Transglycosylation abilities of β-d-galactosidases from GH family 2

Abstract

The ability to predict the transglycosylation activity of glycosidases by in silico analysis was investigated. The transglycosylation abilities of 7 different β-d-galactosidases from GH family 2 were tested experimentally using 7 different acceptors and p-nitrophenyl-β-d-galactopyranoside as a donor of galactosyl moiety. Similar transglycosylation abilities were confirmed for all enzymes originating from bacteria belonging to Enterobacteriaceae, which were able to use all tested acceptor molecules. Higher acceptor selectivity was observed for all others used bacterial strains. Structure models of all enzymes were constructed using homology modeling. Ligand-docking method was used for enzymes-transglycosylation products models construction and evaluation. Results obtained by in silico analysis were compared with results arisen out of experimental testing. The experiments confirmed that significant differences in transglycosylation abilities are caused by small differences in active sites composition of analyzed enzymes. According to obtained result, it is possible to conclude that homology modeling may serve as a quick starting point for detection or exclusion of enzymes with defined transglycosylation abilities, which can be used for subsequent synthesis of e.g., pharmaceutically interesting glycosides.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13205-021-02715-w.

Introduction

The concept of carbohydrate function has been gradually changing during history. While previously they were considered mainly as a source of energy and building blocks, nowadays it is known, that they possess many different specifics and irreplaceable functions for the living organisms (Nelson and Cox 2005). Their connection to different types of molecules can dramatically change the properties of these molecules, such as solubility, bioaccessibility, stability or immunogenicity. Other events, such as degradation or protein folding, may be influenced as well. However, carbohydrates can also contribute to the function of the resulting glycoconjugate as they are able to mediate specific interactions. This ability of saccharidic structures is employed in many very important biological processes including fertilization, embryogenesis, development or immune response, but it plays also an important role in symbiosis or pathogenic events. The possibility to applicate different types of oligosaccharides, polysaccharides or glycoconjugates in the pharmaceutical, food or cosmetic industry makes them an interesting target of scientific research. However, the synthesis of defined saccharide structures is still far from being straightforward, as multiple reactive groups with similar reactivity are present in one molecule and correct stereochemistry must be also ensured (Wang and Huang 2009; Lu et al. 2020).

Enzymatic synthesis seems to be more attractive than usual chemical production as it enables avoiding harsh conditions, toxic compounds, and a cascade of laborious steps typical for chemical synthesis. Two different types of enzymes are able to catalyze the formation of the glycosidic bond. Application of glycosyltransferases (responsible for the synthesis of glycosides in vivo) in biotechnology is slowed by the high price of the substrate molecules and strict acceptor specificity. These problems may be solved by the second class of enzymes, by glycosidases. Their natural function is to cleave different types of glycosidic bonds, however, under specific conditions, they may be used for glycosidic bond synthesis, although the final yields are usually lower than in reactions catalyzed by glycosyltransferases. Genetic engineering may solve this drawback (Wang and Huang 2009; Bojarová and Křen 2011).

Glycosidases catalyze the hydrolysis of glycosidic bonds by different types of catalytic mechanisms – retaining or inverting configuration. Retaining enzymes, using a two-step double displacement mechanism resulting in the maintenance of anomeric configuration in the arising product, are often able to catalyze so-called transglycosylation. Another hydroxyl group-possessing acceptor molecule, other than water (e.g. an alcohol, a saccharide, etc.) is used for the cleavage of the covalent intermediate formed during the first hydrolytic step in this case. The possibility to use glycosidases for glycosides production was demonstrated in many cases. Genetically engineered biocatalysts with improved transglycosylation ability were also prepared (Wang and Huang 2009; Lombard et al. 2014).

As for other enzymes, classification based on their substrate specificity can be used also for glycosidases. However, the broad substrate specificity of some glycosidases complicates this simple approach. Therefore, another classification system, accessible in the CAZy (Carbohydrate-Active Enzyme) database, using the categorization according to the amino acid sequence similarity of enzymes, is the most commonly used. Up to date, 167 families of glycosidases were defined. Enzymes in one family usually share the same catalytic mechanism although a few exceptions can be found. However, for some families the used catalytic mechanism is still unknown or only inferred (e.g., family GH35, which includes also β-d-galactosidases). In some cases, one family contains only enzymes with the same substrate specificity (e.g., family GH147 containing only β-d-galactosidases with EC 3.2.1.23), but it is more common that enzymes with different substrate specificities (different EC numbers) are members of one family. It is important to emphasize that enzymes, which should be newly assigned to this classification system, do not need to be biochemically characterized (Lombard et al. 2014; Henrissat and Davies 1997; Naumoff 2011). β-d-Galactosidases (EC 3.2.1.23) are members of 6 CAZy families (GH1, GH2, GH35, GH42, GH59, GH147 and GH165), all of them using most probably retaining mechanism of catalysis. All these families (except GH165) belong to the Clan GH-A, which is characterized by (β/α)₈ structure. One of the best-known glycosidases is β-d-galactosidase from E. coli. This enzyme from the glycoside hydrolase family 2 was studied deeply and its structure and catalytic mechanism were elucidated. β-d-Galactosidases have a quite wide range of industrial and medicinal applications, including lactose-free milk production (using their hydrolytic activity) or prebiotics synthesis, and bioactive compounds modifications (using the transglycosylation abilities). In combination with other suitable enzymes β-d-galactosidases can be used for the synthesis of molecules containing complex saccharidic parts. For this reason, many current studies have focused on β-d-galactosidases (Lu et al. 2020; Lombard et al. 2014; Saqib et al. 2017).

A few years ago, many modern methods of in silico analysis were evolved. An increasing number of experimentally determined protein structures as well as so-called homology models, more powerful hardware, and sophisticated software tools, have made molecular docking an important tool used in various fields of biochemical research. Thus, a new modern era of research has started with the possibility to predict the properties of new compounds (e.g., drug candidates) without the need of demanding laboratory experiments (structure-based virtual screening, SBVS) (Šícho and Svozil 2017; Jorgensen 2004; Pagadala et al. 2017). Provided that it is possible to predict the probability of the course of the transglycosylation reaction on the basis of in silico analysis, this would mean a significant shift in the synthesis of oligosaccharides and glycoconjugates with defined structures, as it would be possible to avoid lengthy experimental work for suitable enzyme selection. At the same time this information could be used in the design of further modified enzymes (by genetic engineering methods) to optimize their transglycosylation properties in relation to a particular substrate.

The aim of this project was to analyze the possible relationship between the structure and transglycosylation abilities of selected β-d-galactosidases, classified to the CAZy family 2 (GH2), using experimental and in silico methods. Laboratory transglycosylation experiments with different acceptor molecules were compared with results obtained from the ligand-docking method, where transglycosylation products represented the ligand.

Materials and methods

Bacterial strains and cultivation conditions

The strains Citrobacter freundii DBM 3127, Enterobacter cloacae DBM 3125, Escherichia coli DBM 3125, Klebsiella pneumoniae DBM 3186, Lactobacillus delbrueckii subsp. bulgaricus CCDM 1005, Lactococcus lactis subsp. lactis CCDM 1005, and Streptococcus thermophilus CCDM 70 came from a collection of microorganisms at University of Chemistry and Technology Prague. De Man, Rogosa and Sharpe (MRS) medium (HiMedia) containing 2% lactose was used for cultivation of L. delbrueckii subsp. bulgaricus, L. lactis subsp. lactis, and S. thermophilus and Luria–Bertani (LB) medium (HiMedia) containing 2% lactose was used for C. freundii, E. cloacae, E. coli, and K. pneumoniae. All strains were grown at 37 °C, 200 revolutions per minute (RPM) to the stationary phase.

Enzyme activity assay

The hydrolytic activity of β-d-galactosidases was tested using p-nitrophenyl-β-d-galactopyranoside (pNPβ-d-Gal, Carbosynth, UK), where the released p-nitrophenol was measured spectrophotometrically at 420 nm. The reaction was performed in phosphate buffer (100 mmol.l⁻¹, pH 7.5) at 37° C for 5 min. Substrate pNPβ-d-Gal was used at the concentration 40 mmol.l⁻¹ in the reaction mixture. The reaction was terminated by the addition of an equivalent volume of 10% Na₂CO₃. 1 U was defined as an amount of the enzyme capable of releasing 1 μmol of p-nitrophenol per 1 min.

Transglycosylation reaction

The cell suspension used for the catalysis of the transglycosylation reaction was prepared from cells in the stationary phase. Cells after cultivation were centrifugated (5000 g, 4 °C, 10 min) and washed using 100 mM phosphate buffer, pH 7,5. After washing, the aliquots exhibiting the β-d-galactosidase activity of 16 U.ml⁻¹ were stored at − 20 °C. The transglycosylation activities of β-d-galactosidases were tested using 50 mM pNPβ-d-Gal as a donor of d-galactose. Acceptors d-glucose, d-galactose, d-fructose, d-mannose and maltose were used in the concentration of 250 mmol.l⁻¹ and N-acetyl-d-glucosamine (NAG) and sorbitol in the concentration of 500 mmol.l⁻¹.

The reaction mixture contained a donor of d-galactose unit, the acceptor of d-galactose unit, and cell suspension in ratio 4:5:1 (vol.). Reaction time was 2 or 4 h for all performed reactions. Reaction temperature of 37 °C as the enzyme temperature optimum was used for all cultures except of E. cloacae. For this strain, the reaction was performed at 50 °C - temperature optimum of β-d-galactosidase from this strain.

Thin-layer chromatography

The samples after transglycosylation reactions were applied to a thin layer chromatography (TLC) silica gel plate containing a fluorescence indicator (254 nm). 1 µL of each reaction mixture was spotted on a plate alongside with the standards. Plates containing saccharides as acceptors were developed in a mobile phase consisting of butanol: acetic acid: water (3:3:2). Chromatograms were stained using 0.1 M 2-methylresorcinol in 5% (v/v) solution of sulphuric acid in ethanol and visualized by heating the plates.

Homology modeling and molecular docking

Homology models of β-d-galactosidases used in this work were created using Modeller (Webb and Sali 2016) program. The templates for generating these models were four experimentally determined structures of β-galactosidases originating from Arthrobacter sp. C2-2 (PDB:1YQ2) (Skálová et al. 2005), E. coli (PDB: 5A1A) (Bartesaghi et al. 2015), Bacteroides thetaiotaomicron (PDB: 3BGA) (unpublished) and Thermotoga maritima (PDB: 6S6Z) (Miguez Amil et al. 2020). Several homology models were created, and the highest-quality models (in the terms of structural correctness and possible faulty structural parts) were used for molecular docking. Assessment of the model quality was done by program ProSA (Sippl 1993). Molecular docking was performed using the program Glide, within Schrödinger Materials Science Suite (Materials Science Suite 2018). First, the structure of substrate (pNPβ-d-Gal), sodium and magnesium ions and water molecules number 8894, 8624 and 8701 were taken from E. coli enzyme (PDB ID: 1JYW) (Juers et al. 2020) and placed into the aligned model. Next, as a part of this software package, Schrödinger Protein Preparation Wizard (Sastry et al. 2013) was used to further optimize the enzyme structure (by optimization of a hydrogen bond network and energy minimization). Finally, pNPβ-d-Gal was removed prior docking.

LigPrep program was used to generate different conformations of the ligands. Possible products of transglycosylation reactions were used as ligands. For every saccharide acceptor and for sorbitol its galactosylated forms were built. Program Glide (Friesner et al. 2006) was used to define the binding pocket in each enzyme as a sphere (r = 20 Å) around Trp residue corresponding to W568 in 5A1A. Within this grid, ligands had freedom of flexibly exploring a maximum of 50 binding positions. Sampling of ring conformations was disabled in docking.

Results and discussion

Selection of β-d-galactosidases-producing organisms

Producers of β-d-galactosidases for the experimental work were selected according to two criteria. First, each tested organism produces only one β-d-galactosidase isoform, and second, the enzyme is included in the GH2 family. In this way, β-d-galactosidases from 7 different bacterial strains were selected: β-d-galactosidase from L. delbrueckii subsp. bulgaricus (Uniprot: Q1G9Z4), from L. lactis subsp. lactis (Uniprot: Q48727), from S. thermophilus (Uniprot: P23989), from E. cloacae (Uniprot: Q2XQU3), from K. pneumoniae (Uniprot: B5XQY2), from C. freundii (Uniprot: Q5ETW7) and from E. coli (Uniprot: P00722).

Transglycosylation reactions

The transglycosylation abilities were tested for 7 chosen enzymes using bacterial cultures instead of purified enzyme preparation and different acceptors. Reaction times 2 and 4 hours were tested, and the experiments confirmed that the prolongation of the reaction has no significant effect on the recovery of transglycosylation. The reactions were performed at 37 °C (temperature optimum of β-d-galactosidases) in case of all bacterial cultures except of Enterobacter cloacae. β-d-Galactosidase originating from this bacterium has the temperature optimum at 50 °C and for this reason this temperature was used as more appropriate. The production of β-d-galactosylated acceptors was analyzed using TLC chromatography (data shown in Supplementary data - Fig S1-S7) and the intensity of the spots was evaluated on a relative scale, expressing the yield of β-d-galactosyl residue transfer to the various acceptors (Table 1). Although all enzymes were from the same GH family, therefore, they should have similar structure motives, they differed significantly in their transglycosylation properties. It is possible to trace similar transglycosylation abilities for all enzymes originating from bacteria belonging to Enterobacteriaceae, it means E. coli, E. cloacae, K. pneumoniae, and C. freundii. In experiments with these bacterial cultures all used acceptor molecules were galactosylated. The transgalacosylation abilities of the strains L. delbrueckii subsp. bulgaricus, L. lactis, and S. thermophilus were substantially different. β-d-Galactosidases originating from bacteria belonging to Streptococcaceae (L. lactis and S. thermophilus) seem to be significantly more selective with respect to the transglycosylation acceptor, as only in the case of maltose, sorbitol, and NAG (for L. lactis) or d-galactose, sorbitol, and NAG (for S. thermophilus) transglycosylation products were detected in the reaction mixtures.

Table 1.

Relative yield of β-d-galactosyl residue transfer to different acceptors. Rankings ++, + and − were made based on TLC analysis

Source	Acceptor
	d-glucose	d-galactose	d-fructose	d-mannose	Maltose	Sorbitol		NAG
L. delbrueckii subspec. bulgaricus	+	+	−	−	+	+		+
L. lactis	−	−	−	−	+	+		+
S. thermophilus	−	+	−	−	−	+		+
E. cloacae	+	+	+	+	+	+		++
K. pneumoniae	++	++	++	++	++	++		++
C. freundii	+	+	++	+	++	++		++
E. coli	+	+	+	+	+	+		+

NAG N-acetyl-d-glucosamine

Homology modeling

Individual sequences of the chosen enzymes were aligned using MUSCLE program (Edgar 2004) (Table 2) as a multiple sequence alignment, from which template-model alignments were extracted as binary sub-alignments. In the next step, homology models were created and analyzed. The quality of homology models was characterized by Z-score, which is automatically assigned to each analyzed structure by ProSA-web program (Sippl 1993). Z-score gives an estimate of how close the quality of the homology model is to the experimentally determined structures of proteins of a similar size. In other words, values of this score for acceptable and high-quality models are within a certain range, characteristic for native proteins. For the models used in this work, Z-scores are presented in Table 3. Corresponding models are depicted in Fig. 1.

Table 2.

The alignment of amino acid sequences of β–d-galactosidases (Muscle)

% identity		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
B. thetaiotaomicron*	(1)	100	39	32	34	30	32	33	35	34	33
T. maritima*	(2)	39	100	37	38	35	36	36	39	36	37
L. bulgaricus	(3)	32	37	100	49	32	32	32	33	33	34
S. thermophilus	(4)	34	38	49	100	32	31	31	33	32	33
A. sp. C2-2 (iso-1)*	(5)	30	35	32	32	100	30	32	34	32	34
L. lactis	(6)	32	36	32	31	30	100	48	49	51	51
K. pneumoniae	(7)	33	36	32	31	32	48	100	57	58	59
E. cloacae	(8)	35	39	33	33	34	49	57	100	64	65
C. freundii	(9)	34	36	33	32	32	51	58	64	100	82
E. coli*	(10)	33	37	34	33	34	51	59	65	82	100

*Enzymes with known 3D structure used as a templates for modeling

Table 3.

Comparison of Z-values of individual structures

Bacteria producing β-d-galactosidase	Z-score of the homology model
E. cloacae	− 10.65
C. freundii	− 10.34
K. pneumoniae	− 9.98
L. delbrueckii	− 6.69
L. lactis	− 8.09
S. thermophilus	− 8.23
Arthrobacter sp. C2-2 (iso-1)	− 11.42*
E. coli	− 10.15*
Thermotoga maritima	− 12.69*

*This is not the score of the homology model but the score for the structure of β-galactosidase from Arthrobacter sp. C2-2 (iso-1) (PDB:1YQ2), E. coli (5A1A), and Thermotoga maritima (6S6Z), used here for the comparison

Fig. 1.

Experimentally determined structures (a-c) and homology models (d-i) of studied β-galactosidases. Models of E. coli (a, PDB ID: 5A1A), Arthrobacter sp. C2-2 (b, PDB ID: 1YQ2), B. thetaiotaomicron (c, PDB ID: 3BGA), L. delbrueckii subsp. bulgaricus (d), S. thermophilus (e), L. lactis subsp. lactis (f), K. pneumoniae (g), E. cloacae (h) and C. freundii (i) are shown in the same orientation. Catalytic glutamates are depicted as sticks. Main differences are highlighted by different colors (residues corresponding to 421–427 in blue, 505–530 in green and 795–803 in magenta, E. coli numbering)

Lower (more negative) values of the Z-score indicate higher quality of the homology model. It can be seen from the data in Table 3 that few models had lower quality comparing to others, having in mind that the Z-score is assigned based on evaluation of the whole structure. Clearly, lower quality of enzyme homology model would lead to less reliable docking results. However, for docking purposes it is particularly important to achieve high-quality modeling of the active site.

Molecular docking

In this study, we docked products of transglycosylation reactions. The main job of enzymes as catalysts is to stabilize the transition states of the reaction. An ideal option would be to dock transition states. As far as our knowledge goes, transition state docking has been tested in various molecular modeling studies of enzymes, however, this approach is far from trivial, especially for reactions involving covalent catalysis, which is the case of retaining glycosidases. As a compromise, we docked products of transglycosylation reactions. Docking of products may elucidate the capability of the enzyme to recognize the galactosyl moiety and acceptor in the mutual orientation suitable for the transglycosylation reaction.

Positions of water molecules in a proximity to the active site were determined based on the position of water and ion molecules in experimentally determined structures, PDB: 1JYW (Juers et al. 2020). The best docking results were obtained with extra precision (XP) mode of the Glide program and the water molecules positioned based on water and ions in the active site from PDB: 1JYW (β-galactosidase from E. coli in complex with pNPβ-d-Gal).

Algorithm implemented within Glide divides each ligand into its core and rotamer groups (groups attached to the core by rotatable bonds). In the first step of molecular docking, different conformations of the core are generated and for each of them, possible position and orientation within the protein grid is explored. In this work, we defined the ligand center as the atom in the middle of the distance between the two furthest atoms of the core. This ligand center was allowed to explore every part of the grid, while the rest of the ligand was allowed to leave the borders of the grid. This ensured smaller constraints of the ligand position but also a more exhaustive search of its best pose (term pose is here used to define a position, orientation towards the enzyme active site, core conformation and rotamer group conformation of the ligand). To supplement and compare in silico results with the experimental data, we docked all glycosylated saccharide acceptors as ligands (Supplementary data - Tab S1).

Best docking poses for each docked compound were visually inspected in the Maestro environment (Materials Science Suite 2018). Docking poses in which the galactose moiety of ligand was aligned with the galactose moiety of pNPβ-d-Gal from 1JYW were considered as suitable for transglycosylation (+ in Table S1). Acceptors were classified (Table 4) as good acceptors (++), moderate acceptors (+) or poor acceptors (−) depending on the number of hydroxyl moieties suitable for transglycosylation, considering the total number of hydroxyl moieties in the acceptor molecule.

Table 4.

Consensus results of docking. All possible transglycosylation products were docked into active sites of all enzyme (see Supporting Information Table S1)

Source	Docking of individual galactosylated acceptors Acceptor
	d-glucose	d-galactose	d-fructose	d-mannose	maltose	sorbitol		NAG
L. delbrueckii subspec. bulgaricus	+	+	++	++	+	+		++
L. lactis	++	++	++	++	++	++		++
S. thermophilus	+	+	+	+	−	+		+
E. cloacae	++	++	++	++	+	++		++
K. pneumoniae	++	++	++	++	++	++		++
C. freundii	+	++	++	++	++	++		++
E. coli	+	+	+	++	++	++		++

Correct binding poses of the galactose moiety were interpreted as the ability of the enzyme to catalyze corresponding transglycosylation reaction. Rankings ++, + and − were made based on a number of possible transglycosylation products

Comparison of transglycosylation reactions and protein–ligand docking

In this study, we use protein-ligand docking of transglycosylation products to predict the transglycosylation properties of the studied enzymes. An ultimate answer to transglycosylation preferences can be, at least in principle, obtained by mixed quantum mechanical and molecular mechanical (QMMM) studies, which can predict energy barriers, i.e. kinetic constants. There is a pioneer study focused on E. coli β-galactosidase (Brás et al. 2010). However, comparison of different enzymes and acceptors by this method is difficult due to high computational costs and due to tiny differences in activation barriers.

Results of docking are summarized in Table 4. According to these results, enzymes from E. cloacae, C. freundii and L. lactis subsp. lactis were predicted as the most transglycosylating ones. This prediction was correct for E. cloacae and C. freundii. Enzymes from K. pneumoniae and E. coli were also predicted to be good catalysts of transglycosylations. This prediction was correct for E. coli. In contrast, the enzyme from K. pneumoniae was the best catalyst of transglycosylation in the experiment. Enzymes from L. delbrueckii subsp. bulgaricus and S. thermophilus were correctly predicted as poor catalysts of transglycosylations. Figure 2 shows an example docking poses of galactosylated N-acetylglucosamine in the binding site of the enzyme from K. pneumoniae (experimentally verified as a good catalyst for this reaction).

Fig. 2.

Docking of transglycosylation products into the active site of K. pneumoniae β-galactosidase. Binding poses of β-d-Gal-(1 → 3)-α-d-GlcNAc (a), β-d-Gal-(1 → 4)-α-d-GlcNAc (b) and β-d-Gal-(1 → 6)-α-d-GlcNAc (c) are shown in the same orientation

To explain differences between activities of enzymes and failures of some predictions we visually inspected active sites of all enzymes. Active sites are highly conserved. One visible difference was in the orientation of Phe601 (numbering according to E. coli) (Fig. 3a). This residue was conserved in all tested enzymes, but these enzymes differ in a side chain rotamer. In models of enzymes predicted as poor transglycosylators this residue was partially blocking the active site. This may be an artefact of homology modeling as well as a mechanism by which residues in the vicinity of this phenylalanine affect transglycosylation by influencing its side chain rotamer equilibria. Further studies would be required to support or exclude this effect.

Fig. 3.

Structural differences of studied enzymes. a Comparison of orientation of phenylalanine residues corresponding to Phe601 (E. coli numbering). The residue from L. delbrueckii subsp. bulgaricus and S. thermophilus are in blue, others in orange. Other residues were

taken from E. coli; b Loop formed by residues 414–427 (red) in the enzyme from L. delbrueckii subsp. bulgaricus (orange, in comparison with E. coli in blue) and c Substitution of His357 (E. coli numbering, histidine residues of all enzymes are in orange) by Leu (blue) in the enzyme from Enterobacter cloacae

Other differences (Fig. 3b, c) included, for example, the fact that β-galactosidase from L. delbrueckii subsp. bulgaricus contains a loop (414–424) which differs in position from loops of other enzymes, which are located compactly on the side of the active centers. Another interesting difference was found in the structure of β-galactosidase from Enterobacter cloacae, which contains Leu instead of His (used for transition state stabilization) in the position 359. All these variations seem to have an impact on the final transglycosylation process.

Rutkiewicz and co-workers (Rutkiewicz et al. 2019a, 2020) performed a structural biology study on transglycosylation reactions catalysed by a cold-adapted β-galactosidase from Arthrobacter sp. 32cB. This enzyme has interesting transglycosylation properties, including synthesis of lactulose and glycosylation of non-sacharidic acceptors (Pawlak-Szukalska et al. 2014). It has been also intensively studied to explain its adaptation to low temperature (Rutkiewicz et al. 2019b). Mutation of key catalytic residues enabled to determine structures of complexes with transglycosylation products (Rutkiewicz et al. 2020). A distant binding mode was identified for binding of sucrose. An importance of the transition of the acceptor from this distal site to the active site was highlighted by the study of Rutkiewicz and co-workers. Existence of the distal and the proximal site is relevant to binding modes of galactosylated maltose predicted in this study.

Conclusions

The goal of the work was to compare transglycosylation abilities of different β-d-galactosidases from the GH2 family using different acceptors of galactosyl moiety. Classical laboratory experiments were complemented by new in silico approaches. The experiments confirmed significant differences in transglycosylation abilities of individual enzymes. Although all tested enzymes were able to cleave artificial chromogenic substrate pNP-d-Gal and use it as a donor of galactosyl moiety and although all enzymes are according to their sequential and structural similarities classified into the same glycoside hydrolase family, they differ in the acceptor specificity. The reason was easy to find during the in silico analysis because despite the structural similarity of particular enzymes their active sites exhibit small but important differences. It is possible to conclude that enzymes with high identity and structure similarity in the active site region would have also similar transglycosylation properties. However, the fact that two enzymes are in the same glycoside hydrolase family is not sufficient information for their transglycosylation abilities prediction. Our experiments confirmed that there is a relationship between the structure of the active site of the enzyme, which can be modeled according to accessible structure data, and the structure of the molecule, which can be used as an acceptor for transglycosylation. The relationship between the number of ligand positions in a suitable orientation in the active site with transglycosylation properties of β-d-galactosidase was found. Thus, the methods of molecular modeling could be a useful tool for quick in silico searching for glycosidases that would probably be able to transglycosylated defined acceptor molecules and vice versa to exclude enzymes which probably will not be able to catalyze required tranglycosylation reaction. This was confirmed in this work, where β-galactosidases from E. cloacae, C. freundii and K. pneumonie exhibited high identity (more than 50 %) with the enzyme from E. coli, prepared models were evaluated as reliable also in the active site region, and molecular docking provided a number of ligand positions in a suitable orientation. Obtained results were in good agreement with experimental outcomes.

However, it is worth noting, that molecular docking using Glide has its limitations. Relatively rigid enzyme structure leaves the possibility of spatial clashes with ligands, making them incapable of fitting into the active site. Clearly, this becomes an even bigger issue if there are induced-fit mechanisms involved in the binding process or if the ligand binds to the alternative enzyme conformation.

The future plans involve not only the use of other docking software but also molecular dynamics simulations. By simulating the dynamics of a studied system, we will be able to achieve greater insights into ligand-enzyme interactions. Also tests with a larger number of enzymes and with other types of glycosidases could be useful.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors’ Contributions

EB: draft of the manuscript, interpretation of the acquired data. ZS: performing of in silico analysis, homology modeling and molecular docking. MT: performing of experimental analyses of transglycosylation abilities of studied enzymes. VS: design of in silico experiments, interpretation of the acquired data. PL: design of transglycosylation experiments, interpretation of the acquired data.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest in the publication.