Download - University of Surrey

1
Docking and DFT Studies on Ligand Binding to Quercetin 2,3-Dioxygenase
Aramice Y. S. Malkhasian1 and Brendan J .Howlin2
1
Department of Chemistry, King Abdulaziz University, Faculty of Science, Jeddah, 21589, Saudi Arabia
2
Department of Chemistry, FEPS, University of Surrey, Guildford, Surrey, GU2 7XH, UK
ABSTRACT
Simple molecular docking calculations on quercetin, kojic acid and diethylcarbamatodithoic acid
using the software package MOE are shown to be close to the geometries reported in the X-ray
crystal structures of the protein co-crystallised with the respective ligands. Furthermore DFT
optimization of the docked conformations is shown to reproduce the essential features of previous
studies on quercetin, showing that docking can be used to provide good starting structures for
mechanistic study. The flavone ligand, lacking the hydroxyl group of the quercetin is shown by
docking to be unable to approach closely the copper atom, indicating the necessity of the presence
of the hydroxyl group and providing a prediction of the likely binding environment of this ligand.
Keywords: Quercetin; docking; DFT; mechanism
2
INTRODUCTION
Quercetin 2,3-dioxygenase (2,3-QD) is a member of the cupin superfamily, which is a group of
proteins possessing a common fold (Dunwell, Culham, Carter, Sosa-Aguirre and Goodenough,
2001). 2,3-QD is responsible for degradation of quercetin (3,3',4,,5,7 pentahydroxyflavone) via
coordination of the substrate to the copper in the active site and subsequent oxidative cleavage to
yield the corresponding 2-protocatechuoylphloroglucinol carboxylic acid and carbon monoxide.
During the reaction, dioxygen is consumed, Scheme 1.
Scheme 1. Mechanism of Quercetin 2,3-dioxygenase
The structure of 2,3-QD has been determined by Fusetti et al, (Fusetti et. al., 2002). The crystal
structure at 1.6 Å resolution shows that the Cu2+ is bound distorted tetrahedally to His66, His68 ,
His112 and a water molecule. The X-ray absorption spectroscopy (Steiner, Kalk and Dijkstra, 2002)
and EPR (Kooter, et al, 2002) data indicate that, on anaerobically binding of the flavonol substrates
[enzyme_substrate (E_S) complexe] the site becomes penta-coordinate and ordered. The copper
ion maintains a formal Cu(II) oxidation state. The reaction is the breaking of oxygen to release
carbon monoxide by breaking the O-heterocycle ring in flavones. This differs from non-heme iron
intra and extradiol dioxgenases reactions (Que and Ho, 1996), (Solomon et al., 2000). According to
Malkhasian et al. (Malkhasian, Finch, Nikolovski, Menon, Kucera and Chavez, 2007) the following
3
mechanism has been suggested between the enzyme and the substrate and is illustrated in
Scheme 2. The copper environment in 2,3-QD from Aspergillus japonicus has been investigated by
X-ray crystallography (Steiner, Kalk and Dijkstra, 2002), (Steiner, Kooter and Dijkstra, 2002) and
electron paramagnetic resonance (EPR) (Kooter, et al., 2002). The 1.6 Å resolution crystal structure
revealed that the metal is primarily bonded to three histidine nitrogens (His-66, His-68, His-112)
and a water molecule in a distorted tetrahedral fashion. A minor mixed trigonalbipyramidal/square-pyramidal geometry is also present. In the latter, the water is positioned in the
equatorial plane and Glu-73 is additionally coordinated. X-ray studies have revealed that anaerobic
binding of the flavonol substrate results in an ordered copper(II) coordination with displacement of
the water molecule by a deprotonated hydroxy group of the flavonol and additional coordination
by Glu-73 (Siegbahn, 2004). In these studies, it is suggested that the carboxylate group of Glu-73
may serve as a stabilizing group for the bound substrate (Scheme 1). It has also been found that
deletion of this group results in an inactive enzyme (Seigbahn, 2004).No x-ray structures of bound
flavones have been deposited, so it was interesting to see if the lack of the hydroxyl group would
affect the binding. This is not to imply that it is not possible to produce stable crystals of flavone
soaked Quercetin 2,3-dioxygenase just that currently no x-ray structures exist. In this paper we
employ computational techniques similar to those used by Tao et al. (Tao et al, 2009) except that
our work does not include Molecular Dynamics. Vasilescu and Girma reported the first molecular
orbital calculations on Quercetin (Vasilescu and Girma, 2002). Fiorucci at al. have studied the
interactions of quercetin with the enzyme by DFT calculations (Fiorucci, Golebiowski, Cabrol-Bas
and Antonczak, 2006), (Fiorrucci, Golebiowski, Cabrol-Bas and Antonczak, 2007), (Fiorruci,
Golebiowski, Cabrol-Bas and Antonczak, 2004) as has Siegbahn (Siegbahn, 2004). We have
employed the technique of molecular docking, more commonly used in drug design, to produce
starting structures for a mechanistic investigation of Quercetin 2,3-dioxygenase, indicating the
potential for this technique in future studies of enzyme mechanism.
4
Computational Methods
The x-ray structure of Quercetin 2,3-dioxygenase (1JUH.pdb) with a resolution of 1.6Å was used
throughout (Fusetti et al., 2002).
The substrate (quercetin or 3,4,5,7-tetrahydroxyflavonol),
intermediates and product (2-protocatechoylphloroglucinol carboxylic acid) were built using the
BUILDER module of MOE 2012.10 (Chemical Computing Group Ltd.) The small molecules were
energy minimised to convergence using the smart minimiser in MOE with the MMFF94X forcefield
(Halgren, 1996) using a permittivity of 1 and a non-bonded cutoff of 8Å. MMFF94x uses an internal
bond-charge increment charge model. Substrates, intermediates and product were docked into the
active site of the protein (where the copper ion is located) using the DOCK module of MOE. The
SITE finder module of MOE was used to locate suitable spaces in the protein structure and the site
located closest to the copper was filled with dummy atoms to specify the site. The ligands were
docked using the induced fit method (this allows both the ligand and protein to change shape to
ameliorate the docking) with the triangle matcher using London DG rescoring and refined using the
AMBER99 forcefield (Wang, Cieplak and Kollman, 2000) with GBVI/WSA DG rescoring. The ligand
with the lowest value of S (corresponds to ΔG) was selected from the database of docked ligands in
every case and the interactions visualised using the LIGAND INTERACTIONS module of MOE.
Molecular orbital calculations were carried out on the docked complexes using GAUSSIAN09.
(Frisch et al., 2009) Structures were optimised using DFT under B3LYP with a DGTZVP basis set.
Vibrational frequencies were calculated from the optimised geometries and the frequencies were
searched for negative frequencies which correspond to transition states. Mulliken population
analysis was used to determine point charges on the copper.
Results and Discussion
The active site of the protein is shown in Figure 1. The x-ray structure is a dimer, so one molecule
was selected for analysis (in this case the A chain).The X-ray determined sugars and waters were
5
left in the structure throughout the analysis. SITE 1 detected by site finder was used for the
docking. Site 1 is the largest site and contains (VAL30 ASP31 THR32 GLN33 TYR35 MET51 GLY52
THR53 ALA55 PRO56 SER58 ALA60 LEU61 GLY62 VAL63 LEU64 HIS66 HIS68 GLU73 ASN74 PHE75
HIS112 PHE114 ILE116 MET123 THR124 GLY125 VAL126 ILE127 PHE132 LEU135 PHE136 LEU139
LEU172 VAL177 CU401). This site also contains the catalytic copper ion. This site is shown in
Figure 2 with the dummy atoms positioned. The mechanism of the protein is shown in Scheme 2.
Docking of the substrate (quercetin) resulted in a close approach to the copper atom of 2.19 Å by
the 3-OH of the quercetin (see Scheme 1 for atomic numbering scheme). Table 1 shows the
docking energy and Figure 3 A the ligand interaction diagram. This arrangement agrees with the
scheme presented by Malkhasian et al. (Malkhasian et al., 2007), where the 3-OH is also bound to
the copper (interactions to the copper are not shown in this diagram as their distances are larger
than accepted H-bonding distances) and the amino acid residue GLU73. Following on from this
encouraging result the rest of the molecules in the proposed reaction scheme from Steiner, Kalk
and Dijkstra (Steiner, Kalk and Dijkstra, 2002), were also docked to the same site. The next stage in
the scheme is called 3b in the Dijsktra paper (Steiner, Kalk and Dijkstra, 2002) and requires a
dioxygen to be bound to the copper atom. A dioxygen was added to the copper and the docking
carried out. The poses derived from the docking of each stage in the mechanism are shown in
figure 4 as the 10 lowest energy poses with the lowest energy (used for the calculations) shown in
green. The flexibility of the ligands about their torsion angles can clearly be seen, with stage 5
showing the least variation. There is no significance in the poses showing hydrogen bonding in this
figure as this is an artefact of the the subtraction of the protein for this display type. The ligand
interaction diagram is shown in Figure 3 B. A close approach to the amino acid residue HIS68 is
shown in this docking. The ligand interaction diagram for the dock of Stage 4 is shown in Figure 3
C. Again it is the oxygen identified by Dijkstra et al. (Steiner, Kalk and Dijkstra, 2002 ) that has the
closest approach to the copper atom. The docking of the product (Stage 5) is shown in Figure 3 D).
Again in agreement with Dijkstra et al. (Steiner, Kalk and Dijkstra, 2002) the lowest energy docked
6
structure shows the carbonyl of the carboxylic acid bound to the copper. It is interesting to note
that GLU73 is featured throughout the dockings, again supporting the views of Dijsktra et al.
(Steiner, Kalk and Dijkstra, 2002), that the close approach of this residue is important in the
mechanism for coordinating the copper ion. A consideration of the simulated free energies, given
as ΔG values (MOE reports its docking scores as S values instead of ΔG values because it uses a
linear free energy scaling relationship to approximate ΔG, so these are not true ΔG values) in Table
1, shows that there are favourable free energies for all docked ligands. The free energy for the
substrate is less than the first intermediate (3b), then there is a favourable difference for the
second intermediate (4) before the free energy goes up again for the product (5). These data are
plotted in figure 5, as -1*binding energy to make the diagram positive. It clearly shows how the
free energy changes over the course of the reaction. Likewise the closest approach to the copper is
for the intermediate stage 3b and the distance to the copper increases for the later stages of the
reaction.
Crystal structures of 2 inhibitors docked to the enzyme are known, these are
diethylcarbamatodithoic acid (1GQG.pdb) (Steiner, Kooter and Dijkstra, 2002)[20] and Kojic acid(
1GQH.pdb) (Steiner, Meyer-Klaucke and Dijkstra, 2002), both at relatively high resolution. These
two inhibitors were also docked to the structure and the results are presented in Table 1 and figure
6. Diethylcarbamatodithoic acid docks into approximately the correct position making an
interaction with the copper with the same ligand atom as in the x-ray structure, although the
second sulphur approaches TYR35 in the docked structure, a position that is occupied by one of the
ethyl groups in the x-ray structure. In the Kojic acid, the docked structure again makes one of the
interactions with the copper that are present in the x-ray structure but not the second. The lack of
two interactions with the copper may account for the lower free energy of binding in both of these
cases, i.e. about -4 Kcal/Mol, the chelate effect leading to greater stability. As these are both
inhibitors one would expect their free energy of binding to be greater than the natural substrate
thereby accounting for their ability to replace the substrate. The initial docking of the flavone
(lacking the hydroxyl group) located the ligand in the correct site with a favourable binding energy
7
but the distance from the copper was 4.2 Å, a water molecule occupying the space that would
allow a closer approach to the copper (Figure 7). Redocking this ligand with the water molecule
deleted did not produce any poses with the ligand closer to the copper indicating that the lack of
the hydroxyl group is compensated by binding a water molecule.
The docked intermediate (3b) was optimised by DFT including the copper ion, the dioxygen,
the ligand and the coordinating amino acid residues. The charge on the copper ion was found to be
0.692 electrons which is well within the range found by Siegbahn (Siegbahn, 2004) of 0.67-0.74. The
distances from the copper to the oxygens of the ligand are 1.96Å and 3.13Å respectively, showing
that the ligand has moved further away from the copper from the close approach during the docking
(1.18 Å). The angle between the dioxygen, the copper and the closest oxygen of the ligand is 95°,
close to the octahedral geometry expected. A comparision of the HOMO of the complex with the
LUMO shows that the HOMO mostly covers the copper and the dioxygen whereas the LUMO covers
the copper, dioxygen and the first half of the ligand containing the oxygen atoms (Figure 8). In
Frontier Molecular orbital theory the HOMO of one reactant reacts with the LUMO of the other, the
energy difference between these two orbitals being (-5.993857-(-4.39001) eV = 1.60384 eV = 36.8
kcal/mol. This energy difference is rather high and indicates the need to include the protein in the
calculation to facilitate this barrier.
Zhang and Musgrave (Zhang and Musgrave, 2007) have
surveyed the HOMO-LUMO gaps calculated by DFT methods and concluded that the energies of the
HOMO and LUMO differ significantly from the negative ionization potentials and electron affinities.
They have proposed an empirical equation that corrects the energy of the HOMO and applying this
we get -8.612628 eV for the HOMO and for the LUMO, -7.00879 eV. This gives the same band gap as
we have only shifted the energies not improved the accuracy of the basic calculation. Interestingly
the stage known as 3a in the scheme has almost the same energy as stage 3b, i.e. -83531.87 eV for
stage 3a compared to -83532.69 eV for 3b. Therefore there is a difference of 0.82 eV in terms of
energy calculated by DFT for either stage, this corresponds to 18.9 Kcal/mol, indicating a preference
for stage 3b over stage 3a. The intermediate known as stage 4 has an energy of -83534.59 eV by DFT
8
compared to -83532.69 eV from stage 3b. This stage has a partial charge of 0.632 electron on the
copper atom, indicating that the charge has decreased slightly from stage 3. The distances of the
oxygens of the ligand to the copper are 2.01 and 1.92, i.e. both closer than in stage 3b, this
represents the change from both being double bonded oxygens to one being single bonded. Both
are significantly less than in the original docked complex. The HOMO of stage 4 (figure 8 C) shows
that it occupies half of the ligand, there is a small contribution from the copper but none from the
protein. The LUMO (figure 8 D) occupies the other half of the ligand which is perfectly consistent
with the accepted mechanism which involves attack from one half of the ligand to the other
resulting in breaking the bond in the central ring. Stage 5 has an energy of -80454.54 eV
Conclusion
The simple docking of the substrate, intermediates and product of the reaction of quercetin with
quercetin 2,3-dioxygenase is able to produce good starting structures for further optimisation by
DFT calculations which are shown to reproduce closely the structure of the proposed structures in
the mechanism proposed by Dijkstra et al. (Steiner, Kalk and Dijkstra, 2002). As the simple docking
procedure does not allow for covalent bond formation or bond breaking it is necessary to
additionally employ methods that do allow this in order to further investigate enzyme
mechanisms. These encouraging results show the potential for using modern computational
docking for investigating enzyme mechanism and also for generating starting structures for more
accurate molecular orbital calculations.
9
Legend to Figures
Figure 1: Structure of the Enzyme from PDB file 1JUH.pdb.
Figure 2. Active site of protein as defined by site finder. Dummy atoms defining the docking site are
shown.
Figure 3. A) Ligand interaction diagram for quercetin. The names in the spheres show the amino
acids in the vicinity of the ligand. Hydrogen bonds are shown with dotted lines. B) Ligand
interaction diagram for stage 3b of mechanism. C) Ligand interaction diagram for stage 4. There is
an interaction with KMP in this diagram because two oxygens in the molecule are close enough to
be regarded as an interaction. D). Ligand interaction diagram for stage 5.
Figure 4 Overlay of A) the 10 lowest energy poses of stage 3 docked to the protein. The lowest
energy dock is shown in green. B) stage 3b. C)Stage 4 (only 5 lowest energy poses shown here) and
D) Stage 5.
Figure 5 Plot of the free energy changes in Kcal/mol over the course of the reaction from the
docking energies (plotted as -1*binding energy for clarity) against the stage of the reaction. Stage 1
is the energy of the substrate. Stages 3b2, 4 and 5 refer to the scheme in Scheme 2.
Figure 6. Ligand interaction diagram for docking of Diethylcarbamodithoic acid, A) docked
structure, B) X-ray structure. Ligand interaction diagram for docking of Kojic acid C) docked
structure, D) X-ray structure.
Figure 7 Ligand interaction diagram for docking of flavone.
Figure 8 HOMO (A) and LUMO (B) of stage 3b after optimisation by DFT. HOMO (C) and LUMO (D)
of stage 4 after optimisation by DFT.
10
Figure 1
PDB:
IJUH
11
Figure 2
12
Scheme 2. Proposed reaction scheme for quercetin 2,3 dioxygenase mediated dioxygenation of
flavonols (taken from Dijkstra et al.3)
13
Ligand
Quercetin (substrate)
3b (intermediate)
3b2 (intermediate)
4 (intermediate)
5 (product)
Diethylcarbamodithoic acid
Kojic Acid
Flavone
ΔGbind (Kcal/mol)
-6.3632
-5.6611
-7.6598
-7.5595
-6.6057
-4.1739
-4.1600
-5.6578
Table 1 Docking energies for ligands.
Distance to copper (Å)
2.19 (3-OH)
1.18 (3-O)
2.51
2.65
2.56
3.29 (to sulphur)
3.55 (to ring oxygen)
4.20
14
Figure 3
A)
C)
B)
D)
15
Figure 4 Overlay of A) the 10 lowest energy poses of stage 3 docked to the protein. The lowest
energy dock is shown in green. B) stage 3b. C)Stage 4 (only 5 lowest energy poses shown here) and
D) Stage 5.
A)
B)
C)
D)
16
Figure 5
17
Figure 6
A)
C)
B)
D)
18
Figure 7
19
Figure 8
A)
C)
B)
D)
20
References
Chemical Computing Group Ltd, (www.chemcomp.com)
Dunwell, J. M., Culham, A., Carter, C. E., Sosa-Aguirre, C. R., Goodenough, P. W. (2001). Evolution of
Functional Diversity in the Cupin Superfamily. Trends Biochem. Sci, 26, 740–746.
Fiorucci, S., Golebiowski, J., Cabrol-Bas, D. & Antonczak, S. (2006). Molecular Simulation Reveals a
New Entry Site in the Quercetin 2,3-dioxygenase. A Pathway for Dioxygen?. Proteins, 64, 845-850.
Fiorucci, S., Golebiowski, J., Cabrol-Bas, D. & Antonczak, S. (2007). Molecular Simulations Bring New
Insights into Flavonoid/Quercetinase Interaction Modes. Proteins, 67, 961-970.
Fiorucci, S., Golebiowski, J., Cabrol-Bass, D. & Antonczak, S. (2004). Oxygenolysis of Flavonoid
Compounds: DFT Description of the Mechanism for Quercetin. ChemPhysChem, 5, 1726-1733.
Frisch, M. J., Trucks, G. W., Schlegel, H. B., Scuseria, G. E., Robb, M. A., Cheeseman, J. R., & Petersson,
G. A. (2009). Gaussian 09, revisions B. 01 and A. 02; Gaussian Inc.: Wallingford, CT.
Fusetti, F. , Schroter, K. H., Steiner, R. A., Van-Noort ,P. I., Pijning,T. , Rozeboom, H.J., Kalk, K. H.,
Egmond, M. R., Dijkstra,B. W. (2002) Crystal Structure of the Copper-Containing Quercetin 2,3Dioxygnease from Aspergillus Japonicus. Structure 10, 259-268.
Halgren, T.A. (1996) The Merck Force Field. J. Comp. Chem. 17, 490–641.
Kooter, I. M., Steiner, R. A., Dijkstra, B. W., van Noort, P. I., Egmond, M. R., Huber, M. (2002) EPR
characterization of the mononuclear Cu-containing Aspergillus japonicus quercetin 2,3-dioxygenase
reveals dramatic changes upon anaerobic binding of substrates. Eur. J. Biochem, 269, 2971-2979.
Malkhasian, Aramice Y.S., Finch, M.E., Nikolovski, B., Menon, A., Kucera, B.E., Chavez, F.A. (2007) N,
N’-dimethylformamide-derived products from Catalytic oxidation of 3-Hydroxyflavone. Inorg.
Chem., 46, 2950-2952
Tao, P., Fisher, J.F., Shi,Q., Vreven, T., Mobashery, S. and Schlegel, H.B.(2009) Matrix
Metalloproteinase 2 Inhibition: Combined Quantum Mechanics and Molecular Mechanics Studies of
the Inhibition Mechanism of (4-Phenoxyphenylsulfonyl)methylthiirane and Its Oxirane Analogue.
Biochemistry, 48, 9839–9847.
Que, L. Jr, Ho Ry.(1996) Dioxygen Activation by Enzymes with Mononuclear Non-Heme Iron Active
Sites. Chem. Rev., 96, 2607-2624.
Siegbahn, P. M. E. (2004) Hybrid DFT Study of the Mechanism of Quercetin 2,3-Dioxygenase. Inorg.
Chem., 43, 5944-5953.
21
Solomon, E. I., Brunold, T. C., Davis, M. I., Kemsley, J. N., Lee, S.K., Lehnert, N., Neese, F., Skulan, A. J,
Yang, Y.S., Zhou, J. (2000) Geometric and Electronic Structure/Function Correlations in Non-Heme
Enzymes, Chem. Rev., 100, 235-349.
Steiner, R. A., Kalk, K. H., Dijkstra, B. W. (2002) Anaerobic enzyme⋅substrate structures provide insight
into the reaction mechanism of the copper-dependent quercetin 2,3-dioxygenase. Proc. Natl. Acad.
Sci., 99, 16625-16630.
4
.
Steiner, R. A. , Kalk, K. H. , Dikstra, B. W.(2002) Anaerobic enzyme⋅substrate structures provide insight
into the reaction mechanism of the copper-dependent quercetin 2,3-dioxygenase. PNAS, 99, 26,
16625-16630.
Steiner, R.A., Kooter, I.M., Dijkstra, B.W.(2002) Functional analysis of the copper-dependent quercetin
2,3-dioxygenase. 1. Ligand-induced coordination changes probed by X-ray crystallography: inhibition,
ordering effect, and mechanistic insights. Biochemistry, 41, 7955-7962.
Steiner, R.A., Kooter, I.M., Dijkstra, B.W. (2002) Functional analysis of the copper-dependent quercetin
2,3-dioxygenase. 1. Ligand-induced coordination changes probed by X-ray crystallography: inhibition,
ordering effect, and mechanistic insights. Biochemistry, 41: 7955-7962.
Steiner, R. A., Meyer-Klaucke, W., Dijkstra, B. W. (2002) Functional analysis of the Copper-dependent
quercetin 2,3-dioxygenase. 2. X-ray Absorption studies of native enzyme and anaerobic complexes
with the substrates Quercetin and Mericytin. Biochemistry , 41, 7963-7968.
11)
12) Vasilescu, D., Girma, R. (2002) Quantum Molecular Modelling of Quercetin - Simulation of the
Interaction with the Free Radical t-BuOO. International Journal of Quantum Chemistry, 90, 888-902.
Wang, J., Cieplak, P., Kollman, P.A. (2000) How Well Does a Restrained Electrostatic Potential (RESP)
Model Perform in Calculating Conformational Energies of Organic and Biological Molecules?;J.
Comput. Chem. 21, 1049–1074.
22) Zhang, G. and Musgrave, C.B.(2007) Comparison of DFT Methods for Molecular Orbital
Eigenvalue Calculations. J. Phys. Chem. A, 111, 1554-1561.