Rational Design of Cyclodextrin Glycosyltransferase with Improved Hesperidin Glycosylation Activity

Abstract

Cyclodextrin glycosyltransferase (CGTase) can catalyze the glycosylation of hesperidin, resulting in α-glycosyl hesperidin with significantly improved water solubility. In this study, a rational design of CGTase to improve its hesperidin glycosylation activity was investigated. The strategy we employed involved docking hesperidin in its near-attack conformation and virtually mutating the surrounding residues, followed by calculating the changes in binding energy using Rosetta flex-ddG. The mutations with a stabilization effect were then subjected to an activity assay. Starting from CGTase-Y217F, we obtained three double-point mutants, Y217F/M351F, Y217F/M351L, and Y217F/D393H, with improved hesperidin glycosylation activities after screening twenty variants. The best variant, Y217F/D393H, exhibited a catalytic activity of 1305 U/g, and its kcat/KmA is 2.36 times higher compared to CGTase-Y217F and 15.14 times higher compared to the wild-type CGTase. Molecular dynamic simulations indicated that hesperidin was repulsed by CGTase-Y217F when bound in a near-attack conformation. However, by introducing a second-point mutation with a stabilization effect, the repulsion effect is weakened, resulting in a reduction in the distances between the bond-forming atoms and, thus, favoring the reaction.

1. Introduction

Hesperidin is a natural bioflavonoid enriched in citrus peel. As a member of the vitamin P family, hesperidin is functional in regulating the permeability of capillaries [1]. Other activities, such as anti-oxidation [2], anti-inflammation [3], anticancer [4], and regulating blood sugar and cholesterol levels, have also been reported [5,6]. Typical applications of hesperidin include being used as a pharmaceutical to treat hemorrhoids, varicose veins, and lymphedema, as well as in daily products such as cosmetics and functional foods [7,8,9,10]. Hesperidin is more bioavailable in its non-glycosylation state, hesperitin, which is formed after oral intake and hydrolysis by gut microbiota [11]. Unfortunately, gut microbiota can hardly process hesperidin as it is nearly insoluble in water [12]. Poor water solubility also causes problems when trying to apply hesperidin in water-based formulations.

Glycosylation has a profound impact on the water solubility of hesperidin, with the water solubility of monoglucosyl hesperidin reaching as high as 1970 g/L. The study of hesperidin glycosylation was first reported by Kometani et al., who applied cyclodextrin glycosyltransferase (CGTase) originated from an alkalophilic Bacillus sp. to catalyze the transglycosylation from starch to hesperidin [13]. The final product involves a series of hesperidin glycosides, including its monoglucosyl and oligoglucosyl forms. Following studies have reported the use of a commercial glucoamylase that hydrolyzes hesperidin oligoglucosides selectively to α-monoglucosyl hesperidin [14]. Currently, α-monoglucosyl hesperidin has been applied in skincare products to protect tissues from cold shock and promote skin rejuvenation [15,16].

As an enzyme capable of catalyzing hesperidin glycosylation, CGTase belongs to the glycoside hydrolase 13 (GH13) family, which is a group of carbohydrate-active enzymes that act on hydrolyzing and transferring α-glucoside linkages through a retaining mechanism [17]. The GH13 family is the largest glycoside hydrolase family, with over 150,000 identified enzymes [18]. It is noteworthy that, in addition to hesperidin, enzymes in the GH13 family have been extensively used to glycosylate a variety of aglycons, including glycerol [19], L-ascorbic acid [20], hydroquinone [21], resveratrol [22], ginsenoside [23], etc., and are deemed industrially important enzymes to synthesize valuable glycosides.

On the other hand, hesperdin glycosylation catalyzed by wild-type CGTase has poor efficiency, with a k_cat/K_m value of 0.07 g/L·min, as investigated in our previous study [24]. To improve the synthetic rate, we applied a size/polarity-guided triple-code strategy to engineer the +1 and +2 subsites, as such areas closely interact with aglycon during transglycosylation. Three residues, F205, Y217, and G283, were identified to significantly impact the catalytic rate, and the change in binding energy of hesperidin in its near-attack conformation in the catalytic pocket appears to be the key reason [24].

Binding energy refers to the energy required to bring and stabilize a ligand molecule into an active site. Ignoring inactive binding states, the energy required for a ligand to bind in its near-attack conformation contributes significantly to the overall energy barrier and can, therefore, serve as a valuable reference during enzyme design [25]. In this work, we started with our previous mutant CGTase-Y217F and attempted to identify more key residues rationally by calculating the change in binding energy after virtual mutations using flex-ddG, a Rosetta-based modeling suit [26]. After screening 20 mutants, 3 positive mutants were obtained, and M351 and D393 were identified as two new key residues.

2. Results

2.1. Selection of Mutagenesis Sites and Rational Design

The crystal structure of CGTase studied in this research (NCBI accession number: BAA31539.1) has not yet been resolved. However, the crystal structure of CGTase from Bacillus circulans strain 251, which shares 73.6% sequence similarity, is available (PDB: 1CXL) in a covalent complex with a glycosyl donor [27]. Using this as a reference, a theoretical model of CGTase was constructed, and molecular docking was performed to obtain the structure model of CGTase-Y217F in a complex with hesperidin, as shown in Figure 1a.

The catalytic process is driven by a double-displacement mechanism that is catalyzed by a catalytic triad consisting of D251, E279, and D350. In this process, D251 and E279 play the roles of electron donors and acceptors, respectively. D350, on the other hand, does not participate in the reaction but instead serves to accurately position the glucose unit at the −1 position [27]. In this case, in addition to the catalytic triad and the first-generation mutagenesis (Y217F), ten residues within a 4 Å radius of the docked hesperidin were selected for rational design, which include F205, S211, A252, K254, H255, F281, M351, N392, D393, and R397 (Figure 1b). A saturated virtual mutation was applied to these selected residues, and the change in binding energies before and after the mutation was calculated using Rosetta flex-ddG. Figure 2 shows the results that are considered to stabilize hesperidin in its near-attack binding conformation (∆∆G ≤ −1.0 kal/mol), which are subjected to mutagenesis and an assay. The entire calculation results are given in the Supplementary Materials Section (Figure S1).

2.2. Library Construction and Screening

A mutagenesis library was constructed using pET-22b(+)/CGTase-Y217F-6×His as the template and corresponding primers for each of the 20 designed mutants. The resulting plasmids were confirmed to carry the CGTase-expressing gene through gel electrophoresis (Figure S2), and the target mutations were confirmed by DNA sequencing.

The activities of heterogeneously expressed CGTases were assayed through two consecutive rounds of screening, either using whole cells or purified enzymes. SDS-PAGE was applied to characterize the purified enzymes (Figure S3). The screening results are shown in

Table 1. Activities of CGTase variants assayed with whole cells (relative activity) and purified enzymes (specific activity).

Enzyme	Relative Activity (%)	Specific Transglycosylation Activity (U/g)
WT	100	134.4 $\pm$ 7.0
Y217F	671.6	916.6 $\pm$ 22.2
S211T/Y217F	574.2	-
Y217F/K254V	458.1	-
Y217F/K254L	684.1	478.3 $\pm$ 20.7
Y217F/K254H	575.6	-
Y217F/H255S	16.3	-
Y217F/H255L	89.4	-
Y217F/H255F	41.4	-
Y217F/F281M	333.1	-
Y217F/F281Y	743.3	805.6 $\pm$ 6.3
Y217F/F281R	68.2	-
Y217F/F281H	157.9	-
Y217F/M351T	402.5	-
Y217F/M351F	774.5	1194.7 $\pm$ 12.5
Y217F/M351L	791.8	1126.1 $\pm$ 6.4
Y217F/M351A	287.5	-
Y217F/M351Y	547.0	-
Y217F/M351H	238.0	-
Y217F/N392V	533.8	-
Y217F/N392M	474.4	-
Y217F/D393H	716.4	1306.4 $\pm$ 21.6

. The mutations at M351 and D393 were found to favor the target reaction. The specific transglycosylation activities of the positive mutants Y217F/M351F, Y217F/M351L, and Y217F/D393H were assayed to be 1171 U/g, 1126 U/g, and 1305 U/g, which are 1.25, 1.20, and 1.39 times higher compared to the first-generation mutant Y217F and significantly superior than the wild-type CGTase (134 U/g). The disproportion kinetics of the three positive variants were analyzed, as shown in

Table 2. Disproportion kinetics of the wild-type CGTase and its variants.

Enzyme	kcat (min−1)	KmA (Hesperidin) (g/L)	KmB (Maltodextrin) (g/L)	kcat/KmA (L/g·min)
WT	0.47	6.52	0.92	0.07
Y217F	4.85	10.86	2.11	0.45
Y217F/M351F	11.58	16.43	6.48	0.71
Y217F/M351L	16.31	26.42	4.87	0.62
Y217F/D393H	14.74	13.87	2.84	1.06

, based on a ping-pong mechanism. In the case of k_cat/K_mA, representing the enzymes’ hesperidin glycosylation ability in abundance of maltodextrin, the best variant Y217F/D393H is 2.36 times higher than Y217F and 15.14 times higher compared to the wild type. The original kinetic data are displayed in Figure S4.

3. Discussions

The primary application of CGTases is to produce cyclodextrins, which are macrocyclic molecules used to encapsulate other molecules, resulting in the formation of inclusion complexes with improved water solubility and stability [28]. Recently, the use of CGTase to catalyze intermolecular transglycosylation has attracted wide attention as the reaction uses cheap glycosyl donors, such as starch and maltodextrin, to produce valuable glycosides. Besides hesperidin, the glycosylation of its aglycon, hesperitin, using CGTase has also been reported [29]. Several studies have reported the engineering of CGTase to improve its ability to catalyze intermolecular glycosylation reactions. For example, Han et al. applied site-saturation mutagenesis to K47 of CGTase from Paenibacillus macerans and improved the yield of AA-2G, a glycosylation product of L-ascorbic acid, by 64.2% [30]. Tao et al. engineered the residues near the catalytic site, and the enzyme’s k_cat/K_m in synthesizing AA-2G was increased by 2.69-fold [31]. Han et al. engineered the nonconservative site of CGTase and improved its selectivity for synthesizing long-chain glycosylated sophoricosides [32]. Ara et al. engineered three residues near the acceptor subsites and obtained variants with improved coupling activities for synthesizing alkyl glycosides with elongated carbohydrate chains [33].

Although previous studies have been successful in regulating intermolecular glycosylation abilities, rational engineering of a CGTase based on a target transglycosylation reaction has not yet been achieved. It is worth noting that several methods have been proposed to rationally design an enzyme, which are based on the principle of reducing activation energy by minimizing the energy of the enzyme–ligand complex along the reaction pathway but with different algorithms. For instance, a fixed-backbone design only permits side-chain rotamers to be varied during enzyme design while keeping the backbone conformation fixed [34]. In contrast, a backrub design allows for backbone conformational variations before designing for each member of the ensemble [35]. Although fixed-backbone design is computationally efficient, backrub design samples the flexibility in enzyme structure that is closer to the real case of enzymatic catalysis and theoretically provides more accurate results.

In a previous study, we engineered the aglycon-attacking site of CGTase from the alkalophilic Bacillus sp. A2-5a and obtained eight positive mutants at three sites [24]. Upon characterizing the positive effects, it became apparent that the stabilization of hesperidin in its near-attack conformation was the major contributing factor. In this case, we started with the best first-generation mutant Y217F and further applied a flex-ddG design inoculated with the backrub algorithm in this study, aiming to identify additional residues whose mutations could potentially contribute to the energy profile along the glycosylation pathway.

The CGTase-catalyzed glycosylation of hesperidin proceeds through a double-displacement mechanism, where hesperidin attacks the enzyme–glycosyl intermediate from the reducing end in the second displacement step, ultimately leading to the formation of an α-1,4-glucosidic bond (Figure 3) [27]. In the docked structure shown in Figure 1a, the glucose unit of hesperidin is positioned beneath the glycosyl intermediate chain, with the nucleophile O₄ 2.3 Å away from C_α, and the distance between the electron donor, OE2 in E279, and H4 in hesperidin is 2.4 Å. These distances fall within the average van der Waals distance between carbon and oxygen (3.35 Å) and oxygen and hydrogen (2.6 Å), indicating that the bond-forming atoms are within near-attack distance to form an α-1,4-glucosidic bond. Based on this observation, the selected docked structure has the potential to be developed for the target reaction, and a rational design aimed at promoting the formation of this structure could potentially benefit the reaction.

Based on the design and screening results, when using the near-attack conformation as the starting structure, a total of twenty designs demonstrated stabilization effects, and three were screened to be positive. It is important to note that rational design at the current stage may not achieve 100% accuracy due to several reasons. For instance, Lodola et al. demonstrated that amide hydrolysis by an amide hydrolase occurs more readily in a high-energy conformation than in the ground state [36]. Additionally, Romero-Téllez et al. suggested that, due to the flexibility of the enzyme–ligand complex, the energy barrier of an enzyme-catalyzed reaction is not determined by a single reaction pathway [37]. The different precatalytic structures may result in a dispersion of energy barriers that combine statistically to determine the reaction kinetics. Thus, due to the conformational effect, rational design from a static Michaelis complex structure, which is usually used as the starting point for rational design, may not capture the full picture of the catalytic mechanism. Wijma et al. investigated the rational design of the enantioselectivity of an epoxide hydrolase [38]. They propose to further evaluate the designed mutant by high-throughput molecular dynamic simulation to further confirm the frequency of forming near-attack conformations.

To gain a deeper understanding of the positive effect, we conducted molecular dynamic simulations on CGTase-Y217F and the three identified positive mutants. The relative positions of the atoms involved in bond formation play a crucial role in determining the likelihood of their effective collision, which is critical for the success of the reaction. By fixing the relative positions of C_α and OD2 in Asp251, we determined the distances between the bond-forming atoms, as shown in Figure 4. During the 10 ns simulation, the distances between the bond-forming atoms increased, suggesting repulsion between the two molecules. Our previous study showed that although the Y217F point mutation removes the bump at the pocket entrance and facilitates hesperidin transportation, it has a destabilizing effect on hesperidin binding. In this case, introducing mutations with stabilization effects, as performed in this study, can potentially overcome the Y217F mutation’s defect and further improve the enzyme’s catalytic performance. By introducing mutations at M351 and D393, the repulsion effect is weakened and the average distances between the bond-forming atoms are reduced, as shown in

Table 3. Average distances between bond-forming atoms during 10 ns simulation.

Mutants	Average Distances during 10 ns Simulation (Å)
	O4 and Cα	H4 and OE2 in E279
Y217F	0.832	0.686
Y217F/M351F	0.540	0.645
Y217F/M351L	0.531	0.444
Y217F/D393H	0.729	0.634

, which explains the positive effect observed.

4. Materials and Methods

4.1. Materials, Plasmids, and Strains

The nucleotide sequence encoding CGTase (NCBI accession number: BAA31539.1) was synthesized with a polyhistidine tag attached on the C-terminus [39]. Recombinant vectors and strains were obtained following previously reported protocols [24]. Hesperidin, dextrin, and other chemicals were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). His60 Ni Superflow resin and gravity columns were purchased from SunResin Co. (Xi’an, China). Genetic engineering reagents, including a Super-Fidelity DNA Polymerase Kit, Dpn I digesting enzyme, and a One Step Cloning Kit, were purchased from Vazyme Biotech (Nanjing, China).

4.2. Structure Modeling and Rational Design

The theoretical structure of CGTase-Y217F in a complex with hesperidin was obtained in our previous study [24]. Residues within the 4 Å radius of the docked ligand were virtually mutated to the other 19 common amino acids, and changes in the ligand binding energy after mutation were calculated using Rosetta flex-ddG. Results with ΔΔG ≤ −1.0 kal/mol were employed for mutagenesis and an activity assay. The user-defined options in Rosetta flex-ddG include the following: repetitive computation: “nstruct = 15”; max minimization gradient descent steps: “max_minimization_iter” = 5000; maximum allowed score change threshold of minimization: “abs_score_convergence_thresh” = 1.0; number of backrub sampling steps: “number_backrub_trials” = 35000; steps of ddG calculation during backrub: “backrub_trajectory_stride” = 35000; residue to mutate: “residue_to_mutate = (‘protein chain name’, ‘residue number‘)”; target mutation: “mut_aa in ‘ACDEFGHIKLMNPQRSTVWY’”. A “chain_to_move.txt” file was written to define the ligand molecule.

4.3. Preparation of Mutant Libraries

Primers were synthesized as listed in

Table 4. Primers used for point mutations.

Primers	Sequence (5′−3′ Direction) xxx: Mutation Site	Target Mutation (Codon)
S211-R	aatctgtttgatctggcggattacgatctg	T
S211-F	cgccagatcaaacagattgcggtaaatxxxatcctcata	ggt
K254-R	ttgatggcattcgcgtggatgcggttxxxcacatgagcgaag	V	L	H
K254-F	tccacgcgaatgccatcaatgcctttgtcca	gtg	cat	ctg
H255-R	gaaggctggcagactagcctgatgagcgatatt	S	L	F
H255-F	gctagtctgccagccttcgctcatxxxtttaac	gct	cag	gaa
F281-R	agcggcgaagttgatccgcagaaccatcat	M	Y	R	H
F281-F	cggatcaacttcgccgctgcccagxxxccattc	cat	ata	gcg	atg
M351-R	agcggcgaagttgatccgcagaaccatcat	T	F	L	A	Y	H
M351-F	gctctgttcaaagctaaagcggctxxxgtcatggtt	cgt	gaa	cag	cgc	ata	atg
N392-R	gaaaaccgcaaaccgatgagcgattttgat	V	M
N392-F	tcatcggtttgcggttttccggatcxxxgccgcccgtcagat	cac	cat
D393-R	gaaaaccgcaaaccgatgagcgattttgat	H
D393-F	atcggtttgcggttttccggxxxattgccgcccgtcagata	atg

. The target mutation was constructed by amplifying the entire plasmid pET-22b(+)/CGTase-6×His using the corresponding primers. The PCR conditions were followed according to the instructions of the Phanta^® Max Super-Fidelity DNA Polymerase Kit (Vazyme Biotech, Nanjing, China). The PCR product was then digested with Dpn I (Vazyme Biotech, Nanjing, China) and circulated with the One Step Cloning Kit (Vazyme Biotech, Nanjing, China), leading to a plasmid with the target mutation. The plasmid was then transformed into competent E. coli BL21 (DE3) cells. The above procedures were repeated with different primers to create libraries with CGTase mutants for screening.

4.4. Screening of Positive Mutants

Two rounds of screening were performed to assay the enzyme activities using whole cells and purified enzymes. In the first round of screening, recombinant E. coli BL21 (DE3) cells were inoculated into 50 mL aliquots of LB medium with 10 g/L glycerol and 100 mg/L ampicillin in shake flasks. After culturing at 37 °C for 6 h, 50 μL of IPTG (500 mM) was supplemented, and the incubation temperature was reduced to 22 °C for enzyme expression. After another 4 h of culturing, the cells were harvested, and the OD₆₀₀ of each culture was determined. Whole-cell catalysis was performed for 60 min with the culture at 40 °C and pH = 10.0, adjusted by 1M NaOH. Dextrin (100 g/L; DE = 13.31, determined by the DNS method) and hesperidin (10 g/L) were supplemented as glycosyl donors and acceptors, respectively. The samples were added to 1M NaOH to quench the reaction. The concentration of hesperdin and its glycosylation products were determined by HPLC with an InertSustain C18 (GL Sciences, Tokyo, Japan) column. One unit of whole-cell transglycosylation activity (WCTA) was defined as 1 μmol of substrate converted per minute by a unit concentration of cells. The mutants with WCTA larger than CGTase-Y217F were subjected to a second round of screening.

In the second round of screening, the purified CGTases were obtained using His60 Ni Superflow resin and gravity columns. The purity of the obtained CGTases was confirmed by SDS-PAGE, and the protein concentrations were determined by the Bradford method. The enzyme activities were assayed by adding 20 μL of enzyme to a 2 mL solution containing dextrin (100 g/L; DE = 13.31) and hesperidin (10 g/L). The reactions were conducted at 40 °C for 15 min at pH = 10.0, adjusted by 1M NaOH. The samples were added to 1 M NaOH to quench the reaction and then analyzed by HPLC. One unit of specific enzyme activity was defined as 1 μmol of substrate converted per minute by a unit concentration of enzyme.

4.5. Kinetic Analysis

The glycosylation kinetics were investigated with the purified CGTases. The reaction was initiated by adding 20 μL of enzyme to a 2 mL solution containing hesperdin and maltodextrin in various concentrations. Catalysis was performed at 40 °C and pH = 10.0 for 15 min to determine the initial rates. The obtained experimental data were fit to Equation (1) on behalf of a ping-pong mechanism, where v (g/L·min) stands for the disproportion rate, E₀ is the enzyme concentration, and [A] and [B] stand for the hesperidin and maltodextrin concentrations (g/L), respectively.

$$\mathrm{v}=\frac{k_{\mathrm{c}}{\mathrm{E}}_0\left[\mathrm{A}\right]\left[\mathrm{B}\right]}{K_{\mathrm{m}\mathrm{B}}\left[\mathrm{A}\right]+K_{\mathrm{m}\mathrm{A}}\left[\mathrm{B}\right]+\left[\mathrm{A}\right]\left[\mathrm{b}\right]}$$

(1)

4.6. Molecular Dynamic Analysis

Gromacs was employed for a molecular dynamic analysis. The theoretical structure of CGTase-Y217F with hesperdin docked in its near-attack conformation was used as the starting structure. The theoretical structures of the mutants were generated with FoldX. The starting structure was solvated in a dodecahedron box with a TIP3P water model. Na⁺ and Cl^- ions were used for system neutralization. A CHARMM36 forcefield was applied for protein topology generation and ligand parameterization [40]. The starting structure was first minimized with the steepest descent algorithm until a maximum force lower than 10 kJ/mol∙nm was reached. The system was then equilibrated in an NVT ensemble at 300 K for 50,000 steps, 2 fs for each step. The non-bonded interactions were cut off at 12 Å and updated every 20 steps. Another equilibration was performed in an NPT ensemble with the same parameters. A dynamic simulation was then conducted for 10 ns, with the coordinates saved every 10 ps.

5. Conclusions

Rational enzyme design typically involves the steps of docking a ligand in its near-attack conformation, selecting surrounding residues, and calculating binding energies after virtual mutations. Although achieving absolute design accuracy is difficult, this protocol usually generates an operable and referable library. In this study, we aimed to further improve the hesperidin glycosylation activity of our previously obtained single-point mutant CGTase-Y217F by rationally designing the enzyme using Rosetta flex-ddG. A library of twenty variants was designed, and three were found to have improved activity. A molecular dynamic simulation showed that introducing a second-point mutation with a stabilization effect kept the bond-forming atoms closer together and significantly reduced repulsion between the enzyme and hesperidin.