Glucose oxidase catalyses the oxidation of ß-D-glucose in a highly specific way. The enzyme has found a variety of applications in food technology and as a biosensor, but the mechanism of catalysis is not known. Recently the X-structure of this flavoenzyme has been solved. We carried out density functional studies to determine properties of the coenzyme. Docking simulations with correlation techniques have been used to generate an initial structure of the enzyme substrate complex, which has been refined subsequently with MD-simulations. Finally, a small of the active site model has been constructed to derive information about the oxidative and reductive half reaction of the catalytic cycle with semiempirical QM/MM calculations. The active site model consists of small molecules having the same functionality as the chemically active residues of the protein. A point charge model protein environment consisting of 53 amino acids has been constructed to take into account the influence of the protein environment on the active site. Furthermore, the solvent influence on the active site model has been estimated using the SM3.1 model. The calculations indicate that only a hydride ion is transferred from glucose to the flavin coenzyme, whereas the proton is transferred to one of the active site histidines. This is different compared to redox reactions of free flavins. A second histidine at the active site is important for substrate binding and stabilization of intermediates. The solvent interchanges relative energies of the protonation sites of this histidine.
Glucose oxidase (GOX) has found a variety of applications, especially as
a biosensor for the quantitative determination of ß-D-glucose in liquids.
Similar to other flavoenzymes the enzymatic reaction can be divided in two
steps. In the reductive half raction two protons and electrons are transferred
from ß-D-glucose to the enzyme yielding d-gluconolactone as shown in Fig. 1. In the oxidative half reaction the enzyme is oxidized by molecular oxygen
yielding hydrogen peroxide. Finally d-gluconolactone can be hydrolysed non-enzymatically
to gluconic acid. This hydrolysis has been investigated already with semiempirical
PM3 calculations [1].
Fig. 1. The enzymatic reaction catalysed by glucose oxidase (GOX).
We carried out docking and force field studies to determine the structure
of the enzyme substrate complex. Our study reveals new aspects of the interactions
between glucose, FAD and active site residues. Additionally, we computed
the relative heats of formation for different tautomeric states of the reduced
enzyme with a mixed QM/MM method and investigated the stabilization of a
C4a-hydroperoxy dihydroflavin by the enzyme environment, which might be
a general feature of the flavoenzymes called GMC oxidases [2].
Protein structure
The coordinates for the force field and semiempirical calculations have
been taken from the X-ray structures of GOX refined to 2.3 [3] (PDB entry 1GAL) and 1.9 Å resolution.
Docking
Docking simulations with ß-D-glucose and delta-gluconolactone have been
carried out using correlation techniques [4].
Force field calculations
Subsequently to the rigid body docking of glucose into the substrate-binding
pocket of GOX described above, a flexible refinement with molecular mechanics
methods was carried out to form the non-covalently bonded Michaelis complex
using the AMBER all-atom force field [5] with additional parameters for glucose and FAD.
Fifteen crystallographic water molecules from the active site region were
added to the system and the space within a 10 Å radius of the N5 of isoalloxazine
with TIP3P water molecules [6]. The water molecules were constrained to
the centre of the sphere by a harmonic potential with a force constant of
0.5 kcal/mol2. Only the atoms of the residues inside the 10 Å radius were allowed to
move during these simulations. The force field calculations were carried
out at a dielectric constant 1 and an 8 Å cutoff distance for non-bonded
interactions was used. During the molecular dynamics simulations the SHAKE
algorithm [7] was used. The structures were energy-minimized and then equilibrated
by molecular dynamics for 50 ps at 298 K.
Semiempirical calculations
The PM3 method [8] implemented in Vamp and AMSOL was used throughout. The interaction between the quantum system and the
classical enzyme environment has been taken into account using a point-charge
model, which includes an electrostatic perturbation of the quantum system
wave function and van der Waals interactions with the environment.
Quantum region
The number of atoms which can be included in the quantum region representing
the active site is limited. Therefore we restricted this region to FAD and
the residues being potentially involved in the redox reaction, i. e. His
516, His 559 and Glu 412. For the investigation of the complexes glucose
and gluconolactone have been added at active site positions and in conformations
corresponding to the results of the docking and MD simulations. We substituted
the quantum system residues with smaller molecules which were in the same
relative orientation as in the X-ray structure. FAD is represented by lumiflavin
(Fl = 7,8,10-trimethyl isoalloxazine), His is represented by methyl imidazoles
(MI516, MI559) and a propionate ion (P412-) is used as a mimic for Glu. These model molecules shown in Fig. 2 have the same functionality as FAD and the corresponding amino acid residues.
We keep the numbering of the residues and their individual atoms to indicate
the assignment between the model and the enzyme. Kinetic studies [9] have
shown that a deprotonation step is necessary at low pH to obtain a catalytically
active GOX. We assumed a high pH with histidines are already unprotonated
and an overall quantum system charge of -1 according to Fig. 2. There are no further residues close to N5 of the FAD, which can act as
proton or electron relay system on the substrate.
Fig. 2. GOX quantum region. FAD is represented by lumiflavin {FloxH(N3)}, His by methyl imidazoles {MI}and Glu by propionate {P-}
Active site environment
The point charge environment was limited to 53 GOX residues located close
the quantum active model residues. This environment includes the substrate
binding residues and the immediate neighbours of the catalytically active
part, which do not participate in a direct way in the chemical reaction.
Atomic charges were computed using the natural atomic orbital / point charge
model [10].
Single point calculations were carried out at the optimized quantum system
without the environment to estimate the protein influence on the active
site . The influence of the solvent on the quantum system has been investigated
using the PM3-SM3.1 model [11].
DFT calculations
Isoalloxazine relevant for flavin redox reactions has been studied at B3LYP/6-31G*
level [12,13].
Docking simulations
Docking calculations with glucose indicate that there is only one substrate
orientation at the active site corresponding to a high correlation, i. e.
a good shape complementarity. This is a substrate orientation at the bottom
of the active site funnel with the hydrogen atom H(C1) of the substrate
located in the vicinity of FAD N5, the hydride acceptor. The hydroxy group
at C1 is directed roughly between His 516 and 559. Docking simulations with
gluconolactone led to a similar relative orientation at the active site.
Force field calculations
Initial force field calculations were then carried out in order to analyse
the protonation and orientation of the His 516 in the presence of glucose.
In a first step we generated six different structural data sets with three
different protonation states of the nitrogen atoms Ne and Nd of this His
and two orientations of its side chain for each protonation state. Furthermore,
initial structures with different torsional angles of the exocyclic hydroxy
methylene group of ß-D-glucose have been generated to find the conformer
with the lowest energy and to investigate the hydrogen bonds, which may
stabilize orientation of the substrate. Then we minimized the energy of
all complexes. To look for additional low energy conformations of glucose
at the active site we carried out MD simulations starting from selected
low energy structures and minimized the energy again.
The energetically most favourable system we found, is the one with hydrogen
atoms at His 559 Ne and at His 516 Nd. The carboxy oxygen Oe1 of Glu 412
is located at a distance of 2.8 Å to His 559 Ne which allows the formation
of a hydrogen bond to H(Ne). Each hydroxyl group of ß-D-glucose acts as
hydrogen donor and hydrogen acceptor at the same time. Therefore nine hydrogen
bonds stabilize the orientation of b-D-glucose at the active site.In addition
to these highly specific hydrogen bonds which are only possible in this
orientation of glucose the hydrophobic parts of glucose are in contact with
unpolar surface areas of the protein (Tyr 68, Phe 414, Trp 426).
Fig. 3. The hydrogen bond network between ß-D-glucose and the substrate binding site of GOX. The hydrogen atoms at C1 and O1 of glucose are shown in yellow.
The optimized distance between H( C1) of ß-D-glucose and N5 of the FAD
co-enzyme is 2.4 Å. The distance between H(O1) of the substrate and the
unprotonated Nd of His 559 is 1.8 Å. The short distances between FAD and
His residues of GOX on the one hand and H(O1) and H(C1) of b-D-glucose on
the other hand induce a hypothesis for the reaction mechanism, which is
deduced from the geometry (Fig. 2, Fig. 3). It seems most likely that H(C1) of ß-D-glucose is transferred to N5 of
FAD, and H(O1) may be transferred to Nd of His 559, which thereby becomes
positively charged. Subsequently the proton at Ne of the protonated His
559 may be transferred to the Oe1 of Glu 412. This potential mechanism of
the enzymatic reaction differs from the corresponding two electron reductions
of free flavins with a transfer of the hydrogens to N1 and N5 of the isoalloxazine
ring. A proton transfer to N1 of the FAD of GOX is unlikely because N1 is
located at the opposite side of the estimated substrate position (Fig. 3) and is surrounded by enzyme residues. Similarly a proton transfer to Ne
of His 516 seems less probable since this His is not activated by Glu or
Asp.
QM/MM calculations for the enzyme substrate and product complexes
Oxidized and reduced active site
The most stable enzyme substrate complex has a proton at MI559 Ne forming
a hydrogen bond with P412-, as indicated in Fig. 2. The PM3 optimized position of glucose at the active site is nearly independent
of the protonation state and it is close to the structure determined by
MD simulations.
The most stable enzyme product complex consists of FlredH2(N3N5)- (Fig. 4), MI 516 and MI 559 with hydrogen atoms at Nd and propionic acid with the
hydrogen at Oe1. The energy difference between the enzyme substrate and
product complexes is 55 kcal/mol. The transfer of both hydrogens from glucose
to the flavin N1 and N5 atoms, which occurs in free flavins, is 37 kcal/mol
less favourable in the enzyme. Gluconolactone does not form a hydrogen bond
with FlredH2(N3N5)- because this ligand is located somewhat more distant to the co-enzyme than
glucose in the enzyme substrate complex.
Fig. 4 The most stable two electron reduced active site
Reduced active site with oxygen
The oxidative half reaction of GOX is a re-oxidation with molecular hydrogen
(Fig. 1). Two electron reduced flavins are known to form C4a-hydroperoxy dihydroflavins
with molecular oxygen. If molecular oxygen reacts with FlredH2(N3N5)- to form FlredOOH2(N3N5)- and the protonation of other active site residues remains unchanged, a
hydrogen bond of 1.68 Å lengths can be formed between MI 559 H(Nd) and the
terminal oxygen of FlredOOH2(N3N5)- (Fig. 5, top). Density functional calculations for isolated flavins predict a tautomerisation
from the hydrogen at N5 to the terminal oxygen atom [15]. Similarly, PM3
calculations indicate that such a tautomerisation at the GOX active site
may lead to a more stable system with hydrogen bond to the terminal oxygen
instead of N5. For the tautomerized system with FlredOOH2(N3O)- the hydrogen bonding system is different. For this tautomeric state a hydrogen
bond with a distance of 1.85 Å is formed between the oxygen atom at C4a
and MI 516 Ne after imidazole ring rotation and in addition to the intermolecular
hydrogen bond an intra molecular hydrogen bond C4a-OOH... N5 of 1.84 Å stabilizes
the flavin (Fig. 5, bottom).
Fig. 5 Stabilization of C4a-hydroperoxy dihydroflavin anion and a more stable
tautomer with hydrogen bonds at the GOX active site.
top: FlredOOH2(N3N5)-, MI559(Nd), PH412, MI516(Nd), Gln 329
bottom: FlredOOH2(N3O)-, MI559(Nd), PH412, MI516(Ne), Gln 329
Influence of the protein environment and the solvent
The results of the previous section show that the catalytic function of
GOX depends not only of the flavin co-enzyme but also of other residues
at the active site. But a further influence of the enzyme environment on
the chemically active residues is also an important factor for the function.
This is demonstrated by a comparison of the relative heats of formation
for the quantum system with and without the PCM environment. For the oxidized
and two electron reduced enzyme the most stable system identical in the
presence and absence of the environment, except for a tautomerisation at
MI516. But the relative heats of formation for unfavourable tautomers are
much higher in the presence of the environment. This means that the protein
environment stabilizes specific tautomers in course of the catalytic cycle.
Similarly, FlredOOH2(N3O)- is substantially favoured at the active site by the enzyme environment
relative to FlredOOH2(N3N5)-.
The energy difference for the different protonation sites of His 516 are
small. A hydrogen bond to the carboxamide oxygen of Gln 329 can be formed,
if His 516 is protonated at Nd. If this residue is removed from the PCM
environment, lower heats are computed for a hydrogen atom at Ne of His 516,
except for the system shown in Fig. 5.
To estimate the influence of a solvent on the model system, we have calculated the free energy of solvation for the quantum system at fixed geometries without the enzyme environment. The calculations indicate that the most stable model systems for the enzyme substrate and product complexes are not changed upon solvation, except for a tautomeriztion at His 516. These calculations can only provide a rough estimate of solvent effects since only a small part of the catalytic residues is solvent accessable.
The active site of GOX is located at the bottom of a deep pocket of the enzyme, which has the shape of a funnel with a cross section of 10 Å X 10 Å at the surface and a distance between the surface and N5 of FAD in the range between 13 and 18 Å [3]. Test calculations with an enzyme environment augmented with additional residues consisting to certain secondary structure elements like helix 13 pointing towards the FAD do not lead to significant changes of the relative heats of formation. This indicates the selected number of environment residues are sufficient. The applied point charge model used for the enzyme environment can describe molecular multipole moments and electrostatic potentials with a high accuracy [10]. Continuum model calculations indicate that the most stable enzyme substrate and product complex remains unchanged solvent influence.
A two electron reduction with a N1 protonation of FAD is unlikely (see above) because the N1 is located at the opposite side of H(O1) of glucose (Fig. 3). Additionally, the semiempirical calculations for the model system exclude this possibility because a system with a lower heat of formation exists. The calculations indicate that the most stable product for the enzyme part of the reductive half reaction (Fig. 1) is the system with a hydride ion from the glucose transferred to the N5 of FloxH(N3) yielding FlredH2(N3N5)- and a proton transferred to Nd of MI 559 with a subsequent transfer of the hydrogen from Ne of MI 559 to Oe1 of P412-. The predicted reaction is different from the reduction of free flavins with a transfer of both hydrogen atoms to N1 and N5 of the isoalloxazine ring. This is certainly part of the role of the enzyme environment in catalysis.
Comparison with experiments
Manstein [14] reconstituted GOX with FAD analogs and derived the absolute
stereochemistry for the interaction with glucose. The experiments indicate
in correspondence with our computed enzyme substrate structures that the
substrate interacts from the re face of the flavin. Furthermore, the hydrogen atoms at O1 and C1 of glucose
are located at short distances relative to the acceptor atoms, which is
necessary for hydrogen transfer.
Even though the GOX structure was not known at that time, Weibel and Bright
[9] proposed on the basis of kinetic data that the hydrogen at N3 is connected
via a hydrogen bond with the enzyme and the formation or breaking of this
bond may cause changes of the planarity of the isoalloxazine nucleus, which
should influence the reactivity. A hydrogen bond between the hydrogen atoms
located at N3 and the backbone oxygen of Thr 119 of the enzyme indeed exists
in the experimental protein structure [3]. Though free neutral flavins are
planar in the oxidized state, the co-enzyme has a nonplanar butterfly-type
structure in the protein with a fold angle of 16.4° along the N5-N10 axis
even in the oxidized state of the GOX, which is fixed by hydrogen bridges
and sterical constraints which probably activates FAD. Therefore one may
think that the removal of the hydrogen bond may lead to a serious deformation
of the fold angle and thus may influence the reactivity, because the isoalloxazine
ring is already folded in an angle, which is energetically favourable in
the reduced state. The energy minimum of the anionic two electron reduced
isoalloxazine corresponds to a fold angle of 14.7° [15]. But the energy
difference between the planar and folded structure is very small in the
two electron reduced state. Therefore the fold probably affects only the
activation energy of the flavine redox reaction.
Our prediction is also in agreement with an 15N NMR study [16]. The analysis of the spectra shows that the reduced flavin
of GOX is negatively charged with hydrogen atoms at N3 and N5. N1 of the
reduced FAD does not contain a hydrogen. These experimental results are
indeed part of our prediction.
Related Enzymes
The proposed enzyme co-enzyme interaction is also relevant for the homologous
enzyme family termed GMC oxidoreductases [2], which include glucose dehydrogenase,
choline dehydrogenase, cholesterol oxidase, alcohol oxidase in addition
to GOX. In these enzymes a His is conserved, which is located for cholesterol
oxidase (PDB entries 3COX, 1COY) in the same relative orientation to FAD as His 516
of GOX. Thus this residue is also involved in the oxidative half reaction
of the cholesterol oxidase catalytic cycle.
In summary, we think that the combination of classical and quantum mechanical
methods is a powerful tool for the study of enzymatic catalysis. We have
determined substrate binding residues presented a catalytic mechanism for
GOX, which is consistent with experimental data. The calculations show that
only the hydride ion is transferred to the FAD co-enzyme. His 516 plays
an important role in the oxidative half reaction. In addition to the chemically
active residues the enzyme environment is important for the stabilization
of the products of the reductive half reaction.
We thank H. J. Hecht for providing the X-ray structure of glucose oxidase
and H. M. Kalisz (GBF, Braunschweig) stimulating discussions. A portion
of this work was supported by a European Union Access to Large Scales Facilities
grant (ERBCHGECT940062) to EMBL.