyme inactivation. However, the Withers group demonstrated almost 20 years ago that the activity of such a mutated GH can actually be rescued by the use of nucleophiles with a low pK a value such as thiols, which leads to the formation of thioglycosides [142]. The SpHex-E314A mutant was not only able to catalyze synthesis of thioglycosides from pNP-GlcNAc and Glc-oxa as donor substrates and the respective thio-sugars as acceptors, but could also catalyze the transfer of a GlcNAc moiety to the free amino acid Cys as well as to Cys-containing peptides and proteins leading to GlcNAc-peptide/protein conjugates [24]. However, such a mutant is not expected to synthesize conventional O-glycosidic bonds due to the rather high pK a values found in carbohydrates. Product formation of O-glycosides was only observed when using the low pK a nucleophiles pNP and 2,4-dinitrophenol [24].

Mutation of Other Conserved Active Site Residues
Another generic approach to increase trans-glycosylation activity of CAZymes is the mutation of other conserved residues in the active site to structurally related residues (e.g., from Tyr to Phe), which was first described in 2014 for a GH1 β-glycosidase [143] and has been transferred to many other GH families since then [144][145][146]. Recently, six conserved positions were identified from an alignment of 585 GH20 sequences and their effect on the trans-glycosylation activity of BbhI was studied [102]. The two mutants R577K and W288H ( Figure 7A) led to a doubled and quadrupled yield of LNT II, respectively, in reactions with Lac and pNP-GlcNAc as substrates (Table 5) [102]. These residues are present in both fungal and bacterial GH20 β-N-acetylhexosaminidases (residues highlighted in dark blue in Figures 5 and 7). Thus, the corresponding residues could be targeted in other GH20s exhibiting some natural trans-glycosylation activity, as the latter is a requirement for this engineering strategy [102].  [138]) with the mutated W805 [104] highlighted in orange and GlcNAc bound in the active site; (B) homology model of Hex1GTEPG [117] with the inserted loop highlighted in red and the newly positioned R360 in orange; the reaction intermediate Glc-oxa was docked into the structure [117]. The catalytic Asp-Glu pair is highlighted in yellow, the water-stabilizing Tyr residue is highlighted in purple, bound or docked ligands are highlighted in green, and conserved residues (Arg and Trp) are highlighted in dark blue.

Conclusions
GH20 β-N-acetylhexosaminidases are gaining significant attention for production of functional molecules, especially biomimetic human milk oligosaccharides, via their ability to catalyze enzymatic transfer of GlcNAc-and to a lesser extent GalNAc-via reverse hydrolysis or trans-glycosylation. Careful assessment of the available literature data showed that trans-glycosylation is more favorable than reverse hydrolysis with regard to efficiency and product yields. Indeed, a significant body of data have been reported on trans-glycosylation reactions promoted by microbially derived GH20 β-N-acetylhexosaminidases. Notably, a large amount of data have been reported for reactions catalyzed by the AoHex enzyme and the many different mutants of this enzyme derived from Aspergillus oryzae, and more recently also for other GH20 enzymes, in particular from Bifidobacterium bifidum. Surprisingly, it is not possible to discern any clear trends with respect to pH and temperature on GH20 trans-glycosylation reactions, but reaction engineering involving various types of donor activation, high A:D ratio, high substrate concentration in general, lowering of aw by addition of cosolvents, salts or addition of cyclodextrins can lead to increased yields or altered regioselectivity. The distinct catalytic mechanism of the GH20 β-N-acetylhexosaminidases, involving substrate-assisted catalysis in which the 2-acetamido group acts as an intramolecular nucleophile leading to formation of an oxazolinium ion intermediate, has proven uniquely useful as a blueprint for using oxazolineconjugated substrates for trans-glycosylation. Although chitin, e.g., from shrimp or crab waste streams, may seem an obvious substrate for sustainable β-N-acetylhexosaminidase catalyzed trans- Figure 7. Active sites models of two bacterial β-N-acetylhexosaminidases: (A) homology model of BbhI from Bifidobacterium bifidum (created with YASARA [138]) with the mutated W805 [104] highlighted in orange and GlcNAc bound in the active site; (B) homology model of Hex1GTEPG [117] with the inserted loop highlighted in red and the newly positioned R360 in orange; the reaction intermediate Glc-oxa was docked into the structure [117]. The catalytic Asp-Glu pair is highlighted in yellow, the water-stabilizing Tyr residue is highlighted in purple, bound or docked ligands are highlighted in green, and conserved residues (Arg and Trp) are highlighted in dark blue.

A Non-Conserved Loop Close to the Active Site as Hotspot for Beneficial Mutation?
Lastly, we would like to highlight a potential hotspot for mutations that drive trans-glycosylation activity in GH20 enzymes. We previously demonstrated that introduction of a specific loop, which was identified in related GH20 sequences from pathogens, into Hex1 led to a >5-fold increased trans-glycosylation product yield in reactions with Lac as acceptor and (GlcNAc) 2 as donor ( Figure 7B, Table 8) [117]. Additionally, in a recent mutational study on BbhI, the mutant W805R ( Figure 7A) was identified as the one with superior trans-glycosylation activity [104].
Interestingly, when comparing the homology models of BbhI and the Hex1-GTEPG loop mutant (Figure 7), we noticed that the mutated W805 in BbhI and the newly positioned R360 in Hex1-GTEPG are in a similar position with respect to the active site. Therefore, we dare to speculate that an Arg residue in this position is beneficial for trans-glycosylation reactions using GH20 β-N-acetylhexosaminidases. Such an Arg residue might be responsible for modulating the water network or binding water in general, which leads to lowered availability of water for hydrolysis, as proposed for GH33 trans-sialidases [147].
However, in case of Hex1 the inserted loop sequence itself also seems to have an important role with respect to trans-glycosylation activity, since not all alternative loop sequences of the same length were as beneficial as others [117]. Only the additional loops carrying a negative charge in the middle of the five amino acid sequence (GTEPG and GTDDA) led to an increased trans-glycosylation yield. Other tested loops carrying a positive charge (SFRTP) or a negative charge in a different position (DFVTP) led to a decreased trans-glycosylation activity [117].

Conclusions
GH20 β-N-acetylhexosaminidases are gaining significant attention for production of functional molecules, especially biomimetic human milk oligosaccharides, via their ability to catalyze enzymatic transfer of GlcNAc-and to a lesser extent GalNAc-via reverse hydrolysis or trans-glycosylation. Careful assessment of the available literature data showed that trans-glycosylation is more favorable than reverse hydrolysis with regard to efficiency and product yields. Indeed, a significant body of data have been reported on trans-glycosylation reactions promoted by microbially derived GH20 β-N-acetylhexosaminidases. Notably, a large amount of data have been reported for reactions catalyzed by the AoHex enzyme and the many different mutants of this enzyme derived from Aspergillus oryzae, and more recently also for other GH20 enzymes, in particular from Bifidobacterium bifidum. Surprisingly, it is not possible to discern any clear trends with respect to pH and temperature on GH20 trans-glycosylation reactions, but reaction engineering involving various types of donor activation, high A:D ratio, high substrate concentration in general, lowering of a w by addition of co-solvents, salts or addition of cyclodextrins can lead to increased yields or altered regioselectivity. The distinct catalytic mechanism of the GH20 β-N-acetylhexosaminidases, involving substrate-assisted catalysis in which the 2-acetamido group acts as an intramolecular nucleophile leading to formation of an oxazolinium ion intermediate, has proven uniquely useful as a blueprint for using oxazoline-conjugated substrates for trans-glycosylation. Although chitin, e.g., from shrimp or crab waste streams, may seem an obvious substrate for sustainable β-N-acetylhexosaminidase catalyzed trans-glycosylation processes, we anticipate that oxazoline-conjugated donor substrates have more potential for industrial development of these reactions. Insight into the details of the active site topology and the function of the different amino acids in both the fungal and bacterial GH20 enzymes indicate that trans-glycosylation activity may be promoted by the following protein engineering steps: Mutation of the aglycone binding site to large hydrophobic residues, mutation of the water-stabilizing Tyr as well as other conserved residues, notably Trp and Arg. Efficient glycosynthases using oxazoline substrates can be created by mutation of the catalytic Asp. Recent work suggested that introduction or repositioning of an Arg residue near the active site, e.g., by loop engineering, may be a hotspot for creating beneficial mutations for trans-glycosylation. Based on this foundation, we anticipate the development of novel protein engineering and reaction optimization strategies allowing further technological advances to promote exploration of GH20 enzymes for synthesis of distinct functional carbohydrates and glycan conjugates.

Conflicts of Interest:
The authors declare no conflicts of interest.