Glycosynthases have become efficient tools for the enzymatic synthesis of oligosaccharides, glycoconjugates and polysaccharides. Enzyme-directed evolution approaches are applied to improve the performance of current glycosynthases and engineer specificity for non-natural substrates. However, simple and general screening methods are required since most of the reported assays are specific for each particular enzyme. In the present paper, we report a general screening assay that is independent of enzyme specificity, and implemented in an HTS (high-throughput screening) format for the screening of cell extracts in directed evolution experiments. Fluoride ion is a general by-product released in all glycosynthase reactions with glycosyl fluoride donors. The new assay is based on the use of a specific chemical sensor (a silyl ether of a fluorogenic methylumbelliferone) to transduce fluoride concentration into a fluorescence signal. As a proof-of-concept, it has been applied to a nucleophile saturation mutant library of Bacillus licheniformis 1,3-1,4-β-glucanase. Beyond the expected mutations at the glutamic acid (catalytic) nucleophile, other variants have been shown to acquire glycosynthase activity. Surprisingly, an aspartic acid for glutamic acid replacement renders a highly active glycosynthase, but still retains low hydrolase activity. It appears as an intermediate state between glycosyl hydrolase and glycosynthase.
Retaining glycosidases, through their ability to catalyse transglycosylation reactions, have many synthetic applications for the preparation of oligosaccharides and glycoconjugates . With the exception of natural transglycosidases (which act by the same retaining glycosidase mechanism, but with essentially no hydrolase activity), the use of retaining glycosidases for kinetically controlled transglycosylation is severely hampered by the predominant hydrolase activity, where the product formed must be a hydrolysable substrate.
The glycosynthase concept was introduced to overcome these limitations for the efficient enzymatic synthesis of oligosac-charides and glycoconjugates. Glycosynthases are engineered glycoside hydrolases in which the catalytic nucleophile has been replaced by a non-nucleophilic residue. They are inactive hydrolases, but efficiently catalyse glycosyl transfer to an acceptor when using activated glycosyl fluoride donors with the opposite anomeric configuration from the original substrate of the parental wt (wild-type) hydrolase reaction (Figure 1A). Since the first reports in 1998 [2,3], retaining glycosidases from more than 12 different glycoside hydrolase families (according to the CAZy database classification ) have been converted into glycosynthases. They cover different specificities and relevant applications have been developed (for reviews, see [5–8]).
Fluoride chemosensor assay for monitoring glycosynthase reactions
Expanding the proficiency of current glycosynthases towards novel specificities and improved efficiencies is a major goal in enzyme engineering. The main drawbacks of the glycosynthase technology are: (i) reactions catalysed by glycosynthases are rather slow in comparison with their parental glycosidase activities (reduced kcat), requiring a large amount of enzyme and extended incubation times; (ii) reaction conditions, with regard to temperature and pH, are restricted to those in which the activated glycosyl donors (glycosyl fluorides) are stable enough and, therefore, often far from the conditions of optimal enzyme performance; (iii) for many exo-glycosynthases, the synthase reaction is not highly regiospecific, often yielding mixtures of products with different glycosidic bonds, whereas endo-glycosynthases are often highly regiospecific; and (iv) enzyme specificity limits the acceptance of modified donor and acceptor substrates to extend the application to non-natural substrates.
Improving the performance of glycosynthases with regard to the above drawbacks is being addressed by enzyme-directed evolution approaches [9,10]. The success of these approaches not only depends on the strategy to create mutant libraries (random or focused libraries), but also depends on the selection or screening methodology for the property of interest. In particular, screen-ing methods have to be implemented to allow for the analysis of large libraries with good sensitivity and robustness .
Few methods have been reported for the screening of directed evolution libraries of glycosynthases. Most methods rely on the detection of the product formed in the glycosynthase reaction: (i) coupled enzyme assay in which the product formed by the glycosynthase is a substrate of another glycosidase that releases an easily detectable chromophoric aglycone (e.g. β-glucosidase from Agrobacterium ); (ii) ELISA-based assay where the acceptor is immobilized on a plate and the glycosynthase product is detected by a specific antibody (e.g. Rhodococcus sp. endo-glycoceraminidase [13,14]); and (iii) chemical complementation in which a yeast three-hybrid system is used to link glycosynthase activity to the transcription of a reporter gene making cell growth dependent on glycosynthase product formation (e.g. Humicola insolens cellulase 7B [15,16]). These methods have shown good screening capacity, yet at the price of a narrow applicability since they are specific for each enzyme. In the search for universal methods to screen glycosynthase activity, a pH-based assay has been developed recently, taking advantage of the hydrofluoric acid released as a by-product of the glycosynthase reaction, which is detected by colour change in a pH indicator (e.g. β-xylosidase from Geobacillus stearothermophilus) . Although it is a general method, pH changes as the result of glycosynthase activity in culture plates and cell extracts when screening a library are not very sensitive and difficult to implement due to matrix sample variations.
In the present paper, we report a novel general assay for the screening of glycosynthases, independently of the enzyme specificity and on the basis of the quantification of the fluoride ion released as a by-product in all glycosynthase reactions with glycosyl fluoride donors (Figure 1). It is based on the use of a specific chemical sensor to transduce fluoride concentration into a fluorescence signal which enables the quantification of glycosynthase activity. To implement the concept, we first selected an appropriate ‘fluoride sensor’ compatible with HTS (high-throughput screening) of glycosynthase reactions with cell extracts expressing the enzyme variants to be analysed. The relevant classes of fluorescent fluoride chemosensors reported for analytical applications include amine co-ordination fluorophores [by ICT (intramolecular charge transfer)] [18,19], boron co-ordination compounds (in boron-containing π-systems) [20–22] and silyl ethers of fluorogenic alcohols [23–27]. We selected the last class since they can be used in aqueous solutions, a stable Si–F bond is formed resulting in a stable signal and they are highly specific not to expect interferences when used with cell extracts. As a proof-of-concept, the HTS method was validated and applied to the screening of a nucleophile saturation mutant library of Bacillus licheniformis 1,3-1,4-β-glucanase. It is a retaining endoglycosidase that hydrolyses mixed-linked 1,3-1,4-β-glucans (such as barley β-glucan and liquenan). The catalytic machinery involves a catalytic triad formed by Glu138 and Glu134 as general acid and catalytic nucleophile respectively, and Asp136 as an assisting residue that participates in the modulation of the pKa values of the catalytic residues . Glycosynthase variants of this enzyme have been characterized and used in oligosaccharide synthesis [29–32]. In a previous study, they were shown to be efficient catalysts for the preparation of artificial polysaccharides with regular structures through glycosynthase-catalysed donor self-condensation , but improved glycosynthase variants are required to reach high-molecular-mass polysaccharides. The screening assay applied to a nucleophile library has identified an unexpected mutation with glycosynthase activity. The fluoride sensor screening assay discussed in the present paper is of general application for further evolution experiments addressed to engineer new specificities and highly active glycosynthases.
MU (4-methylumbelliferone) was from Fluka, and t-butyldimethylsilyl chloride and Glc-pNP (p-nitrophenyl β-D-glucopyranoside) were from Sigma–Aldrich. The fluoride sensor MUTBS (4-methylumbelliferyl t-butyldimethylsilyl ether) was prepared as described in . Glcβ3GlcαF (α-laminaribiosyl fluoride) and Glcβ4Glcβ3GlcαF [β-D-glucopyranosyl-(1→4)-β-D-glucopyranosyl-(1→3)-α-D-glucopyranosyl fluoride] were prepared as reported previously .
Nucleophile saturation mutagenesis library generation
Saturation mutagenesis libraries of the Bgl nucleophile position were prepared using the gene encoding wt 1,3-1,4-β-glucanase in a pET16b vector , and partially overlapping oligonucleotides containing the NNK degenerated codon for the Glu134 codon: 5′-TGGGATNNKATCGACATCGAATTTCTAGG-3′ and 5′-ATGTCGATMNNATCCCAAGGCGTACC-3′, following a modified QuikChange protocol with extended primers . The mutagenesis reaction was directly used to transform electrocompetent BL21 Star Escherichia coli cells. Colonies (168) were replated in LB medium (50 mg/l ampicillin) and used to inoculate 96-deepwell plates containing 1.4 ml of LB medium (50 mg/l ampicillin). The library growth was carried out in deepwell plates for 24 h at 37°C at 150 rev./min in an orbital shaker. The final cultures were centrifuged (1250 g, 15 min, 4°C), the supernatants discarded and cell pellets were resuspended in 200 μl of LB medium containing 50 mg/ml ampicillin and 15% glycerol. These were then transferred to new 96-well microplates for storage at −80°C to be used as an inoculum.
Fluoride sensor assay
The assay to quantify the fluoride ion released in the glycosynthase reaction is based on the fluoride sensor MUTBS  under conditions adapted for polystyrene microplates and conditions compatible with a HTS implementation of glycosynthase reactions using cell extracts.
The reaction conditions were set with fluoride standards. A solution of potassium fluoride (0–500 μM) in 50 mM phosphate buffer, pH 7.0, was mixed with a MUTBS sensor (0.5 or 1 mM) in DMF (dimethylformamide) at ratios from 95:5 to 75:25 (final volume 100 μl), and left to react for 40 min at 25°C. Then, 100 μl of NH3/NH4Cl buffer (0.5 M, pH 9.4) was added, and the fluorescence measured at λex=365 nm and λem=460 nm in a microplate reader. The corresponding blank fluorescence (F0, at 0 mM fluoride) was subtracted from each measurement to give the F−F0 value which was plotted against fluoride concentration in the reaction. The reaction time with the sensor was assessed with a 300 μM fluoride solution following the same procedure as above, at a fluoride to MUTBS ratio of 90:10 (v/v), and stopping the reaction at different time intervals (0–120 min).
The final fluoride sensor assay conditions were as follows: 30 μl of 0.5 mM MUTBS in DMF/water (1:1) was added to 120 μl of fluoride solutions (standards or glycosynthase reactions), and the mixture was incubated for 40 min at 25°C. Then, 50 μl of NH3/NH4Cl buffer (0.5M, pH 9.4) was added and the fluorescence was measured at λex=365 nm and λem=460 nm. A standard curve (F−F0) compared with [F−] (μM) was determined (three repetitions) to validate the procedure.
Library expression and screening
Deepwell microplates containing 1.4 ml of LB medium (supplemented with 50 mg/l ampicillin and 2% glucose) were inoculated with 20 μl of the −80°C glycerol stock library. Each microplate contained 84 clones, two replicas of negative controls (pET16b and pET16b-Bgl), and eight wells were reserved for fluoride standards and glycosyl fluoride donor substrate control. This plate design was maintained along all of the cell growth and analysis procedures. Plates were sealed with a gas-permeable film and incubated for 8 h at 37°C in an orbital shaker at 150 rev./min. Once growth reached the late exponential phase, the growth medium was changed to 1.4 ml of LB medium (containing 50 mg/l ampicillin and 1 mM IPTG) to induce expression, and cells were grown overnight at 25°C. The overnight cultures were rinsed once with 0.9% NaCl solution, and lysed by ten consecutive freeze–thaw cycles (2 min at −80°C, followed by 8 min at 4°C), followed by toluene permeabilization of the cellular pellet in 150 μl of reaction buffer (50 mM phosphate buffer, pH 7.0, 0.1 mM CaCl2 and 3% toluene). After 30 min of incubation at 4°C, the microplates were centrifuged for 1 h at 1250 g and 120 μl of the supernatant was recovered for each well. Glycosynthase reactions were carried out in a new microplate. The lysate supernatant (60 μl) containing each mutant were mixed with 60 μl of glycosyl fluoride in 50 mM phosphate buffer, pH 7.0, to give a final 1.5 mM substrate concentration. The microplates were sealed and incubated for different times at 30°C. Next, 30 μl of 0.5 M MUTBS solution in DMF/water (1:1) was added. The microplates were incubated further for 40 min with protection from light at room temperature (23°C). After the addition of 50 μl of NH3/NH4Cl buffer (0.5 M, pH 9.4), fluorescence was measured at λex=365 nm and λem=460 nm. Fluoride concentrations were calculated by interpolation into a fluoride standard curve (0–500 μM) included in each microplate.
Expression and purification of glycosynthase mutants
Positive hits from the above screening were sequenced and recovered from the −80°C library stock for preparative expression and purification. A 3 ml culture (LB medium, 100 mg/ml ampicillin and 2% glucose) was inoculated with the corresponding glycerol stock and incubated overnight at 37°C. This was then used to inoculate 200 ml of LB medium (100 mg/ml ampicillin and 2% glucose) which was incubated for 8 h at 37°C and 250 rev./min until the late exponential phase. The medium was changed to 200 ml of LB medium containing 1 mM IPTG and 100 mg/ml ampicillin for protein expression (16 h at 25°C). Cells were harvested by centrifugation, rinsed with 200 ml of 0.9% NaCl, and resuspended in 50 ml of 50 mM phosphate buffer, pH 7.0, and 0.1 mM CaCl2. The proteins were purified essentially as reported for the wt enzyme using metal affinity chromatography of the His-tagged proteins . Cells were lysed by sonication and centrifuged at 17000 g. The supernatant was loaded on to a HiTrap 1 ml column (GE Healthcare) previously equilibrated with loading buffer (50 mM phosphate, pH 7.0, and 0.1 mM CaCl2). The column was rinsed to remove unbound proteins and then eluted with a gradient of 0.5 M imidazole in 50 mM phosphate, pH 7.0, and 0.1 mM CaCl2. The protein fractions were dialysed twice against 50 mM phosphate, pH 7.0, and 0.1 mM CaCl2, followed by a last dialysis against water. The proteins were freeze-dried for storage, and redisolved in 50 mM phosphate, pH 7.0, and 0.1 mM CaCl2 before use. Enzymes were >95% homogeneous as determined by SDS/PAGE (14% gel). The concentration was determined by absorbance at 280 nm using a molar absorption coefficient of 3.53×105 M−1·cm−1 .
Enzyme kinetics of the glycosynthase variants
Kinetic parameters for glycosynthase activity were determined with Glcβ4Glc3GlcαF as donor and Glc-pNP as acceptor substrates . The reactions were prepared in microplates by mixing 20 mM Glc-pNP (saturating) and varying concentrations of Glcβ4Glc3GlcαF donor (0–8 mM) in 50 mM phosphate buffer, pH 7.0, and 0.1 mM CaCl2, and initiated by adding the enzyme (0.5 μM) to a final volume of 200 μl. The reactions were incubated at 35°C with shaking. At different time intervals, 20 μl aliquots were withdrawn and added to 180 μl of 2% formic acid in a new microplate. These samples were analysed by HPLC (Agilent HPLC 1100 chromatograph) in a Nova-Pak® C18 column (4.9 mm×150 mm, 4 μm; Waters) with isocratic elution with water/methanol (87:13) at 45°C and UV detection at 300 nm. The product was quantified by interpolation of peak areas on an external Glc-pNP standard curve. Initial rates were calculated as the slope of product concentration against time. Kinetic parameters (kcat and Km for the donor substrate) were determined by fitting initial rates against donor concentration to the Michaelis–Menten equation by non-linear regression with Fig.P® software.
Possible donor hydrolyses by the mutant enzymes were checked under the same experimental conditions, by incubating the donor Glcβ4Glc3GlcαF (2 mM) and the acceptor Glc-pNP (20 mM) substrates with each mutant enzyme (0.5 μM) for 30 min at 35°C. The reaction mixtures were analysed by TLC on Silica plates (Silicagel 60 F254; Merck Millipore) eluted with acetonitrile/water (7:3). No hydrolysis products were detected at the initial reaction times.
The pH dependence of the glycosynthases activity was determined from kcat/Km values at different pH levels. The reactions were carried out as above in 50 mM phosphate, 50 mM citrate buffer and 0.1 mM CaCl2, and at a constant ionic strength of 0.5 M with added KCl, adjusted to different pH values in the range 5–7 (at pH>7.5, the solution becomes cloudy due to protein precipitation). The kcat/Km values plotted against pH data were adjusted to a single ionization curve for base catalysis (eqn 1) by non-linear regression with Fig.P® software:
Specific activities for hydrolysis of the chromogenic substrate Glcβ4Glc3GlcβMU were evaluated as reported for wt 1,3-1,4-β-glucanase . The kinetics was performed by following the changes in UV absorbance due to the release of MU using an Evolution 300 spectrophotomer (Thermo Scientific). The rates of the enzyme-catalysed hydrolyses were determined by incubating the Glcβ4Glc3GlcβMU substrate (5 mM final concentration) in citrate-phosphate buffer (6.5 mM citric acid and 87 mM Na2HPO4), pH 7.2 and 0.1 mM CaCl2 for 5 min in the thermostatically controlled cell holder at 55°C. The reactions were initiated by the addition of the enzyme (0.1–1 μM) and monitoring the absorbance change at λ=365 nm (Δε=5440 M−1·cm−1). Specific activities were expressed as v0/[E] (s−1).
The pH profile of the hydrolase activity was performed by determining v0 at 1 mM Glcβ4Glc3GlcβMU substrate in the pH range of 5–7, in 50 mM phosphate, 50 mM citrate buffer and 0.1 mM CaCl2, and at a constant ionic strength of 0.5 M with added KCl. The molar absorption coefficient of MU was determined under the same conditions at each pH value. Data were adjusted to a single ionization curve for acid catalysis (eqn 2) by non-linear regression with Fig.P® software:
RESULTS AND DISCUSSION
Fluoride sensor assay
The principle of the assay is to transduce the fluoride ion released in the glycosynthase reaction into a fluorescence signal. The selected fluoride sensor consists of a silyl ether of a fluorogenic coumarin, which reacts with the fluoride ion to form a stable Si–F bond with release of the fluorescent probe (Figure 1B). The use of MUTBS has been reported for the quantification of fluoride ion in toothpaste and tap water samples . The original protocol used acetone as an organic co-solvent to enhance the nucleophilicity of fluoride ion in the aqueous sample, but the high acetone content (70%) is not appropriate for HTS implementation. Using a set of standard fluoride solutions in the same buffer and concentration range to be used in the glycosynthase activity assays (0–500 μM fluoride ion in 50 mM phosphate buffer, pH 7.0), other co-solvents were assayed and reaction conditions were optimized. DMF at 10% in aqueous solution proved to give the best results. With the optimized protocol (see the Experimental section), a standard curve (F−F0) against [F−] was established (Figure 2A). The response was proportional up to 500 μM fluoride with a variation coefficient <5% at 500 μM and 15% at 25 μM, and the quantification limit was determined to be approximately 10 μM.
Set-up of the fluoride chemosensor assay with positive and negative controls
Developing the assay for glycosynthase activity
The enzyme used in the present study to develop the proof-of-concept of the glycosynthase HTS assay is a B. licheniformis 1,3-1,4-β-glucanase. The enzyme is intracellularly expressed in recombinant E. coli cells harbouring the pET16b-Bgl plasmid . The expression set-up in deepwell microplates was first optimized with the wt enzyme. The best conditions were induction after 8 h growth at 37°C and expression at 25°C for 16 h, followed by freeze–thaw cycles combined with membrane cell permeation with added toluene  to release the intracellular enzyme. These conditions were then used in the glycosynthase screening assay.
Next, the screening assay was validated using inactive and active glycosynthase enzyme as control references. Cells containing an empty plasmid (pET16b), the plasmid encoding the wt enzyme (pET16b-Bgl, template for mutant library generation) and a plasmid expressing a known glycosynthase mutant (pET16-Bgl-E134A) were subjected to the full screening assay: protein expression, cell lysis, glycosynthase reaction, fluoride sensor assay and fluorescence measurement.
E. coli cells harbouring the corresponding plasmids were grown, induced for protein expression and lysed. Glycosynthase activity was evaluated with Glcβ3GlcαF as a substrate, which is known to act as donor and acceptor in self-condensation reactions with the E134A glycosynthase mutant of 1,3-1,4-β-glucanase . Equal volumes of the clarified cell lysate containing the enzyme and substrate solution in reaction buffer (final 1.5 mM concentration in 50 mM phosphate, pH 7.0, and 0.1 mM CaCl2) were incubated at 30°C. The fluoride released by the glycosynthase reactions and fluoride standards (0 and 200 μM) were then quantified by the fluoride sensor as described above. The results are summarized in Figure 2(B), where mean fluorescence values (as F−F0) and S.D. values for each sample (eight repetitions) are given. Visual inspection clearly shows good discrimination between the positive and negative samples. The quality of the method in HTS assays is often expressed by Z-factors defined as in eqn (3) :
where μ is the mean signal value [(F−F0) in this case] for positive (+) and negative (−) levels, and σ+ and σ− are their corresponding S.D. values. The intrinsic Z-factor applies to the sensor reaction (standard fluoride solutions at 0 and 200 μM as negative and positive levels respectively), and the Z′-factor evaluates the quality of the overall process of library screening. E. coli cells harbouring the empty vector (pET16b) and the vector expressing the wt enzyme (pET16b-Bgl) are taken as negative levels, whereas cells expressing a glycosynthase mutant (pET16-Bgl-E134A) are the positive level. For the fluoride standards, the Z-factor was 0.72, whereas for enzyme samples Z′=0.69 (glycosynthase mutant against empty plasmid) or Z′=0.64 (glycosynthase mutant against wt glucanase). These values indicate a good discrimination between negative and positive hits . The variation coefficients for the F−F0 values are lower than 5% for all samples, which is a low variability considering the number of steps involved in the overall assay. This complies with a reliable method for the screening of glycosynthase activities implemented in an HTS format.
Saturation mutagenesis library at position 134 of 1,3-1,4-β-glucanase
Mutation of the catalytic nucleophile Glu134 of 1,3-1,4-β-glucanase to alanine, glycine or serine residues renders a glycosynthase . As a proof-of-concept of the screening assay, and to search for any other possible nucleophile mutant with glycosynthase activity in our target enzyme, a saturation mutagenesis library at the nucleophile residue was prepared. Mutagenesis used the gene encoding the wt (hydrolase) enzyme as template and partially overlapping oligonucleotides containing a NNK degenerated codon. Clones (168) were plated into two 96-well microplates. The library was screened according to the validated protocol (see the Experimental section). The final fluorescence measured in each well was converted into the fluoride concentration by means of the fluoride standards included in each plate. The results are shown in Figure 3, in which R-factor=([F−]−[F−]0)/[F−]0 for each clone are presented, where [F−] is the fluoride concentration of a given clone and [F−]0 is the fluoride concentration of the negative control (the one with the highest signal of pET16b or pET16b-Bgl). Two clusters are clearly segregated: those clones with an R-factor of approximately 0, corresponding to negative clones, and those with an R-factor of >1.5. The 30 positives (out of 168 clones) were recovered and had their plasmid DNA sequenced. Mutations found in the positive clones were serine, alanine, glycine, cysteine, threonine, and aspartic acid (Table 1). The first three (serine, alanine and glycine) residues were already known glycosynthase variants for this enzyme, and are the common amino acid substitutions reported for glycosynthases, where their relative glycosynthase activity depends on the particular enzyme . The cysteine mutant was not known previously as having glycosynthase activity for 1,3-1,4-β-glucanase, but it has been reported as a glycosynthase mutation in few other enzymes . Interestingly, two new amino acid substitutions were demonstrated in the present study to render active glycosynthases: threonine, which is a reasonable substitution similar to serine, and aspartic acid, which was fully unexpected.
Library screening with the fluoride sensor for screening
|Amino acid||Clone number||Codons found||Codons in NNK|
|Serine||6||1×AGT, 2×TCG and 3×TCT||3|
|Alanine||4||1×GCG and 3×GCT||2|
|Threonine||8||5×ACT and 3×ACG||2|
|Amino acid||Clone number||Codons found||Codons in NNK|
|Serine||6||1×AGT, 2×TCG and 3×TCT||3|
|Alanine||4||1×GCG and 3×GCT||2|
|Threonine||8||5×ACT and 3×ACG||2|
To assess that the positive hits were true glycosynthases and that the activity observed as fluoride release was not due to hydrolysis of the donor substrate (although not expected due to the α-fluoride configuration, opposite to the normal substrate for hydrolysis by the wt β-glycosidase), one clone of each mutation was expressed and purified for enzyme characterization. For general application with larger libraries (i.e. coming from random libraries), a secondary screening can be included to discard mutants with high hydrolase activity on the α-glycosyl fluoride donor (e.g. by a reducing sugar assay) before proceeding with the characterization of the positive glycosynthase mutants.
Characterization of positive glycosynthase hits
The six hits (E134A, E134S, E134T, E134C, E134G and E134D) were grown in cultures of 200 ml. Glucanase mutants were expressed and purified by affinity chromatography as reported for E134S . Their glycosynthase activities were determined using Glcβ4Glc3GlcαF as donor and Glc-pNP as acceptor substrates at pH 7.0 and 35°C, monitoring product formation by HPLC. In all cases a time-dependent formation of the condensation product was observed, and no donor hydrolysis was detected (TLC). Hydrolase activity with the chromogenic substrate Glcβ4Glc3GlcβMU was also evaluated to detect any residual β-glycosidase activity of the mutant variants.
Glycosynthase kinetics are plotted in Figure 4 and kinetic parameters are summarized in Table 2. In terms of kcat/Km values, the serine mutant was the most efficient glycosynthase, mainly due to its low Km for the donor substrate, followed by glycine, aspartic acid, cysteine, alanine and threonine mutants. But comparing kcat values, E134D was the most active. Nucleophile mutations to serine, alanine and glycine residues were already known glycosynthases for 1,3-1,4-β-glucanase [3,33]. The cysteine mutant also behaved as a glycosynthase as observed in other enzymes, and the threonine mutant (to our knowledge, a mutation not reported for other glycosynthases) was also active, but had the lowest activity. Yet, the aspartic acid mutant was not expected to have glycosynthase activity. This is a surprising result since it is a conservative substitution of the wt glutamate residue acting as nucleophile in the hydrolase reaction catalysed by the wt enzyme. When testing the hydrolase activity with a β-glycoside substrate (Glcβ4Glcβ3GlcβMU), all mutants gave no detectable activity except E134D, which retained 2% of wt activity. Shortening the side chain of the nucleophile residue (glutamic acid to aspartic acid) was already known to decrease significantly the hydrolase activity of β-glycosidases (e.g. lacZ β-galactosidase , Agrobacterium β-glucosidase  and Bacillus circulans xylanase ). The glycosynthase reaction of E134D was monitored during long incubation times. As shown in Figure 5(A), the condensation product was rapidly formed up to a 90% conversion, but then slowly hydrolysed following long incubation. Therefore E134D can be seen as a transitional mutant between hydrolase and synthase activities.
|Mutant||kcat (s−1)||KM,donor (mM)||(kcat/KM)donor (M−1·s−1)||% kcat||% kcat/KM|
|Mutant||kcat (s−1)||KM,donor (mM)||(kcat/KM)donor (M−1·s−1)||% kcat||% kcat/KM|
Kinetics of the glycosynthase variants
Time-course and pH dependence of the reactions catalysed by the E134D mutant
The E134D nucleophile mutant is an intermediate between hydrolase and glycosynthase
On the basis of the 3D structure of the enzyme–product complex of the hydrolase reaction  and our theoretical calculations on the hydrolase mechanism of the 1,3-1,4-β-glucanase , the glycosynthase mutants with bound α-glycosyl fluoride substrate were modelled in an attempt to rationalize the results obtained in the present study (Figure 6). Whereas a β-glycoside substrate binds with a 1S3 distorted conformation to the wt enzyme in the Michaelis complex (pre-activated conformation for catalysis in the hydrolase mechanism [46,47]), the α-glycosyl fluoride donor in the glycosynthase mutants remains in a chair (4C1) conformation (as recently observed in the X-ray 3D structures of glycosynthase mutants of rice BGlu1 ). The shorter side chain of the mutants at position 134 relative to the wt creates a cavity that accommodates the fluoride aglycone with α-configuration. For the serine mutant (Figure 6A), the hydroxy oxygen is at a distance of approximately 4.3 Å (1 Å=0.1 nm) from the fluoride atom, with an orientation that may allow for a weak hydrogen bonding interaction to assist the departure of the fluoride aglycone in the SN2 displacement reaction by the acceptor. This rationale has also been given to explain the higher activity of serine mutants compared with alanine or glycine mutants in other glycosynthases [49,50]. Additionally, the occurrence of this hydrogen bond R-F··H-O-Ser will result in better binding, which is in agreement with the lower Km (donor) observed for the E134S mutant.
Modelled structure of enzyme–donor complexes for E134S and E134D mutant 1,3-1,4-β-glucanases with the Glcβ3GlcαF donor
An essential aspect of the glycosynthase mechanism is the need for a deprotonated carboxylate to act as a general base to activate the acceptor substrate (Figure 1). In the wt 1,3-1,4-β-glucanase, Glu138 is the general acid (with a pKa value of 7.2) in the first step leading to the covalent glycosyl-enzyme intermediate in the hydrolase mechanism . This high pKa is, in part, due to the presence of the negatively charged catalytic nucleophile Glu134, which destabilizes the conjugate base of Glu138. It drops in the glycosyl-enzyme intermediate due to neutralization of the negative charge in the covalent complex allowing Glu138 to act as a general base in the deglycosylation step, leading to the hydrolysis product . In the glycosynthase mechanism catalysed by the original E134A mutant, Glu138 has a lower pKa of 5.2  owing to the alanine for Glu134 substitution. The same concept applies for the other glycosynthase variants having a neutral residue in the position of the original nucleophile. However, identification of the E134D mutation in the present study is surprising. For the mutant E134D to behave as a glycosynthase, Glu138 must also be able to act as a base. The E134D mutant shows a pH dependence corresponding to general base catalysis, with a kinetic pKa (on kcat/Km) of 5.7 and maximum activity at pH 7 (Figure 5B). At pH values higher than 7.5, the protein is less stable and precipitates, precluding kinetic measurements. The observed pKa is similar to that obtained previously with the original E134A glycosynthase mutant of 1,3-1,4-β-glucanase (pKa=5.2), which corresponds to Glu138 acting as general base (as assigned in the E134A mutant by titration with a water-soluble carbodiimide) . This is also consistent with the pH profile of the residual hydrolase activity of the mutant with a β-glycoside substrate (Figure 5C), where the hydrolase activity decreases with pH, with a pKa value of 5.9, assigned to Glu138 as a general acid (pKa 7.2 in the wt enzyme).
The possible reasons for the low pKa of Glu138 in the E134D mutant are that the longer distance between the carboxylates reduces the electrostatic effect of the negatively charged Asp134, or that the carboxylate of Glu138 is stabilized by a hydrogen bonding interaction with Asp136, a third auxiliary residue of the so-called catalytic triad in the glycoside hydrolases 16 family which participates in modulating the pKa values of the catalytic residues. The latter is consistent with the observed X-ray structures of the free wt enzyme (Asp136 hydrogen bonds with Glu134), of a covalent enzyme–ligand complex (Asp136 has been rearranged and forms hydrogen bonds with Glu138), and of an enzyme–product complex (where Asp136 occupies an intermediate position) . Therefore it is reasonable in the E134D glycosynthase that, even with Asp134 being deprotonated, the carboxylate of Glu138 might be stabilized by Asp136, thus being able to act as a base (Figure 6B). Another possible effect is the partial shielding of the negative charge of Asp134 by an halogen bonding interaction with the fluorine atom of the glycosyl donor in the enzyme–substrate complex. Although this interaction might be very weak with fluoride (more significant for heavier halogens), the modelled structure in Figure 6(B) proposes a distance F··O of 3.3 Å and a C1–F··O angle of ≈160° that perfectly fit the average geometry of halogen bonding interactions found in protein–ligand complexes . Work is in progress to get a deeper understanding on the mechanism of the different glycosynthase variants, and explain the efficiency of the E134D mutant.
The glycosynthase screening assay reported in the present paper based on a chemical fluoride sensor is a general method applicable to any glycosynthase with a glycosyl fluoride donor. It is independent of the substrate specificity of the particular enzyme since it is based on the detection and quantification of the fluoride ion released, which is a general by-product in the reaction. There are two main objectives in evolving a glycosynthase: modify substrate specificity to accept a non-natural (or modified) substrate, and improve the efficiency of a given glycosynthase reaction. In the first case, a glycosynthase mutant active on its ‘natural’ substrate and inactive on the new ‘non-natural’ substrate will be the negative control and template for evolution. The assay used in the present study will identify positive variants with glycosynthase activity. In the second case, an active glycosynthase with the substrate to be used in the screening is the initial template for evolution, and therefore the parameter to be evaluated is the increase in glycosynthase activity. The assay allows for two variables to be adjusted to different levels of activity; reduce the amount of enzyme in the glycosynthase reaction and/or reduce the incubation time in order to bring the initial glycosynthase template to a negative level for the fluoride sensor assay. The sensitivity window is therefore shifted to detect more active glycosynthase variants.
Application of the screening assay to a nucleophile saturation library of 1,3-1,4-β-glucanase identified, in addition to the known alanine, serine and glycine mutations, two new variants, threonine and aspartic acid. The E134D mutant was an unexpected glycosynthase with improved kcat value, but still with residual hy-drolase activity. It appears as an intermediate state between hydrolase and glycosynthase for which further studies will contribute to decipher the fine tuning of active site pKa values in the glycosynthase mechanism.
4-methylumbelliferyl t-butyldimethylsilyl ether
Antoni Planas conceived the work and designed the experiments. Eduardo Andrés and Hugo Aragunde performed the experiments. Eduardo Andrés, Hugo Aragunde and Antoni Planas discussed the results. Antoni Planas, Eduardo Andrés and Hugo Aragunde wrote the paper.
We thank Dr Xevi Biarnés (Laboratory of Biochemistry, Institut Químic de Sarriá, Universitat Ramon Llull, Barcelona, Spain) for modelling studies, and Estela Castilla for pH studies of the E134D mutant.
This work was supported, in part, by the Ministry of Economy and Competitiveness, Spain [grant BFU2010-22209-C02-02]. A predoctoral fellowship from Generalitat de Catalunya is acknowledged (to H.A.).