Recognition of reactive high-energy conformations by shape complementarity and specific enzyme–substrate interactions in family 10 and 11 xylanases

Binding of xylotetraose to the family 10 xylanase (CEX) from Cellulomonas fimi and the family 11 xylanase (BCX) from Bacillus circulans was investigated using quantum mechanical calculations, molecular dynamics simulations and molecular mechanics–Poisson–Boltzmann surface area (MM-PBSA) free energy calculations.

In the reactive enzyme–substrate conformation, in which the catalytic groups are precisely positioned for the catalytic reaction, the −1 sugar moiety of the substrate adopted a B3,O (boat) conformation for CEX and a 2SO (skew-boat) conformation for BCX.

Upon binding the internal energies of the reactive substrate conformations are increased by about 9 kJ mol−1.

In contrast to this, the internal energies of substrates in all-chair conformation are increased by 33–43 kJ mol−1, with somewhat larger increase for BCX than CEX.

As a consequence, the stability order of the all-chair conformation, which is the most stable conformation in aqueous solution, and the distorted reactive conformations is reversed upon binding to the enzyme.

Thus, in addition to specific enzyme–substrate interactions, CEX and especially BCX recognise the reactive high-energy conformation by the substrate-binding site of several sugar-binding subsites complementary in shape to the reactive conformation.

The role of substrate distortion in glycosidase-catalysed reactions is discussed.


Xylanases (β-1,4-d-xylan xylanohydrolases, EC are β-glycosidases, which degrade Nature’s most abundant hemicellulose, xylan.

On the basis of amino acid sequence similarities, the majority of xylanases are members of families 10 and 11 of the glycosyl hydrolases.1,2

Xylanases of both enzyme families catalyse the hydrolysis of β-1,4-linked xylose polymer by a double-displacement mechanism and net retention of the anomeric center of the reaction.3

In the active site there are two key catalytic glutamic acid residues, one acting as an acid/base catalyst and another as a nucleophile, which forms a covalent reaction intermediate with the substrate.

The transition states to and from the covalent intermediate have considerable oxocarbenium ion character (Fig. 1a).3–5

Binding of xylotetraose (X4, Fig. 1b) to two xylanases, CEX from Cellulomonas fimi and BCX from Bacillus circulans, was studied computationally in this work.

CEX is a member of family 10 and BCX of family 11 glycosyl hydrolase.1

These two xylanases are good representatives of families 10 and 11 as several X-ray structures and a lot of other biochemical data are available about their functional properties.3

Although CEX and BCX catalyse the hydrolysis of xylan with a similar type of mechanism, there are differences in the details of the reaction.

Mechanism-based inhibitors, like 2′,4′-dinitrophenyl 2-deoxy-2-fluoro-β-xylobioside, form stable covalent glycosyl–enzyme intermediates with xylanases.

In the intermediate the catalytic nucleophile forms a covalent bond with the C1 of the xylose moiety bound in the −1 subsite of the enzyme.

In the case of CEX, the 2-deoxy-2-fluoroxylose residue bound in the −1 subsite adopts a 1C4 (chair) conformation,6 whereas in the case of BCX, the corresponding residue is bound in a 2,5B (boat) conformation.7

These structures have shed light on the atomic details of the covalent reaction intermediate, but it is less clear in which conformation the sugar of the −1 subsite is bound in the Michaelis complex.

The ambiguity is mainly due to difficulties in finding suitable substrates, which are not hydrolysed, occupy at the same time subsites also at the reducing end (+1) of the binding site, and are structurally similar to the natural substrates.3

Distortion of the reactive pyranose ring has been observed in the X-ray structures of several β-glycosidases complexed with uncleaved substrates.8–10

Lysozyme is a classical example of an enzyme binding the substrate’s reactive sugar in a distorted sofa conformation.

It has been suggested that distortion of the reactive sugar moiety is important for the rate of the enzyme-catalysed reaction.4

In contrast to this, it has been suggested that distortion plays a minor role in the catalysis.11,12

Knowledge of the energetics involved in the substrate binding in different conformations can be obtained with computer simulations and further our understanding on factors responsible for enzyme-catalysed reactions.

In this work, the binding of xylotetraose to CEX and BCX was studied computationally.

QM calculations were used to compare the (internal) energy of the substrate X4 free in aqueous solution and complexed with the two xylanases.

MD simulations of 4.0 ns were used to provide the structures of enzyme–substrate complexes and substrate free in aqueous solution, and molecular mechanics–Poisson–Boltzmann surface area (MM-PBSA) free energy calculations13 were applied to study the energetics of the binding.

The MM-PBSA method has recently been used with good success in predicting and explaining the binding of various ligands to proteins14–19.

Computational methods

MD simulations

The starting coordinates for the MD simulations were obtained from the X-ray crystal structures of the endo-1,4-β-xylanases from C. fimi (CEX, PDB code 1FH8 20) and from B. circulans (BCX, 1BVV7) determined to 1.95 Å and 1.8 Å resolution, respectively.

Hydrogen atoms were added with the Xleap program21 to the protein atoms.

His85 and His107 of CEX and His149 and His156 of BCX were protonated at the δ-nitrogen, while ε-tautomers were assigned for the rest of the histidines.

The acid/base catalysts Glu127 (CEX) and Glu172 (BCX) were protonated.

For the MD simulations X4 was docked to the −2, −1, +1 and +2 subsites of the xylanases.

The sugar moieties of the −2, +1 and +2 subsites most probably bind to the xylanases in the 4C1 conformation.

More variations have been observed in the conformations of the −1 subsite7,22,23 and it is possible that xylanases distort the −1 sugar in the Michaelis complex.

Therefore, MD simulations were done for both enzymes with X4 in an all-chair conformation and in a conformation having the −1 sugar in a distorted conformation.

For the latter simulations, X4 was docked to the binding site with the −1 moiety in different skew-boat conformations.

These structures were first visually checked and if they were found to fit in the binding site without unreasonably large steric overlap with the protein atoms, MD simulations for such enzyme–X4 complexes were started.

It turned out that simulations with distorted −1 sugar produced stable trajectories only when the sugar adopted a B3,O conformation for CEX and a 2SO conformation for BCX (Fig. 2).

An appropriate number of counterions (seven Na+ ions for CEX and four Cl ions for BCX) were added to neutralize the simulation systems.

Xylanase-X4 complexes were solvated with a truncated octahedron of TIP3P water molecules extending at least 12 Å from the protein atoms.

Crystal waters were included in the simulations.

This resulted in 9087 water molecules for CEX and 6365 water molecules for BCX.

For the MD simulations the water molecules and hydrogens of protein and substrate were first energy minimized for 1000 steps, heated to 300 K in 7.5 ps and equilibrated by 50 ps at a constant temperature of 300 K and pressure of 1 atm.

After that the whole simulation system was minimized for 1000 steps, the temperature of the system was increased to 300 K in 7.5 ps and equilibrated for 300 ps using a time step of 1.5 fs and the SHAKE algorithm24 to constrain the bonds involving hydrogen atoms to their equilibrium values with a geometrical tolerance of 10−5 Å.

After that the production simulations of 4.0 ns were started.

During the production simulations structures were saved every 5 ps for analyses.

The cutoff for Lennard-Jones interactions was 8 Å and the particle mesh Ewald (PME) method25 was used for the treatment of electrostatic interactions.

A charge grid of ∼1 Å with cubic B-spline interpolation and the direct sum tolerance of 10−5 were used.

Continuum model correction was applied for energy and pressure.

The MD simulations were done with the AMBER 7.0 program26 and the Cornell et al. force field.27

The modified parameters of the GLYCAM_93 set28 were used for the substrates and their atomic point charges were calculated using the RESP method29 at the HF/6-31G(d) level.

Periodic water box simulations of 4.0 ns were done for free xylotetraose in an all-chair (4C1) conformation (X4(c)) and in a conformation in which the −1 sugar moiety was in the 2SO conformation (X4(sb)) while the other moieties were in the 4C1 conformation.

The trajectories were analysed with the ANAL, PTRAJ and CARNAL programs of AMBER 7.0 and visually examined with the assistance of the Moil-view program30.

MM-PBSA free energy calculations

In the MM-PBSA method13 MD simulations in explicit water are first carried out for the system studied.

After that solvent water molecules of the “snapshot” structures collected from the molecular dynamics trajectories are removed, and the average free energies (G) of the protein–ligand complex and separated protein and ligand are calculated as a sums of the average molecular mechanical gas-phase energies (EMM), solvation free energies (ΔGsolv) and entropy contributions (−TS):G = EMM + ΔGsolvTSThe average free energies of the species are then used to calculate binding free energies (ΔG) for protein-ligand association.

In this work the molecular mechanical (EMM) energies were calculated using the SANDER program of AMBER7 with all protein pair-wise interactions included using a dielectric constant (ε) of 1.

The solvation free energy (ΔGsolv) is the sum of electrostatic solvation, calculated by the finite-difference solution of the Poisson–Boltzmann equation (ΔGPB) with the Delphi program,31 and non-polar solvation energy (ΔGnp), calculated from the solvent-accessible surface area (SASA):ΔGsolv = ΔGPB + ΔGnpA probe radius of 1.4 Å and atomic radii of the PARSE parameter set32 was used to determine the molecular surface.

Atomic charges of the Cornell et al. force field27 were used for amino acid residues and RESP charges were calculated at the HF/6-31G* level for the substrate.

An 80% boxfill cubic lattice and a grid resolution of 0.5 Å (grid point)−1 were used in the Delphi calculations.

The MSMS program33 was used to calculate the SASA for the estimation of ΔGnp using eqn. (3) and γ = 0.02268 kJ mol−1−2 and β = 3.85 kJ mol−132Gnp = γ × SASA + βThe calculated ΔGbind values can be divided into contributions of different force-field terms and solvation energies.

This allows us to analyse the energetics of the ligand binding in more detail.

The energy terms used in the binding energy analyses are explained in the footnote of Table 1.

The energy contribution from entropy changes upon ligand binding was not included here primarily because the main interest in this work was in calculating the binding free energies of one substrate molecule in different conformations.

This was justified by the good agreement between the calculated and experimental relative binding free energies in several earlier studies applying the same approximation.13–18

In addition, there is no straightforward way to quantitatively calculate entropy contribution to binding.13,34,35

The single-trajectory method was applied in this work to calculate the binding free energies.

This means that the snapshot structures of the protein–ligand complex and separated protein and ligand were taken from the MD trajectory of the protein–ligand complex.

It has been observed that the single-trajectory method provides fairly good estimates for the relative binding energies.13,15,16

However, in cases where there are reorganizations in the protein structure due to ligand binding, separate trajectories are often needed to get reliable binding energies13,36,37.

QM calculations

Density functional theory and ab initio QM calculations were done with the Gaussian98 program.38

The geometries of xylose monomers were optimised at the B3LYP/6-31G(d) and MP2/6-31G(d) levels.

These geometries were then used to calculate energies with the 6-311G(d,p) and 6-311G(3d,2p) basis sets.

The energies of xylotetraose snapshots collected from the MD simulations were calculated at the B3LYP/6-31G* level without geometry optimisation (B3LYP/6-31G(d)//MM).

Results and discussion

Structures of the xylanase–xylotetraose complexes

MD simulations of 4.0 ns were done for CEX and BCX complexed with X4 in two different conformations.

In the CEX(c) and BCX(c) simulations X4 was in an all-chair (4C1) conformation.

In the CEX(b) and BCX(sb) simulations the −2, +1 and +2 moieties were in the chair conformation and the −1 sugar in a skew-boat starting conformation.

On the basis of the root-mean-squared deviations (RMSD) of the protein atoms and the calculated free energies (see below) all four simulations produced stable trajectories.

The average RMSDs for the Cα atoms from the corresponding X-ray structure during the last 1.0 ns of the 4.0 ns MD simulation were 1.04 Å for the CEX(c), 0.99 Å for the CEX(b), 0.79 Å for the BCX(c), and 0.87 Å for the BCX(sb) simulation.

In the CEX(c) simulation all monomers of the substrate stayed in an almost relaxed 4C1 conformation during the simulation time.

Inspection of the arrangement of the catalytic nucleophile, the acid/base catalyst, the reacting C1 and the glycosidic oxygen of the substrate revealed that this conformation is not a reactive one (i.e. conformational arrangements would be required for the catalytic reaction).

In the case of CEX(b) simulation, the −1 sugar adopted during the equilibration period a slightly twisted B3,O conformation and stayed in this conformation for the rest of the simulation (Fig. 3a).

The average C1-C2-C4-C5 torsion angle during the last 1.0 ns of the production simulation was −2.1° (standard deviation = 8.3°), close to 0° of an ideal boat conformation.

In the B3,O conformation the −1 sugar is in a reactive conformation: (i) the catalytic nucleophile, Glu233, forms a hydrogen bond with the 2-OH of the −1 sugar and is ready to attack the C1 carbon, (ii) the breaking glycosidic bond is in a pseudoaxial orientation suitable for a facile bond cleavage, and (iii) the acid/base catalyst (Glu127) is suitably positioned to deliver a proton to the leaving glycosidic oxygen.

In the BCX(c) simulation the −1 sugar stayed in a slightly distorted 4C1 conformation for the first 1.8 ns.

After that the conformation of the −1 monomer changed to a 1C4 conformation where it stayed the rest of the 4.0 ns simulation.

Like in the CEX(c) simulation, the interactions of the reactive groups were not suitable for the catalytic events without conformational changes.

During the equilibration period of the BCX(sb) simulation, the −1 sugar adopted a 2SO conformation while the rest of the monomers (−2, +1, and +2) stayed in a 4C1 conformation.

On the basis of structures and energies, this X4 conformation was particularly stable during the whole 4.0 ns simulation.

In the 2SO conformation the −1 sugar has the leaving glycosidic bond in a pseudoaxial orientation and the reactive groups ideally positioned for the catalytic reaction (Fig. 3b).

In the reactive substrate conformations of CEX and BCX the leaving +1 sugar is rotated so that the acid/base catalysts, Glu127 and Glu172, may approach the glycosidic oxygens without steric hindrance.

In the case of CEX the acid/base catalyst protonates the glycosidic oxygen from the anti direction relative to the endocyclic C1–O6 bond of the −1 sugar ring and in the case of BCX from the syn direction20.

Single-trajectory binding free energies

The free energies for the binding (ΔGbind) of X4 calculated from the CEX(c), CEX(b), BCX(c) and BCX(sb) simulations as a function of simulation time are shown in Figs. 4a and 4b.

The free energy analysis of ΔGbind values calculated from the last two ns of 4.0 ns MD simulations are listed in Table 1.

Since in the single-trajectory method the structures of the isolated ligand and enzyme are taken from the enzyme-ligand complex, these energies measure the interactions present between the enzyme and ligand in the structure of the complex, and no changes in the internal energy of protein or ligand taken place upon ligand binding are included.

In the case of CEX(c) simulation, the ΔGbind is on average −90.8 kJ mol−1 during the first 1.0 ns, increases to −72.8 kJ mol−1 and −67.8 kJ mol−1 for the next two nanoseconds, respectively, and decreases to −87.9 kJ mol−1 for the last 1.0 ns of the 4.0 ns MD simulation.

More consistent behaviour is seen for CEX(b): ΔGbind is −58.6 kJ mol−1 in the first 1.0 ns but as the simulation progresses it decreases being −77.0 kJ mol−1 when calculated from the structures of the last 1.0 ns.

In the case of BCX, the average ΔGbind of the BCX(c) simulation is −89.1 kJ mol−1 and −85.4 kJ mol−1 in the first and second ns of simulation.

However, at about 1.8 ns a conformational change takes place and the −1 sugar adopts an 1C4 conformation.

As a consequence, the ΔGbind increases sharply being −46.9 kJ mol−1 when calculated from the structures of the last 1.0 ns.

The ΔGbind of BCX(sb) is exceptionally stable and stays between −84.5 and −87.0 kJ mol−1 during the whole 4.0 ns of MD simulation.

In the case of CEX the calculated ΔGbind value predicts (Table 1) that the all-chair substrate conformation (X4(c)) and the reactive conformation (X4(b)) are bound equally tightly (−77.8 kJ mol−1vs. −78.0 kJ mol−1).

The analysis of different components of the ΔGbind values of Table 1 show, that electrostatic enzyme–substrate interactions (ΔEele) clearly favour (by 93.9 kJ mol−1) the binding of the reactive conformation as compared to X4(c).

On the other hand, the van der Waals interactions (ΔEvdW) favour the binding of X4(c) over X4(b) by 28.7 kJ mol−1.

In the case of BCX, ΔEele favours the binding of X4(sb) by 78.1 kJ mol−1 whereas ΔEvdW is slightly unfavourable (4.0 kJ mol−1) for the binding of X4(sb) as compared to X4(c).

Thus, in the cases of CEX and BCX the specific enzyme–substrate interactions favour the binding of the reactive conformations, whereas the unspecific van der Waals interactions are more favourable for the binding of the non-reactive conformations.

On the basis of ΔGbind values, CEX binds xylotetraose in an all-chair conformation and a conformation having a distorted −1 sugar (B3,O) with comparable affinities, and BCX prefers a conformation having the −1 sugar in a reactive (2SO) high-energy conformation.

This may partly explain why CEX binds various xylobiose-derived imino sugars and xylobiose much tighter (102–104 fold) than BCX, which prefers distorted substrates.

For example, the inhibition constant (Ki) for xylobiose is 4800 μM for CEX and 80 000 μM BCX22.

Intramolecular energies of the free and enzyme-bound xylotetraose conformations

When a substrate is bound to the enzyme active site its structure is deformed to optimise the enzyme–substrate interactions.

This evidently results in increase in the substrate’s internal energy (deformation energy) as compared to its energy in aqueous solution.

The estimation of the substrate deformation energy was one of the main goals of this study.

To this end, MD simulations of 4.0 ns were done also for free X4 in aqueous solution in an all-chair conformation (X4(c)) and in a conformation in which the −1 sugar was in the 2SO conformation and the other moieties in the 4C1 conformation (X4(sb)).

The internal energies obtained from these simulations were compared to those of the enzyme-bound X4.

The 4C1 conformation has been experimentally found to be the most stable conformation for most natural carbohydrates.39

In the following discussion X4(c,aq) and X4(sb,aq) denote the all-chair and skew-boat (−1 sugar) xylotetraose conformations free in aqueous solution, X4(c,CEX) and X4(b,CEX) denote the all-chair and boat (−1 sugar) conformations bound to CEX, and X4(c,BCX) and X4(sb,BCX) denote the all-chair and skew-boat (−1 sugar) conformations bound to BCX.

Because it is not guaranteed that the MM force-field correctly describes the distorted substrate conformations, we first compared the energies of xylotetraose conformations calculated with MM and QM methods.

It must be noted here that MM force-fields in general are not parameterised to reproduce exact gas-phase energies but molecular properties in the condensed phase.

The accuracy of the B3LYP/6-31G(d) method, which was used in the QM energy calculations of xylotetraose conformations, was first checked by calculating the energies of 4C1, 2SO and B3,O conformations of xylose monomer with various QM methods (up to MP2/6-311G(d,p)//MP2/6-31G(d) and B3LYP/6-311G(3d,2p)//B3LYP/6-31G(d) levels).

The B3LYP/6-31G(d) calculations were found to provide the relative energies of the conformations with reasonable accuracy (Table 2).

The energy comparison of X4 conformations was done by calculating the average MM (the same force-field was used as in MD simulations) and QM (at the B3LYP/6-31G(d)) gas-phase energies for snapshot structures collected from the first nanoseconds of the MD simulations.

The average MM and QM energies were calculated for the MD snapshot structures without geometry optimisation.

The relative average gas-phase energies are reported in Table 3.

These results show that excluding X4(c,BCX), the energies of the conformations studied are somewhat overestimated (relative to X4(c,aq)) with the MM force-field as compared to the B3LYP/6-31G(d) method.

However, the conclusions that (i) on binding to enzyme the energies of the reactive conformations (X4(sb,CEX) and X4(sb,BCX)) are increased by significantly less than those of the all-chair conformations (X4(c,CEX) and X4(c,BCX)), and that (ii) the energy increase is larger for binding to BCX than CEX, can be drawn with both energy calculation methods.

These conclusions are also in line with those drawn from the free energies obtained from the MM-PBSA analysis (see below, Fig. 5).

The internal energies of xylotetraoses were studied in more detail by calculating MM-PBSA free energies for free and enzyme-bound substrates over the 4.0 ns MD simulations.

The free energies of enzyme-bound X4s as a function of simulation time are shown in Figs. 4c and 4d, and the relative free energies are presented in Fig. 5.

The average MM-PBSA free energy of X4(c) free in aqueous solution was 640.6 kJ mol−1, calculated from the 4.0 ns simulation.

In the X4(sb) simulation the −1 sugar stayed in the 2SO conformation for the first 1.8 ns after which its conformation changed to the 4C1 conformation.

The average free energy of the X4(sb) was 658.1 kJ mol−1, calculated from the structures of the first 1.8 ns.

Thus, in water X4(c) is calculated to be 17.6 kJ mol−1 more stable than X4(sb).

In the CEX(c) simulation, the energy of the enzyme-bound X4(c) is 675.3 kJ mol−1 during the first 1.0 ns, it decreases by 5–11 kJ mol−1 for the next 2.0 ns, and increases to 682.0 kJ mol−1 for the last 1.0 ns (Fig. 4c).

The changes in the internal energy of X4(c) can be seen to be coupled with the changes in the single-trajectory free energy of binding (ΔGbind, Fig. 4a): the relaxation of the substrate structure to a less distorted one compensates partly for the formation of less tight protein–substrate interactions.

In the CEX(b) simulation the energy of the bound X4(b) was 666.1 to 668.2 kJ mol−1 during the whole simulation, on average 6 kJ mol−1 lower than the energy of X4(c) in the CEX(c) simulation (Fig. 5).

Interestingly, the order of stability of the enzyme-bound X4(c) and X4(b) is reversed from that in aqueous solution.

This is due to the clearly larger deformation energy (32 kJ mol−1vs. 9 kJ mol−1) for the binding of X4(c) than X4(b) to CEX (the energy of the unconstrained 2SO simulation was used here as a reference).

Even more pronounced differences in the deformation energies of X4 conformations take place upon binding to BCX.

During the first 1.0 ns of BCX(c) simulation the energy of X4(c) was on average 683.7 kJ mol−1, 43 kJ mol−1 higher than the energy of the free X4(c) (Fig. 4d).

After 1.8 ns of BCX(c) simulation the conformation of the −1 sugar changed from 4C1 to 1C4 and, as in the CEX(c) simulation, the change in the internal energy of X4 is linked to the changes in the enzyme–substrate interactions (Fig. 4b and d).

In the BCX(sb) simulation the average energy of X4(sb) was 664.4 to 669.9 kJ mol−1 during the simulation, on average 15.6 kJ mol−1 lower than the energy of the enzyme-bound all-chair conformation.

The energy of the enzyme-bound X4(sb) is only 9 kJ mol−1 higher than that of the free X4(sb).

Implications to enzyme-catalysed reactions

Simply stated, enzymes catalyse chemical reactions by decreasing the free energy of activation for reaching the transition state from the ground state.

A wide range of different theories have been developed to provide more detailed explanations for the large rate enhancements of enzyme-catalysed reactions as compared to the corresponding solution reactions.40–42

In general enzymes achieve rate enhancement by providing environments, which specifically stabilize transition states and high-energy intermediates of the reaction using mainly electrostatic interactions of the preorganised environment.11,42–45

In addition, part of the enzymes’ catalytic potency has been suggested to originate from their ability to bind substrate(s) in a reactive conformation,41,44–46 also named as a near-attack conformation (NAC),47 although there has been some disagreement about the importance of NACs in enzyme-catalysed reactions.48

A NAC is an enzyme–substrate conformation closely resembling the reaction transition state through which the reaction must pass to reach the transition state.47

In this conformation, which is not necessarily a stationary point on the (free) energy surface, the catalytic groups are in a precisely correct position for the reaction.

It is likely that enzymes owe their catalytic properties to several different factors, which have varying relative importance in different enzymatic reactions, especially because the nature of chemical transformations catalysed by enzymes varies greatly.

It is likely that significant part of the catalytic properties of CEX and BCX, in analogy to the lysozyme-catalysed hydrolysis studied computationally by Warshel et al.,11,12 is due to the fact that they provide suitable catalytic groups, a nucleophile and an acid/base catalyst, for the reaction and a preorganised environment, which electrostatically stabilises the reaction transition state relative to the ground state.

However, the role of substrate distortion in the catalytic mechanisms of glycoside hydrolases has been of some controversy12,49,50 ever since the publication of lysozyme–substrate X-ray structure, showing that in the Michaelis complex the reactive sugar moiety adopts a sofa conformation.49

Such experimental structural data is not available for xylanase–substrate complexes.

The calculations of this work showed that in the reactive enzyme–xylotetraose conformations of CEX and BCX the −1 sugar adopts a B3,O and 2SO conformation, respectively (Fig. 3).

These reactive conformations were stable during the 4.0 ns MD simulations.

On the basis of binding free energies, CEX binds X4(b) with a comparable affinity to X4(c), and BCX binds X4(sb) tighter than X4(c).

In addition that CEX and BCX form tighter specific enzyme–substrate interactions with the reactive than non-reactive conformations, recognition of the reactive conformations is accomplished by shape complementarity.

Namely, the reactive conformations fit in the active sites of CEX and BCX with only a modest increase in their internal energies (9 kJ mol−1) as compared to the corresponding conformations free in aqueous solution.

In contrast to this, the internal energies of all-chair conformations are increased by 33–43 kJ mol−1 upon binding to the enzymes.

As a result of this, the internal energies of the reactive xylotetraose conformations of CEX and BCX are lower than the internal energies of the bound all-chair conformations.

In aqueous solution the all-chair conformation was calculated to be 17.6 kJ mol−1 more stable than the reactive conformations (Fig. 5).

So, selective binding of reactive high-energy conformations, through which the CEX- and BCX-catalysed reactions pass, likely provides a contribution to the rate enhancement of the enzyme reaction as compared to the reaction in aqueous solution.

Namely, in solution the all-chair xylotetraose conformation was calculated to be about 15 kJ mol−1 more stable than the 2SO conformation (Fig. 5).

If in the xylanase-xylotetraose Michaelis complex xylotetraose is bound in a skew-boat or boat conformation, it can be estimated that the binding of this high-energy conformation may contribute up to 15 kJ mol−1 to the rate enhancement of the enzyme-catalysed reaction as compared to the same reaction in solution.

An accurate estimation of this contribution would require the calculation of the reaction both in aqueous solution and within enzyme, which is out of scope of this work.

It must also be noted that enzyme deformation energy (i.e. change in enzyme energy taking place upon substrate binding) is not included here.

This energy contribution may play some role in determining the relative energies of substrate conformations and different states of the enzyme-catalysed reaction.44

Unfortunately, its accurate calculation with the MM-PBSA approach, and also in general, is a demanding task.36,37

It has been suggested, on the basis of X-ray structures, that glycoside hydrolases distort their substrates by favourable enzyme–substrate interactions, which are only present when the substrate adopts a distorted conformation.49–51

In such a case the increase in the substrate energy is compensated by the enzyme–substrate interactions.

That this suggestion holds also in the case of CEX and BCX is supported by the calculated structures (Fig. 3) and more favourable specific enzyme–substrate interactions (ΔEele, Table 1) for reactive than non-reactive conformations.

It has also been suggested that substrate distortion is not important because enzymes and substrates are so flexible that no significant distortion is possible in enzyme–substrate complexes and, therefore, distortion of substrate does not play important role in enzyme catalysis.11,12,48

This idea is based on the argument, that because energetically significant substrate distortion involves small movements in the positions of atoms, it is not likely that protein structure could exert large enough forces to distort substrates.48,50

That CEX and BCX do not exert large forces on individual substrate monomers is likely true.

However, because the substrate-binding site with several subsites is complementary in shape with the relatively rigid carbohydrate chain having the reactive −1 sugar in a high-energy conformation, the all-chair substrate conformation does not fit well in the binding site resulting in considerable deformation energy (33–43 kJ mol−1), if such a conformation is tightly bound.

Since the overall shapes of saccharide polymers are sensitive to the conformations of individual subunits, in this case of the reactive −1 sugar, the recognition of substrates in reactive conformations by shape complementarity may be operating also in the mechanisms of other glycosyl hydrolases.

Distortion of oligosaccharide substrate, which seems to be more pronounced for the family 11 xylanase BCX than the family 10 CEX, requires that substrate occupies several sugar-binding subsites in order to become strong enough to enhance substrate hydrolysis rate.

If substrate distortion plays role in the catalysis and is larger for BCX than CEX, should be seen in different substrate specificities for the enzymes.

This suggestion is supported by the fact that xylanases of family 10 hydrolyse more readily and show higher affinity for shorter xylooligosaccharides than family 11 xylanases.23,52

To conclude, the results of this work led us to suggest that selective binding of reactive high-energy substrate conformations by favourable specific enzyme–substrate interactions and by shape complementarity is an important contribution, especially in the case of BCX, to the catalytic properties of xylanases.