Methylumbelliferyl-β–cellobioside (MUF–G2) is a convenient fluorogenic substrate for certain β–glycoside hydrolases (GH). However, hydrolysis of the aglycone is poor with GH family 6 enzymes (GH6), despite strong binding. Prediction of the orientation of the aglycone of MUF–G2 in the +1 subsite of Hypocrea jecorina Cel6A by automated docking suggested umbelliferyl modifications at C4 and C6 for improved recognition. Four modified umbelliferyl-β–cellobiosides [6–chloro-4–methyl- (ClMUF); 6–chloro-4-trifluoromethyl- (ClF3MUF); 4–phenyl- (PhUF); 6–chloro-4–phenyl- (ClPhUF)] were synthesized and tested with GH6, GH7, GH9, GH5 and GH45 cellulases. Indeed the rate of aglycone release by H. jecorina Cel6A was 10–150 times higher than with MUF–G2, although it was still three orders of magnitude lower than with H. jecorina Cel7B. The 4–phenyl substitution drastically reduced the fluorescence intensity of the free aglycone, while ClMUF–G2 could be used for determination of kcat and KM for H. jecorina Cel6A and Thermobifida fusca Cel6A. Crystal structures of H. jecorina Cel6A D221A mutant soaked with the MUF-, ClMUF- and ClPhUF-β–cellobioside substrates show that the modifications turned the umbelliferyl group ‘upside down’, with the glycosidic bond better positioned for protonation than with MUF–G2.
Structured digital abstract
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The filamentous ascomycete fungus Hypocrea jecorina (previously known as Trichoderma reesei) has served as a model organism for research on the mechanism, specificity and synergistic cooperation of cellulose-degrading enzymes. It is also the predominant source of cellulase enzymes for industrial use . The genome of H. jecorina has been sequenced and transcriptome analysis identified &sim 35 genes coding for known or putative biomass-active proteins that were upregulated on cellulose-containing media . Major components are two cellobiohydrolases (CBH), Cel7A (40–50%) and Cel6A (∼20%), characterized by having their active sites enclosed in a cello-oligosaccharide binding tunnel [3, 4], and the ability to processively cleave off several cellobiose units from the end of a cellulose chain before substrate release [5-7]. The enzymes cooperate synergistically and act preferentially from opposite ends of cellulose chains, Cel7A from the reducing end and Cel6A from the nonreducing end. They also utilize different reaction mechanisms. Cel7A hydrolyses glycosidic bonds with net retention of the β-anomeric configuration, whereas Cel6A is an inverting enzyme .
Aryl saccharides have been used as chromogenic substrates in cellulase research for over 50 years [9-11]. Chromogenic and fluorogenic cellobiosides and lactosides have been very helpful tools in research on H. jecorina Cel7A and the homologous endoglucanase Cel7B, as well as on GH7 enzymes from other organisms, ranging from simple applications such as expression assays and monitoring of protein purification, to high-throughput screening, pH- and temperature-dependence profiling, enzyme kinetics and inhibition studies [12-20]. However, similar applications have not been readily available for H. jecorina Cel6A and other GH6 enzymes, because of poor hydrolysis of the heteroglycosidic bond and/or preferential cleavage at homoglycosidic bonds in the substrate.
During the 1980s, Claeyssens and co-workers synthesized 4–methylumbelliferyl (MUF) cello-oligosaccharides to use as analytical tools for cellulase enzymes . Although the compounds did not work as fluorogenic substrates with H. jecorina Cel6A, the MUF group served well as a reducing end label in studies of bond cleavage preference within the saccharide part. Furthermore, because release of the fluorophore (4–methylumbelliferone; 1) was negligible, despite strong binding at the active site, they could be used as probes in direct and displacement fluorescence quenching titrations for ligand-binding studies [22, 23].
GH6 cellulases are common in both cellulolytic fungi and bacteria (http://www.cazy.org/) and have been divided into tunnel-type cellobiohydrolases (CBH) and endoglucanases (EG) that have their active sites in a more open cleft and show a higher tendency to cut internal bonds in the cellulose polymer. X–Ray structures are available of both types, five CBHs: H. jecorina Cel6A (PDB code 1CBH) , Humicola insolens Cel6A (PDB code 1BVW) , Coprinopsis cinerea Cel6A (PDB code 3VOH) , C. cinerea Cel6C (PDB code 3A64)  and Chaetomium thermophilum Cel6A (PDB code 4A05) , and three EGs: H. insolens Cel6B (PDB code 1DYS) , Thermobifida fusca Cel6A (PDB code 1TML)  and Mycobacterium tuberculosum Cel6A (PDB code 1UP0) .
Enzyme mechanism of family 6 glycoside hydrolases
Structure studies of ligand binding in GH6 CBHs have provided detailed insights into the enzyme mechanism [25, 30-32]. Six d–glucose units of the nonreducing end of a cellulose chain bind in subsites −2 to +4 with the cellobioside to be cleaved off in subsites −2 to −1. Tryptophan residues provide glucosyl-binding platforms in subsites −2, +1, +2 and +4. An aspartic acid residue, D221 in H. jecorina Cel6A, acts as the catalytic acid and protonates the scissile glycosidic bond between subsites −1 and +1. A water molecule is held in position for nucleophilic attack on the anomeric carbon of the d-glucose unit in subsite −1, but is not in direct contact with any carboxylate side chain. Instead it is hydrogen bonded to a second water molecule, which in turn binds to the side chain of an aspartate residue, D175 in H. jecorina Cel6A, that is believed to act indirectly as the catalytic base, in a so-called Grotthus mechanism. This has also been supported by molecular dynamics simulations . Dual conformations have been observed for both catalytic residues. In several structures, they form a hydrogen bond to each other. This hydrogen bond must be broken and both side chains must be in their ‘active’ conformations to catalyse the reaction. Furthermore, one loop near the catalytic centre is flexible and needs to swing in to a closed conformation to isolate the two water molecules near the anomeric carbon [23, 31].
Mechanistic itinerary of GH6
A glycosidic bond substitution appears to subject the involved glycon ring to a conformational change [33-35]. Within the crystal structures of the GH6 CBHs (Cel6A) from H. jecorina (PDB codes 1QK2 and 1QJW) , and H. insolens (PDB codes 1GZ1 and 1OC5)  that are in Michaelis complex with a nonhydrolysable thio-oligosaccharide, the 2SO conformation is observed for the sugar moiety occupying the −1 subsite. Also, in the 3D structures of H. insolens Cel6A and M. tuberculosis Cel6 in complex with a gluco-isofagomine cellobioside analogue (resp. PDB codes 1OCN  and 1UP2 ), the isofagomine moiety occupying the −1 subsite shows a conformation that is close to a 2S5 (equivalent to a 2SO in the pyranoside numbering). The apparent preference for a β-2SO sugar conformation has been interpreted as being indicative of a synperiplanar lone pair hypothesis compliant [β-2SO pre-TS → 2,5B TS → nonground state α-5S1] hydrolysis mechanism itinerary [34, 36], but may just as well be indicative of a standard antiperiplanar lone pair hypothesis compliant [β-2SO or nearby β-1S3 pre-TS → E3 or nearby 4H3 TS → ground state α-4C1] itinerary [33, 34].
Anti versus syn protonation and the exo-anomeric effect
The concept of semi-lateral glycosidic oxygen protonation was introduced in a seminal article by Heightman and Vasella . It was originally only described for beta-equatorial glycoside hydrolases, but appears to be equally applicable to enzymes acting on an alpha-axial glycosidic bond . When dividing subsite −1 by the Ox, C1′ and H1′ plane of the −1 glycoside, many ligand-complexed structures show that the proton donor is positioned either in the syn half-space, near the ring-oxygen of the −1 glycoside, or in the anti half-space at the opposite side of the ring-oxygen. Members of the same GH family are always syn or always anti protonators and this specificity appears to be preserved within clans of families.
Closer inspection of −1/+1 subsite-spanning ligands in crystal structures reveals a further intriguing corollary . In anti-protonating GH enzymes, the scissile anomeric bond (often as thio-analogue) always shows a dihedral angle φ (O5′–C1′–[O,S]x–Cx) that is in the lowest-energy synclinal (gauche) position. A minus-synclinal dihedral angle φ for an equatorial glycosidic bond, or plus synclinal for an axial bond , allows for hyperconjugative overlap of the C1′–O5′ antibonding orbital with an antiperiplanar-oriented lone pair orbital lobe of the glycosidic oxygen, thereby creating partial double-bond character and stabilization of the glycosidic bond by 4–5 kcal·mol−1; the so-called ‘exo-anomeric effect’ [40, 41]. However, anti protonation occurs precisely on the oxygen's antiperiplanar lone pair, which removes the stabilizing effect, thus providing an additional enzyme-strategy for lowering the activation barrier (Fig. 1 centre).
Syn protonators show a different strategy . In many ligand complexes the dihedral angle φ of the scissile bond has been substantially twisted away from its lowest-energy synclinal position: clockwise to minus-anticlinal or antiperiplanar for beta-equatorial; counterclockwise to plus-anticlinal or antiperiplanar for alpha-axial anomeric bonds. This removes the hyperconjugative overlap and thus the stabilizing exo-anomeric effect. Because of this twisting, the lone pair of the glycosidic oxygen is pointing into the syn half-space, ready to be protonated by the syn-positioned proton donor (Fig. 1 right). GH6 enzymes are syn protonators and in several subsite −1/+1 spanning ligand complexes the dihedral angle of the (thio)glycosidic bond is in minus-anticlinal position (e.g. φ2 = −125° in H. jecorina PDB-entry 1QK2, φ2 = −132° in H. insolens 1GZ1 or φ2 = −140° in T. fusca 2BOF).
Consequences for GH6 substrates
Summarizing from the considerations above, with the syn-protonating GH6 enzymes the productive pose of a substrate is expected to possess a ring conformation close to a 2SO skew for the β-d-glucoside moiety in subsite −1, whereas the dihedral angle φ of the scissile anomeric bond is in the minus-anticlinal position. This is a double deformation away from the ground state, and the combined energetic burden (4–5 kcal·mol−1 for the exoanomeric effect [40, 41] plus 5–6 kcal·mol−1 for the 4C1 into 2SO glucoside deformation ) will have to be compensated for by favourable enzyme–substrate interactions. We suspected that 4–methylumbelliferyl-β–cellobioside (MUF–G2, 2) fails to be a good substrate for GH6 enzymes because the 4–methylumbelliferyl moiety cannot make enough favourable interactions with the +1 subsite to help the substrate to be induced into the productive pose, after all, this subsite has been evolved to specifically recognize a d-glucoside, not a 4–methylumbelliferyl group. By contrast, the also syn-protonating GH7 enzymes do recognize MUF–G2 (2) as a good substrate [21, 22] probably from a lucky subsite +1 match.
We thus decided to use automated docking with MUF–G2 (2), with its central d-glucoside in the 2SO skew conformation, into the active site of H. jecorina Cel6A in order to obtain information on how its umbelliferyl group would be oriented versus the +1 subsite and how the aglycone may be modified towards better positioning and more efficient hydrolysis. Four modified umbelliferyl β–cellobiosides were designed, synthesized and tested as substrates for GH6 and other enzymes, and enzyme-substrate complex structures were determined by X-ray crystallography.
Results and Discussion
Predictions from automated docking and rationale for substrate development
Docking with AutoDock-Vina  of MUF–G2 (2, central d-glucoside in 2SO) into H. jecorina Cel6A structure 1QJW did yield the productive pose occupying subsites −2 to +1 with a minus-anticlinal anomeric bond: φ2 (O5′–C1′–O7umb–C7umb) = −114°, albeit only as the 11th found pose with a calculated affinity of −8.6 kcal·mol−1 (Table 1). The latter must be an overestimation because the docking program does not take anomeric effects into account. More importantly, the orientation of the 4–methylumbelliferyl group within subsite +1 (reflected by ψ2 (C1′–O7umb–C7umb–C6umb) = −108°) suggests two positions for substitution towards possibly better recognized variants that can be easily synthesized: (a) the 4–methyl group of the umbelliferone points towards the +2 subsite which can be filled by introduction of a 4–phenyl group; and (b) the C6–H6 bond of the umbelliferone points towards an empty local zone in subsite +1, which in the original (Glc)2–S–(Glc)2–OMe complex is occupied by the hydroxymethyl group of the resident d–glucoside. Introduction of a chlorine at C6 would occupy this zone. This is an electron-withdrawing group that will increase the leaving capacity of the aglycone, and because it is close to the anomeric oxygen it may also decrease the exo-anomeric effect. Concomitant incorporation of a 4–trifluoromethyl group can further increase the leaving capacity of the aglycone.
|Docked ligand||Pose nrb||Dihedral ϕ2c||Dihedral ψ2c||Calculated affinityd|
|Crystal structure||+1 group||ϕ2c||ψ2c||Distance (Å)e|
Vina dockings with ClMUF–G2 (7), ClF3MUF–G2 (8), PhUF–G2 (9) and ClPhUF–G2 (10) indeed indicate a better recognition of the productive pose, each predicting a near-identical orientation of the umbelliferone moiety in subsite +1 (reflected by similar φ2 and ψ2 values; Table 1); this prompted us to carry out their synthesis. Our H. jecorina Cel6A crystal structure in complex with ClPhUF–G2 (10) revealed a near-identical pose as that predicted by the Vina-docking (Fig. S2).
The four modified umbelliferyl β–cellobiosides were synthesized in a short-step straightforward route with good overall yields (Fig. 2). The 6–chloro-4–methyl-, 6–chloro-4-trifluoromethyl-, 4–phenyl- and 6–chloro-4–phenyl-umbelliferones (respectively 3, 4, 5 and 6) were synthesized by Pechmann condensation in trifluoroacetic acid ; an alternative preparation of 6–chloro-4–methylumbelliferone (3) by condensation in sulfuric acid has been described previously, but without analytical data . Aglycone coupling with acetobromo cellobiose (11) to the respective peracetates 12, 13, 14 and 15 was Knorr-type in the presence of silver carbonate and 2,4,6–collidine , which was followed by deacetylation under mild acidic conditions yielding the respective β–cellobiosides 7, 8, 9 and 10.
Fluorescence properties of synthesized compounds
The fluorescence spectra of ClMUF (3), PhUF (5) and ClPhUF (6) showed excitation (Ex) and emission (Em) maxima at about the same wavelengths as MUF (1, ∼360 and ∼450 nm), whereas the trifluoro-methyl group of ClF3MUF (4) caused a shift to longer wavelengths (385 and 495 nm; Table 2). All fluorophores showed higher fluorescence intensity at pH 10.2 than at pH 5.0. ClMUF (3) and ClF3MUF (4) gave approximately three and four times higher intensity at pH 5.0, and somewhat lower intensity at pH 10.2 (∼70% and ∼20%, respectively) than MUF (1), whereas the phenyl substituted ClPhUF (6) and especially PhUF (5) gave weaker fluorescence at pH 5.0, and even more so at pH 10.2, than MUF (1). Emission spectra demonstrate that the background fluorescence of the substrate ClMUF–G2 (7) is very low, even at 0.4 mm concentration and pH 10.2, compared with 16 μm of the free fluorophore ClMUF (3) (Fig. S1).
|Fluorophore||Wavelength maximum (nm)||Fluorescence intensity (a.u.)|
|Ex||Em||pH 5.0||pH 10.2|
Monitoring hydrolysis of fluorogenic substrates on UV table
Different H. jecorina and T. fusca cellulases were assayed for activity against the fluorogenic cellobioside substrates by monitoring the fluorescence on a UV table (Fig. 3). ClF3MUF–G2 (8) was less soluble and had to be used at lower concentration (0.1 mm) than the other substrates (0.5 mm). The substrates themselves did not show any visible fluorescence as seen in the top row.
The free fluorophores ClMUF (3) and ClF3MUF (4) showed similar intensity as MUF (1) under UV light (pH 5.0), but the fluorescence was much weaker with PhUF (5) and ClPhUF (6). Both GH6 enzymes, H. jecorina Cel6A (CBH) and T. fusca Cel6A (endoglucanase) showed clear activity against ClF3MUF–G2 (8), while fluorescence was barely visible with ClMUF–G2 (7), and not at all with MUF–G2 (2). The fluorescence was weak with PhUF–G2 (9) and ClPhUF–G2 (10), but may still indicate significant activity due to the low intensity of the corresponding free fluorophores. H. jecorina Cel7A (CBH) and Cel7B (endo), and T. fusca Cel9B (endo) could hydrolyse all substrates and gave very strong fluorescence, indicating high degree of substrate consumption. No activity was detected against any substrate with T. fusca Cel9A (endo-processive).
Initial hydrolysis rates
Initial hydrolysis rates ([product]/[enzyme]·min−1) with 0.5 mm fluorogenic substrate [0.1 mm with ClF3MUF–G2 (8)] were determined for H. jecorina Cel6A and Cel7B and T. fusca Cel6A by following the increase of fluorescence at 37 °C (Table 3). The GH6 enzymes did actually hydrolyse MUF–G2 (2) at detectable, albeit very low, rates. In all cases, hydrolysis of the modified umbelliferyl substrates was faster. ClF3MUF–G2 (8) gave the highest rate increase compared with MUF–G2 (2) (150 times with H. jecorina Cel6A) followed by ClMUF–G2 (7). HPLC analysis revealed that < 1% glucose was released after > 10% consumption of ClMUF–G2 (7) or ClPhUF–G2 (10), demonstrating preferential cleavage of the heteroglycosidic bond to the aglycone. The hydrolysis rates for the GH6 enzymes were still approximately three orders of magnitude lower than with H. jecorina Cel7B. Interestingly, also with this enzyme ClMUF–G2 (7) and ClF3MUF–G2 (8) were hydrolysed much faster than MUF–G2 (2) (approximately five times). Preliminary tests indicated increased hydrolysability of the synthesized substrates also with H. jecorina Cel5A and Cel45A, but the activity was too low for accurate determination of hydrolysis rates (data not shown).
|Substrate||Hydrolysis rate (min−1)b|
|H. jecorina Cel6A||T. fusca Cel6A||H. jecorina Cel7B|
ClMUF–G2 (7) was chosen as the most suitable substrate for measurement of enzyme kinetics, and was used at substrate concentrations well above and below the KM values. By stopping the reaction and enhancing fluorescence with addition of pH 10.2 buffer, enough product was released within 20 min for determination of kcat and KM for both H. jecorina Cel6A at 37 °C and T. fusca Cel6A at 50 °C (Table 4). A lower KM value for T. fusca Cel6A indicates about twice as strong binding as H. jecorina Cel6A. The kcat values are not directly comparable because of different temperatures.
|Enzyme||kcat (min−1)||KM (μm)||kcat/KM (s−1·m−1)|
|H. jecorina Cel6Ab||0.11 ± 0.01||45 ± 7.9||41 ± 1.7|
|T. fusca Cel6Ac||0.20 ± 0.02||21 ± 5.7||162 ± 27|
Crystal structures of H. jecorina Cel6A D221A substrate complexes
X–Ray crystallographic studies were with the catalytic acid mutant D221A of H. jecorina Cel6A to avoid rapid hydrolysis of the substrates . Soaking experiments were successful with MUF–G2 (2), ClMUF–G2 (7) and ClPhUF–G2 (10); the complexed structures were refined at 1.7, 1.7 and 2.2 Å resolution, and Rfree values of 0.22, 0.24 and 0.25, respectively. Statistics from data processing and structure refinement are summarized in SI Table S1. The location of the active site tunnel and its occupation by ClMUF–G2 (7) and ClPhUF–G2 (10) are shown in Fig. 4. The protein structures are very similar (pairwise RMSD < 0.5) to previous H. jecorina Cel6A structures, including the N- and O–glycosylation sites on the surface. The flexible 177–183 loop near the active centre adopts the closed conformation, as previously observed with the D221A and Y169F mutants (PDB codes 1HGY and 1QJW), except in protein molecules A and C of the MUF–G2 complex where loop residues 179–181 had insufficient electron density to be resolved.
The MUF–G2 complexed crystal contained four protein molecules per asymmetric unit. Molecules B and D displayed practically identical binding of one MUF–G2 molecule at the catalytic centre, with the cellobioside moiety occupying subsites −2/−1 and the MUF group in subsite +1, and without further ligand electron density in subsites +2 and +3 (Fig. 5B). In the other two protein molecules A and C, the MUF moiety was found in subsite +3 and the cellobioside in +1/+2. They also had electron density for another cellobiose moiety in subsites −1 and −2, but without either additional density from MUF at subsite +1 or from an alpha-axial hydroxyl group on the d-glucoside moiety occupying subsite −1. This implies that either the −1 and +1 d-glucoses are connected, or that the electron density is the result of overlapping binding of cellobiose and MUF–G2 at partial occupancies. All four subsequent d-glucoses and MUF could be connected by glycosidic bonds without violation of bond length and angle restraints. Thus we chose to include and refine 4–methylumbelliferyl-β–cellotetraoside (MUF–G4) in protein molecules A and C in the final model (Fig. 5A).
The other two crystals had two protein molecules in the asymmetric unit. With the ClMUF–G2 complex, both A and B chains showed the substrate occupying subsites +1 to +3 (ClMUF group in +3) which is not a catalytically productive binding position (Fig. 5D). The cellobioside moieties bound similarly, but there was a slight deviation in the binding of the ClMUF unit, up to 1.6 Å for atoms at the outer edge, most likely because of interference by a neighbouring protein molecule (symmetry-related chain C) near the ClMUF in chain B, which is not present in chain A. In chain B there was also density for a single d-glucose in subsite −2, but not in chain A. In the ClPhUF–G2 complexed structure, however, the productive pose of the substrate was found in both protein molecules, with almost identical binding of the cellobioside in subsites −2/−1, the 6–chloro umbelliferyl moiety in subsite +1, and the 4–phenyl substitution occupying subsite +2, all as predicted by the docking (Fig. 5C).
All the d-glucose moieties found in subsites −2, +1 and +2 did refine in the stable 4C1 chair conformation, in similar positions and with same interactions with the protein as described in previous H. jecorina Cel6A structures  (e.g. 1QK2, 1QJW). The d-glucose moiety in subsite −1 adopted the previously observed 2SO skew conformation, somewhat tilted towards a B3,O boat, in both ClPhUF–G2 complexed protein molecules (Fig. 5D), and in the MUF–G2 complexes A and C where MUF–G4 was modelled (occupying subsites −2 to +3 with MUF in +3; Fig. 5A). The 6–hydroxyl group in −1 was not positioned above the d-glucose ring as in the wild-type H. jecorina Cel6A complex with Glc2–S–Glc2–OMe (1QK2), but had taken on a more relaxed position pointing away from the d-glucose ring, as observed with the Y169F mutant  (1QJW) However, in the other two protein molecules (B, D) of the MUF–G2 complex, with the substrate occupying subsites −2 to +1 (MUF group in +1), the d-glucoside in −1 adopted the 4C1 chair conformation (Fig. 5B).
Comparison of MUF–G2 and ClPhUF–G2 binding at the catalytic centre
A superposition of the observed binding modes of MUF–G2 (2) (B chain) and ClPhUF–G2 (10) (A chain) within the H. jecorina Cel6A crystal structures is shown in Fig. 6. The catalytic acid is missing in these structures due to the D221A mutation; instead the D221 side chain was taken from the structure of mutant D175A  (1HGW, B molecule), which is the only H. jecorina Cel6A structure where the catalytic acid is in position for protonation of the glycosidic bond .
The binding mode of MUF–G2 (2) at the catalytic centre provides clues to why it is poorly hydrolysed by the enzyme. Although it occupies the correct −2 to +1 subsites, the d-glucoside moiety at −1 is in the unproductive ground state 4C1 chair conformation. The anomeric carbon, or more precisely its anti-orbital to the glycosidic bond, is poorly accessible for nucleophilic attack. Furthermore, the dihedral angle φ2 of the heterosidic anomeric bond remains in a minus-synclinal (φ2 = −88°) position, thus the stabilizing exo-anomeric effect is still present. The heteroglycosidic oxygen is also rather distant from the catalytic acid (∼3.9 Å) and is not pointing a free electron pair towards it (Fig. 6A). The orientation of the MUF group versus subsite +1, as reflected by the dihedral angle ψ2 (C1′–O7umb–C7umb–C6umb) = +21°, comes as a surprise. It is flipped ‘upside down’ compared with the productive pose in the docking of MUF–G2 (2) at the active site (ψ2 = −108°; Table 1 and SI Fig. S2). The 4–methyl group is not pointing towards the +2 subsite, but instead interferes with and prevents closure of the active-centre loop, since it overlaps with the position of Ala178 Cβ in the closed loop conformation.
By contrast, the ClPhUF–G2 complex shows a similar orientation (ψ2 = −137°) of the umbelliferyl moiety within subsite +1 as predicted by the docking (ψ2 = −115°), with the 4–phenyl group occupying subsite +2 (Figs 6B and SI S2). This complex shows all the requirements for the substrate to be in a catalytically productive pose: (a) the d-glucoside moiety at subsite −1 has adopted the pre-TS 2SO skew conformation; (b) the dihedral angle φ2 of the heterosidic anomeric bond is now in minus-anticlinal position (φ2 = −105°) which relieves the stabilizing exo-anomeric effect; and (c) the glycosidic oxygen is closer to the catalytic acid (∼3.0 Å) and now points a free electron pair towards it.
Because of the C6 substitution, ClPhUF–G2 (10) cannot bind in the ‘upside down’ orientation observed for MUF–G2 (2) within the confined space of subsite +1, because the chlorine atom would then clash with C1 and/or O5 of the d-glucoside in subsite −1. Neither can the d-glucoside at −1 relax to a 4C1 chair without O2 of the ClPhUF moiety overlapping with Ala178 from any known conformation of this enzyme (regardless if the active-centre loop is closed or open).
Conversely, the d-glucoside in subsite −1 of the MUF–G2 complex (B and D chains) cannot adopt the 2SO skew conformation with the current ‘upside down’ orientation of MUF because that would place O2 of the MUF moiety in close contact with His266 NE2. However, as shown by the docking, there is no hindrance for MUF–G2 (2) to bind in a flipped orientation similar to ClPhUF–G2 (10) that allows the 2SO skew conformation at subsite −1. It should be noted, however, that subsite +1 is too narrow to allow an umbelliferyl group to flip once inside. Thus a MUF–G2 molecule bound in the nonproductive ‘upside down’ orientation would first need to dissociate from the active site and rebind in the productive pose in order to be hydrolysed.
Productive versus nonproductive binding positions
The observation of two binding positions for MUF–G2 (2), with the MUF moiety either in subsite +1 or +3, correlates with differences in crystal packing. In the B and D chains of the asymmetric unit, subsite +4 is blocked and +3 is partly obstructed by crystal contacts (Leu230–Pro233 of chains C and A, respectively). Apparently this has a negative impact on MUF binding at subsite +3 and shifts the binding preference of MUF–G2 (2) in favour of the −2 to +1 position. In protein molecules A and C, subsite +3 is fully accessible and harbours the MUF moiety, suggesting that MUF–G2 (2) prefers to bind nonproductively at the +1 to +3 position when the enzyme is in solution.
A similar correlation does not occur with the ClMUF–G2 and ClPhUF–G2 complexes. Although subsites +3 and +4 of chain A are obstructed in the same way, while they are open in chain B, ClMUF–G2 (7) binds at the nonproductive +1 to +3 position in both chains. By contrast, ClPhUF–G2 (10) shows the productive pose at subsites −2 to +2 in both. Apparently the 4–phenyl substitution shifts the binding preference, most likely due to its affinity at subsite +2, because there is no steric hindrance for the ClPhUF moiety to bind in subsites +3 to +4 in chain B. Obviously ClMUF–G2 (7) must also be able to bind productively at −2 to +1 and is actually hydrolysed faster than ClPhUF–G2 (10), despite the preference for nonproductive binding (Table 3).
These structures illustrate how the presence of multiple subsites may allow for alternate binding positions along the active site, which complicates enzyme kinetics and ligand-binding studies. Nonproductive substrate-binding modes that overlap and compete with productive binding will lower the apparent kcat and KM values by the same factor (1/[1 + nonproductive Kd/productive KM]) depending on the relative affinities. Weaker productive than nonproductive binding of ClMUF–G2 (2) in H. jecorina Cel6A implies that the ‘true’ intrinsic kcat and KM values must be at least twice as high as the measured apparent values (Table 4). Apparent KM values are thus not necessarily direct measures of binding strength for substrate at the catalytic centre. Differences observed between enzymes and/or substrates may as well be due to differences in nonproductive binding at other positions along the active site.
Comparison of fluorogenic substrate and oligosaccharide binding at the catalytic centre
The configuration at the −1/+1 glycosidic bond in these and comparable crystal structures with GH6 enzymes is listed in Table 1. We note that with d-glucose or thio-d-glucose at subsite +1, the φ2 angle is rotated around −20° further away from the minus-synclinal position compared with the ClPhUF–G2 complex. Perhaps this is necessary for complete removal of exoanomeric stabilization and/or optimal alignment of a free electron pair with the catalytic acid, which would then explain why these fluorogenic substrates are still hydrolysed slowly compared with oligosaccharide substrates. We propose that further development of artificial substrates should strive at optimizing the fit of the aglycone at subsite +1 such that the φ2 angle will be restrained within the range observed when d-glucoside is attached at subsite +1 (Table 1).
This work demonstrates the successful use of substrate docking at the active site in combination with theory on mechanistic requirements on bond geometry for catalysis, as guides for rational modification of the substrate towards better complementarity with the catalytic machinery of the enzyme. Our results indicate that poor hydrolysis of MUF–G2 (2) by GH6 enzymes is due to preferential nonproductive binding at the active site. At least with H. jecorina Cel6A, MUF–G2 (2) prefers to bind out-of-register with the catalytic centre (in subsites +1 to +3), and when bound in-register at the catalytic centre a nonproductive pose is preferentially adopted. Modifications of the umbelliferyl motif by introduction of a chlorine at C6 and/or substitution of the 4–methyl group with 4–phenyl or 4-trifluoromethyl, as predicted by substrate dockings, restrict the conformational freedom in the confined space of the active site and effectively turn the umbelliferyl group ‘upside down’, with the glycosidic bond better positioned for protonation (Fig. 6).
The modified umbelliferyl β–cellobioside substrates are indeed hydrolysed faster than MUF–G2 (Table 3). The largest rate increase with GH6 enzymes, ∼150 times, is obtained with ClF3MUF–G2 (8). But this compound is only soluble up to ∼0.1 mm concentration. Furthermore, the 4–phenyl substituted fluorophores PhUF (5) and ClPhUF (6) show much weaker fluorescence. Therefore, ClMUF–G2 (7) was considered most useful for enzyme kinetics measurements. All the fluorophores exhibit stronger fluorescence at pH 10.2 than at pH 5.0 [∼12–fold for ClMUF (3); Table 2], and thus the sensitivity of activity assays can be significantly enhanced by alkali addition prior to fluorescence measurements.
The rate of fluorophore release from the modified substrates is still two to three orders of magnitude lower with the tested GH6 enzymes than with H. jecorina Cel7B, pointing at a potential for additional improvements. Based on the comparison with relevant GH6 structures with d–glucoside or thio-d–glucoside connected at subsite +1, we speculate that further rate enhancement may be achieved if the φ2 angle of the scissile bond is rotated further away from the minus-synclinal position than observed in the ClPhUF–G2 complex. However, further studies are needed to investigate this.
Materials and methods
Protein molecule A of the H. jecorina Cel6A crystal structure 1QJW, containing methyl 4-S-β–cellobiosyl-4-thio-β–cellobioside (Glc)2–S–(Glc)2–OMe within its −2 to +2 subsites, was selected for docking experiments with AutoDock-Vina  using AutoDockTools  for preparations of enzyme and ligands. The calcium, water and noncovalent ligands were removed, and the Y169F mutation was renatived. The original ligand was successfully redocked as the first pose found by Vina (calculated affinity −11.3 kcal·mol−1) when its four anomeric bonds were set as unrotatable in their original dihedral angles, resp. φ1 (O5″–C1″–O4′–C4′) = −72°, φ2 (O5′–C1′–S4–C4) = −140°, φ3 (O5–C1–O4*–C4*) = −75° and φ4 (O5*–C1*–O4Me–CMe) = −69°. The umbelliferyl–cellobioside ligands, with the central d-glucoside in a 2SO skew conformation, were drawn in 3D with avogadro (http://avogadro.openmolecules.net/) and were minimized with the MMFF94 force field . Dockings were with the anomeric bond between both d-glucoside moieties always locked in the synclinal dihedral angle as obtained after MMFF94 minimization (φ1 (O5″–C1″–O4′–C4′) = −71°), whereas the heterosidic anomeric bond was kept as rotatable. Vina box settings were: size x, y and z = 20 Å centred at x = 28.5, y = −36.3 and z = 35.9; docking exhaustiveness was set at 64. Only those poses were evaluated where the cellobioside moiety occupied subsites −2 to −1 near identical to that in the original ligand.
MUF–G2 (2) was synthesized as described previously . Details of synthesis, analysis and compound data for the four modified umbelliferones, respectively 6–chloro-4–methyl- (ClMUF, 3), 6–chloro-4-trifluoromethyl- (ClF3MUF, 4), 4–phenyl- (PhUF, 5) and 6–chloro-4–phenyl- (ClPhUF, 6) and for the corresponding β–cellobiosides ClMUF–G2 (7), ClF3MUF–G2 (8), PhUF–G2 (9) and ClPhUF–G2 (10) are provided in the Supporting Information (SI).
Enzymes and general experimental conditions
Wild-type enzymes from H. jecorina for hydrolysis experiments, prepared as described from culture filtrate of strain QM9414 (ATCC 26921; formerly T. reesei) were Cel5A and Cel7B , Cel6A , Cel7A  and Cel45A . Enzymes Cel6A, Cel9A and Cel9B (formerly E2, E4, E1) from T. fusca  were kind gifts from David B. Wilson (Cornell University, Ithaca, NY, USA). Construction and expression of the catalytically deficient mutant D221A of H. jecorina Cel6A and preparation of the catalytic domain has been described previously . Unless otherwise stated, hydrolysis experiments were carried out in 240 μL of 25 mm sodium acetate buffer, pH 5.0 or 5.5, in black 96-well microtiter plates with cover (NUNC, Thermo Fisher Scientific, Roskilde, Denmark), and fluorescence was measured at these conditions or after addition of 60 μL 1 m sodium glycine buffer, pH 10.2. MUF was from Sigma-Aldrich (St. Louis, MO, USA). All other chemicals were of analytical grade.
Spectroscopic property analysis
Fluorescence spectra of 20 μm free fluorophores MUF (1), ClMUF (3), ClF3MUF (4), PhUF (5), ClPhUF (6) and 0.5 mm ClMUF–G2 (7) were scanned from 300 to 550 nm wavelength in a plate reader (Cary Eclipse, Varian, Agilent Technologies, Santa Clara, CA, USA) to determine their respective Ex (excitation) and Em (emission) wavelength maxima, where relative fluorescence intensities were measured. This was done at room temperature, both at pH 5.0 and at pH 10.2 after addition of sodium–glycine buffer.
Monitoring hydrolysis of fluorogenic substrates on a UV table
Hydrolysis reactions were set up in 96-well plates at pH 5.0 or 5.5 and 40 °C, with 0.1 mm ClF3MUF–G2 (8) due to limited solubility, and 0.5 mm of the other fluorogenic cellobioside substrates, alone or with the following enzymes and concentrations: H. jecorina Cel6A, 2 μm; Cel7A, 1 μm; Cel7B, 0.1 μm (pH 5.0); T. fusca Cel6A, 2 μm; Cel9A, 1 μm; Cel9B, 0.1 μm (pH 5.5). The fluorescence was monitored on a UV table (UV Transilluminator Biometra TI 1, Biometra GmbH, Goettingen, Germany) and photographed with a Kodak EDAS 290 camera (Thermo Fisher Scientific). The picture in Fig. 3 was taken after 1 h of incubation. Free fluorophore was included as reference at 10 and 20 μm concentration for MUF (1), ClMUF (3) and ClF3MUF (4) or at 10, 20 100 and 500 μm concentration for PhUF (5) and ClPhUF (6), at pH 5.0.
Determination of hydrolysis rates and enzyme kinetics parameters
Hydrolysis rates were measured at pH 5.0 (pH 5.5 for T. fusca Cel6A) and 37 °C, with 0.1 mm ClF3MUF–G2 (8) and 0.5 mm of the other cellobioside substrates, and with H. jecorina Cel6A (1 μm), Cel7B (0.01 μm), Cel5A (1 μm), Cel45A (2 μm) or T. fusca Cel6A (1 μm). The fluorescence was recorded at regular intervals for 10–20 min in a microplate reader (Infinite M200, TECAN, with magellan 6.55 software) at their respective Ex and Em wavelengths. The hydrolysis rates ([P]/t/[E]) were derived from the slopes using standard curves for the respective free fluorophores.
Enzyme kinetics parameters (kcat, KM and kcat/KM) of H. jecorina Cel6A (1 μm; pH 5.0) and T. fusca Cel6A (1 μm; pH 5.5) were determined using ClMUF–G2 (7) as substrate, at concentrations ranging from < 1/5 KM to > 5 KM. After 20 min at 37 °C for H. jecorina Cel6A and 50 °C for T. fusca Cel6A the reaction was stopped by addition of 60 μL 1 m sodium glycine buffer, pH 10.2, and fluorescence intensity was read (Ex 360 nm; Em 450 nm; Cary Eclipse). The substrate consumption was < 10% in all cases. After subtraction of background fluorescence, kcat and KM were extracted by fitting the Michaelis–Menten equation to the experimental data by nonlinear regression using mmfit of the simfit package (http://www.simfit.manchester.ac.uk/) . The efficiency constant, kcat/KM was calculated using rffit of the simfit package by fitting the equation below.
HPLC analysis of hydrolysis products
Hydrolysis reactions were set up in 200 μL buffer, pH 5.0, at 37 °C, with 50 μm substrate and 1 μm enzyme (H. jecorina Cel6A or T. fusca Cel6A). Reactions with ClMUF–G2 (7) were incubated for 2 and 4 h, and with ClPhUF–G2 (10) for 16 h, aiming for > 10% substrate consumption. Glucose and cellobiose in the hydrolysates were separated and quantified by high-performance anion exchange chromatography coupled with pulsed amperometric detection (HPAEC-PAD) on a Dionex ICS-3000 HPLC system equipped with a ED40 electrode, using a PA-200 column (3 × 250 mm; Dionex, Thermo Fischer Scientific), 4 min equilibration with 50 mm NaOH, 20 mm sodium acetate, 5 min isocratic elution and 5 min gradient elution to 200 mm NaOH, 20 mm sodium acetate at 0.4 mL·min−1 flow rate. The ED40 detector was set to use the carbohydrate quadruple waveform as provided by the manufacturer. Free fluorophores ClMUF (3) and ClPhUF (6) were analysed for reference, and standards of glucose and cellobiose were used for quantification (0.5, 1, 2.5, 5, 10, 20 μm).
The Cel6A D221A catalytic domain (10 mg·mL−1) was crystallized by the hanging drop vapour diffusion method  with 20% poly(ethylene glycol) 5000 monomethyl ether (Fluka, Sigma-Aldrich) and 20 mm sodium Mes buffer, pH 6.0, as precipitant solution. Streak and micro-seeding were used to improve crystal quality. Ligands were introduced by transferring crystals to soaking drops containing 1 mm of MUF–G2 (2), ClMUF–G2 (7) or ClPhUF–G2 (10) in precipitant solution with 15% glycerol as a cryoprotectant. After 1.5 h of soaking at room temperature, crystals were flash frozen in liquid nitrogen. X–Ray diffraction data were collected from single crystals at the European Synchrotron Radiation Facility (ESRF; Grenoble, France) beamlines ID14–1 and ID23–1. The datasets were reduced and scaled using the ccp4 suite of programs . The complex structures were solved by molecular replacement using molrep (auto-MR), a program in the ccp4 package, and using the previous structure of H. jecorina Cel6A D221A  (PDB code 1HGY) as search model. A topology file for refinement of substrate structure was obtained by the PRODRG server (http://davapc1.bioch.dundee.ac.uk/prodrg/) . Water molecules were initially added automatically using the water picking function in ARP/wARP . The structures were improved through several cycles of inspection and rebuilding against electron density using coot  and structure refinement with refmac5 . For cross-validation, 5% of the reflections were excluded from refinement for calculation of Rfree. Uppsala Software Factory (http://xray.bmc.uu.se/usf/) programs moleman2  and lsqman  were used for pdb-file handling and structure comparison. Protein structure images were prepared using pymol (PyMOL Molecular Graphics System, Version 184.108.40.206 Schrödinger, LLC). Atom coordinates and structure factors have been deposited with the Protein Data Bank  with accession codes 4AX7 (MUF–G2 (2) soak), 4AU0 (ClMUF–G2 (7) soak) and 4AX6 (ClPhUF–G2 (10) soak).
This work was supported by the Faculty for Natural Resources and Agriculture, Swedish University of Agricultural Sciences; Swedish Farmers Research Foundation (SLF); Swedish Energy Agency; through the research program MicroDrivE – Microbially Derived Energy, and by Japan Society for the Promotion of Science (JSPS) through fellowship to TI. Dr Anu Koivula and Dr Laura Ruohonen, VTT Technical Research Centre of Finland, are gratefully acknowledged for providing the H. jecorina Cel6A D221A mutant protein, and Prof. David B. Wilson, Cornell University, Ithaca, NY, USA, for kind gift of T. fusca enzymes.