Finding molecular dioxygen tunnels in homoprotocatechuate 2,3-dioxygenase: Implications for different reactivity of identical subunits

Abstract

Extradiol dioxygenases facilitate microbial aerobic degradation of catechol and its derivatives by activating molecular dioxygen and incorporating both oxygen atoms into their substrates. Experimental and theoretical studies have focused on the mechanism of the reaction at the active site. However, whether the catalytic rate is limited by O(2) access to the active site has not yet been explored. Here, we choose a recently solved X-ray structure of homoprotocatechuate 2,3-dioxygenase as a typical example to determine potential pathways for O(2) migration from the solvent into the enzyme center. On the basis of the trajectories of two 10-ns molecular dynamics simulations, implicit ligand sampling was used to calculate the 3D free energy map for O(2) inside the protein. The energetically optimal routes for O(2) diffusion were identified for each subunit of the homotetrameric protein structure. The O(2) tunnels formed because of thermal fluctuations were also characterized by connecting elongated cavities inside the protein. By superimposing the favorable O(2) tunnels on to the free energy map, both energetically and geometrically preferred O(2) pathways were determined, as also were the amino acids that may be critical for O(2) passage along these paths. Our results demonstrate that identical subunits possess quite distinct O(2) tunnels. The order of O(2) affinity of these tunnels is generally consistent with the order of the catalytic rate of each subunit. As a consequence, the probability of finding the reaction product is highest in the subunit containing the highest O(2) affinity pathway.

Full text

ORIGINAL PAPER
Finding molecular dioxygen tunnels in homoprotocatechuate
2,3-dioxygenase: implications for different reactivity
of identical subunits
Liang Xu •Weijie Zhao •Xicheng Wang
Received: 7 May 2009 / Revised: 9 September 2009 / Accepted: 24 September 2009 / Published online: 14 October 2009
ÓEuropean Biophysical Societies’ Association 2009
Abstract Extradiol dioxygenases facilitate microbial
aerobic degradation of catechol and its derivatives by
activating molecular dioxygen and incorporating both
oxygen atoms into their substrates. Experimental and
theoretical studies have focused on the mechanism of the
reaction at the active site. However, whether the catalytic
rate is limited by O
2
access to the active site has not yet
been explored. Here, we choose a recently solved X-ray
structure of homoprotocatechuate 2,3-dioxygenase as a
typical example to determine potential pathways for O
2
migration from the solvent into the enzyme center. On the
basis of the trajectories of two 10-ns molecular dynamics
simulations, implicit ligand sampling was used to calculate
the 3D free energy map for O
2
inside the protein.
The energetically optimal routes for O
2
diffusion were
identified for each subunit of the homotetrameric protein
structure. The O
2
tunnels formed because of thermal
fluctuations were also characterized by connecting
elongated cavities inside the protein. By superimposing the
favorable O
2
tunnels on to the free energy map, both
energetically and geometrically preferred O
2
pathways
were determined, as also were the amino acids that may be
critical for O
2
passage along these paths. Our results
demonstrate that identical subunits possess quite distinct
O
2
tunnels. The order of O
2
affinity of these tunnels is
generally consistent with the order of the catalytic rate of
each subunit. As a consequence, the probability of finding
the reaction product is highest in the subunit containing the
highest O
2
affinity pathway.
Keywords Oxygen pathways Oxygen tunnels 
Extradiol dioxygenase Implicit ligand sampling
Introduction
Dioxygenases are non-heme iron-containing enzymes that
play an important role in the biodegradation of catechol and
its derivatives by catalyzing the cleavage of aromatic rings in
either an intradiol or extradiol manner. Generally, the intra-
diol dioxygenases require Fe
3?
to cleave C–C bond between
the phenolic hydroxy groups, producing cis,cis-muconic
acid, whereas the extradiol dioxygenases use Fe
2?
or Co
2?
as
a cofactor to cleave the C–C bond adjacent to the phenolic
hydroxy groups, yielding 2-hydroxymuconaldehyde (Bugg
and Lin 2001; Siegbahn and Haeffner 2004; Georgiev et al.
2008). As a result, an O
2
is activated and both oxygen atoms
are inserted into the products. Several comprehensive
reviews on dioxygenase mechanisms have been published
recently, including those focusing on geometric and elec-
tronic structures (Wallar and Lipscomb 1996; Solomon et al.
2000), biomimetic modeling and intermediates (Costas et al.
Electronic supplementary material The online version of this
article (doi:10.1007/s00249-009-0551-9) contains supplementary
material, which is available to authorized users.
L. Xu X. Wang (&)
Department of Engineering Mechanics, State Key Laboratory
of Structural Analyses for Industrial Equipment,
Dalian University of Technology, 116023 Dalian, China
e-mail: guixum@dlut.edu.cn
L. Xu
e-mail: xuliang0889@yahoo.com.cn
L. Xu
Department of Chemistry, Dalian University of Technology,
116023 Dalian, China
W. Zhao
School of Chemical Engineering, State Key Laboratory
of Fine Chemicals, Dalian University of Technology,
116012 Dalian, China
123
Eur Biophys J (2010) 39:327–336
DOI 10.1007/s00249-009-0551-9

2004; Kryatov and Rybak-Akimova 2005), and reaction
mechanism (Abu-Omar et al. 2005; Bugg 2003; Lipscomb
2008). The catalytic mechanism of the non-heme iron cate-
chol dioxygenase has also been theoretically studied by using
density functional theory (Siegbahn and Haeffner 2004;
Deeth and Bugg 2003).
Many crystal structures of dioxygenases with or without
bound substrates have been determined. From the sequence
alignments of 35 extradiol dioxygenases it is apparent the
three metal ligands His155, His214, and Glu267 (num-
bering from homoprotocatechuate 2,3-dioxygenase (2,3-
HPCD); PDB entry 2IGA or 2IG9), a motif referred to as
the 2-His-1-carboxylate facial triad, are strictly conserved
residues in the active site pocket (Solomon et al. 2000).
There are at least three possible mechanisms that could
account for the dioxygen reactivity and the corresponding
intermediates (Solomon et al. 2000). Recently, a crystal
structure of Fe
2?
-containing 2,3-HPCD (EC 1.13.11.15)
from Brevibacterium fuscum has been resolved success-
fully, with three different intermediates in identical sub-
units of a single homotetrameric enzyme (Kovaleva and
Lipscomb 2007). Thus, a general mechanism strategy for
the diverse extradiol dioxygenases has emerged (Kovaleva
and Lipscomb 2007; Kovaleva et al. 2007; Lipscomb
2008). Furthermore, these intermediates have not been
observed independently in a crystal structure. How these
subtle differences in identical subunits control the O
2
binding, activation, and insertion still remains unclear.
In addition, the crystallization was conducted in the slow
substrate 4-nitrocatechol (4NC) in a low-O
2
atmosphere.
Thus, the precise, atomic-resolution pathways for O
2
migration in the protein, with predicted relative significant
parts of the pathways, should help to rationalize the
selectivity of the specific intermediate at the active site of
each subunit and greatly facilitate the selection of specific
site mutations for such studies.
In contrast with non-heme proteins, O
2
migration path-
ways for myoglobin (Teeter 2004; Cohen et al. 2006; Cohen
and Schulten 2007; Orlowski and Nowak 2007; Ruscio et al.
2008) and copper amine oxidase (Johnson et al. 2007), and
H
2
pathways for [Ni-Fe]-hydrogenase (Cohen et al. 2005;
Leroux et al. 2008) and [FeFe]-hydrogenase (Teixeira et al.
2006) have been explored by experimental and computer
simulations. To date, maps of O
2
pathways in non-heme iron
enzymes have only been characterized by molecular
dynamics (MD) simulations for a small number of proteins,
including quercetin 2,3-dioxygenase (van den Bosch et al.
2004; Fiorucci et al. 2006) and 12/15-lipoxygenase (Saam
et al. 2007). The results from these studies have important
implications for enzymatic activity and for protein engi-
neering applications involving modifications of these path-
ways (Cohen and Schulten 2007). Previous studies have also
consistently shown that gas pathways primarily originate
from localized thermal fluctuations of protein structure.
Hence, these pathways are transient, and permanent gas
tunnels cannot be detected in static crystal structures in most
cases. It should be noted that migration pathways for small
gas molecules inside proteins should be both geometrically
and energetically favorable. Yet most previous studies have
specifically focused on the energy aspect. A small body of
work has investigated the dynamic cavities along the
migration pathways (van den Bosch et al. 2004; Daigle et al.
2009).
In this study, we first report an all-atom MD simulation
on the full length of solvated 2,3-HPCD complexed with
4NC as substrates. Then, the energetically favorable O
2
channels in the four subunits are identified on the basis of
implicit ligand sampling (Cohen et al. 2006,2008).
In combination with possible tunnels from cavities detected
using CAVER software (Petr
ˇek et al. 2006; Damborsky
´
et al. 2007), we found four different O
2
migration pathways
for four identical subunits. Our results provide pertinent
implications for the different catechol dioxygenase reac-
tivity of a single homotetrameric enzyme.
Materials and methods
Force-field parameters for the active site residues
The starting coordinates for simulations were built from the
0.19 nm resolution refined X-ray crystal structure (PDB
entry: 2IGA) (Berman et al. 2000; Kovaleva and Lipscomb
2007). This protein is a homotetrameric structure in which
each subunit contains an active site. In particular, four
subunits (A–D) were bound by product, alkylperoxo,
superoxo, and alkylperoxo intermediates, respectively. The
coordinate chosen for the Fe
2?
site is the five-coordinate
(Solomon et al. 2000,2003) with the conserved 2-His-
1-carboxylate motif (His155, His214, and Glu267) and one
doubly deprotonated 4NC (Kovaleva and Lipscomb 2007)
(Fig. 1). Following previous experiments and computational
Fig. 1 The model structure used to calculate force field parameters
328 Eur Biophys J (2010) 39:327–336
123

work, the His, His, and Glu motif of the active site was
modeled with two imidazoles and one acetate (Bugg 2003;
Deeth and Bugg 2003; Costas et al. 2004), and their initial
orientations were taken from the crystal structure. The initial
orientation of 4NC was chosen as the result of superposition
of 4NC with the three intermediates found in the crystal
structure. The CHARMM force field parameter set of the
model structure, and the partial atomic charges were cal-
culated by using Paratool software, a plugin included with
VMD software (Humphrey et al. 1996). Further computa-
tional details are given in the Supplementary Material. The
resulting atomic charges are also given in Table S1 of the
Supplementary Material.
MD simulations
The substrates in four subunits of the crystal structure (PDB
entry: 2IGA) were replaced by four 4NC dianions. Two
methods, i.e., structure superimposition and Autodock 4.0
(Morris et al. 1998), were used to determine the binding
mode of 4NC in each of the binding sites of the protein. It is
noted that three Fe-bound ligands, (i.e., substrate 4NC, the
alkylperoxo intermediate, and the ring-open product) can be
well superposed (Kovaleva and Lipscomb 2007); therefore,
the original orientation of the 4NC in each subunit adopted
in the MD simulation could be the same for both methods.
Hydrogen atoms were added by using the PSFGEN
plugin of VMD (Humphrey et al. 1996). All His residues
were protonated at epsilon-N except His155 and His214 in
each chain, because they were coordinated to the Fe
2?
via
epsilon-N.
All MD simulations were performed by use of the
software package NAMD 2.6 (Phillips et al. 2005) with the
CHARMM27 force field parameter set (MacKerell et al.
1998). The structure with 4NC bound in each subunit,
together with crystal water molecules, was solvated to form
a10911.8 912.3-nm simulation water box by using the
Solvate plugin of VMD. The resulting solvated system was
neutralized by randomly adding Na
?
ions in the bulk water
by using the Autoionize plugin of VMD. The final system
contains a total of 136,234 atoms, including the model
protein, 37,825 water molecules and the added counterions.
First, the system was energy minimized for 10,000 steps,
keeping protein, 4NC, and crystal water molecules fixed.
Subsequently, the system was heated from 0 K to 300 K
with heavy atoms restrained by using a spring with a var-
iable constant from 20 kcal/mol/A
˚
2
to 10, 5, 2.5, 1.2, 0.6,
and 0.1, with steps of 20 ps during first six stages and until
0 kcal/mol/A
˚
2
. Finally, the system was equilibrated for
100 ps and the simulation then lasted for another 20 ns at
constant pressure and constant temperature (NPT).
Langevin dynamics and the Langevin piston method were
used to maintain the temperature at 300 K and the pressure
at 1 bar. Particle mesh Ewald (PME) was used for long-
range electrostatics. The multiple time steps of 2, 2, and
4 fs were used for bonded, nonbonded, and long-rang
electrostatics interactions, respectively. Trajectory was
saved every 1 ps, resulting in 20,000 frames from the 20-ns
simulation. Then the whole trajectory was separated into
two 10-ns trajectories and used in subsequent implicit
ligand sampling and cavity identification.
Implicit ligand sampling
The implicit ligand sampling method (Cohen et al. 2006,
2008) enables computation of 3D free energy or potential of
mean force (PMF) map for a gas ligand at any location in a
protein. As such, the interaction energy of a gas molecule
with its surroundings is taken into account. This method has
been implemented in VMD. Here, we apply this technique to
our system. One O
2
was placed in every trajectory frame on a
finely-spaced 3D grid. Fifty conformations of O
2
were
sampled on each 0.1 nm
3
voxel. Recommended CHARMM
parameters of O
2
were used. Averaging over 10,000 frames
from each of the 10-ns trajectories, the free energy at every
grid position was obtained. Migration pathways of O
2
could
be constructed by connecting the energetically favorable
areas of the maps.
Identification of tunnels
The software CAVER (Petr
ˇek et al. 2006; Damborsky
´et al.
2007) was applied to each of the 10-ns trajectories, deter-
mining hundreds to thousands of tunnels for each monomer
of the protein. The energy profile along each tunnel was then
constructed by superposing every tunnel profile on to the 3D
free energy map generated using implicit ligand sampling.
The most likely routes for O
2
diffusion from the protein
surface to the active site were thus identified by searching
for the lowest-energy tunnels. Following this procedure for
all detected tunnels allowed us to find the energetically
preferred tunnel in the free energy map for each subunit.
Residues lining the tunnels that could play key roles in
controlling the passage of O
2
were also extracted.
All calculations, including MD simulations were per-
formed on the Lenovo clusters DeepComp 1,800 at the
Department of Engineering Mechanics of Dalian University
of Technology.
Results and discussion
Equilibrium properties of the system
It seems reasonable to replace the original Fe-bound ligand
for a 4NC in each monomer of 2,3-HPCD simultaneously
Eur Biophys J (2010) 39:327–336 329
123

as previous experimental studies have observed a similar
phenomenon (Kita et al. 1999). Comparing the structure of
the four subunits after minimization and equilibration (Fig.
S2 in the Supplementary Material) provides direct evidence
of a\ similar conformation. The non-bonded distances of
interacting atoms at the active site of X-ray and MD-
averaged structures are summarized in Table S2 of the
Supplementary Material. As can be seen, the interactions of
Fe
2?
and His155, His214, Glu267, and 4NC calculated
from the MD simulation are consistent with those observed
in the X-ray structure (PDB entry 2IGA), indicating that
the coordination environment of Fe
2?
has not been per-
turbed during the 20-ns MD simulation (Table S2).
The structural stability of each monomer is evaluated on
the basis of the root-mean-square deviation (RMSD) of
each residue with regard to the crystal structure, as shown
in Fig. 2. The modest deviations from the reference crystal
structure suggest that each subunit is stable during the
time-course of 20-ns MD simulation. However, the termi-
nal residues of 350–364 fluctuate much within the second
10 ns. To compare the effect of structural difference on the
O
2
pathway, the whole trajectory was divided into two
parts and analyzed separately.
O
2
migration pathways in four monomers
The implicit ligand PMF map for O
2
inside 2,3-HPCD
bound with 4NC is shown in Fig. 3. In water, the free
energy is uniform (8.2 kJ/mol). All areas of the protein for
which the PMF value is less than 8.2 kJ/mol are areas in
which it is more likely to find an O
2
molecule than in water
phase of identical volume. In each monomer we identify
one major region that contains the most likely route for O
2
diffusion from protein surface to the active center. The
PMF map shows equally favorable O
2
-accessible regions
inside each monomer. The pathway for O
2
movement with
the lowest barrier starts at the free coordinate site (space
left for O
2
) of Fe, close to both of the catecholate positions,
and reaches to the protein–solvent boundary adjacent to
Trp304, Tyr305, and Thr205 amino acids. Moreover, the
regions with high probability of O
2
occupation are quite
wide. Although we cannot anticipate from the PMF map
that O
2
takes side-on binding rather than end-on orientation
when coordinated to Fe (Kovaleva and Lipscomb 2007),
the wide high-affinity area near the active site should favor
formation of a side-on complex with iron. The C-terminal
residues 323–362 were thought to form a ‘‘lid’’ over the
substrate-binding pocket (Kovaleva and Lipscomb 2007).
Here, Trp304, Tyr305, and Thr205 were found to line
along the O
2
pathway, suggesting that O
2
could not share
the same migration pathway with the substrate 4NC. For
12/15-lipoxygenase, copper amine oxidase, and cholesterol
oxidase, previous studies have also suggested that dioxy-
gen channels and substrate pathways are different (Saam
et al. 2007). In contrast, the route for substrate entrance that
also efficiently functions as O
2
access pathway has been
identified for cyclooxygenases (Saam et al. 2007).
It should be noted that the PMF values presented here
are statistical averages sampled from 10,000 conforma-
tions. The pathways by which O
2
enters the protein are
Fig. 2 Root-mean-square
deviations (RMSD) of each
residue with regard to the X-ray
structure in the first (black) and
second (red) 10-ns MD
simulations
330 Eur Biophys J (2010) 39:327–336
123

assumed to result from diffusion through well-defined
regions of the protein made possible by the protein’s
thermal fluctuation, rather than migration along permanent
channels (Cohen et al. 2006; Cohen and Schulten 2007;
Johnson et al. 2007; Saam et al. 2007). Therefore, the
pathway identified by implicit ligand sampling resembles a
chain of separate cavities that are transiently intercon-
nected. In order to find subtle differences along O
2
path-
ways in each monomer, cavities were characterized by use
of CAVER software (Petr
ˇek et al. 2006; Damborsky
´et al.
2007) based on two 10-ns MD simulations.
As suggested, cavities are short-lived and can hop inside
the protein matrix, whereas elongated cavities that connect
the active site to the solvent are rarely found. Such cavities,
hereafter referred to as tunnels, are still geometrically
favorable gas pathways. The lowest barrier tunnel profiles
for each subunit, i.e., the tunnel radii and the energy values
along the tunnels, are given in Figs. 4and 5(also see Figs.
S3 and S4 for the second 10-ns result). The corresponding
position of Fe was used as the starting point to calculate the
distance but the first five points were not kept, because of
the extremely large PMF values within such a narrow
region, as suggested by the CAVER software (Petr
ˇek et al.
2006; Damborsky
´et al. 2007). The tunnel profiles (Figs. 4
and S3) clearly show that the tunnel in subunit D is
Fig. 3 Implicit ligand PMF
maps for O
2
migration pathways
inside the four subunits (A–D)
of the protein, based on the first
10-ns MD simulation of the
2IGA PDB structure in the
presence of one 4NC dianion in
the active site of each subunit.
Three free-energy isosurfaces
with energy values of 0 kJ/mol
(orange), 7.5 kJ/mol (gray), and
12.5 kJ/mol (pink) are
superimposed. The Fe-bound
ligands, i.e., His155, His214,
Glu267, and the 4NC dianion
are represented as a CPK model.
The green ball represents the Fe
atom. The static surface of each
monomer is displayed as a
transparent violet surf model.
The yellow van der Waals
spheres indicate the most
favorable tunnels
Fig. 4 Optimal tunnel profiles for subunits A–D of 2,3-HPCD
dioxygenase with 4NC, extending from the position of Fe to the
bulk solvent. Only the result of the first 10-ns MD simulation is given.
The result of the second 10-ns MD simulation is given in Fig. S3
Eur Biophys J (2010) 39:327–336 331
123

generally broader than the other three tunnels. Interest-
ingly, tunnel D was observed to have two branches (tunnels
D1 and D2) at the location close to the active center. On the
other hand, the radii of the tunnel in subunit C are generally
narrower, particularly in the position located approximately
1.25 nm from the iron. Another feature of the tunnel
profiles is that there are two broader regions along the four
tunnels. One lies at approximately 0.5 nm, in close prox-
imity to the catalytic center Fe atom; another covers
1–1.3 nm. The latter wide area may serve as the anteroom
to store O
2
, because other studies have also indicated the
presence of such region inside myoglobin (Cohen et al.
2006), copper-containing amine oxidase (Johnson et al.
2007), and 12/15-lipoxygenase (Saam et al. 2007). The
interaction between O
2
and other Fe-bound ligands reduces
the tunnel radii approximately 0.3 nm from the origin
(Fig. 4).
The energy profiles along these tunnels are somewhat
surprising, as shown in Figs. 5and S4. Two local minima
of energy within 0.5–1.2 nm of the starting position were
identified, corresponding to the broader region as men-
tioned earlier and as shown in Fig. 3. One minimum lies at
approximately 0.5–0.7 nm for all of these tunnels, and
another lies approximately 1.0–1.2 nm from the iron
except for tunnel C. The unexpectedly high PMF values in
this region of tunnel C suggest that the chance of finding
O
2
is far lower than in the same region in the other sub-
units; in particular, the probability of finding O
2
in this
region is much lower than in an equivalent volume of
solvent, because the PMF values are above 0 kJ/mol.
Recall the PMF map as shown in Fig. 3, the tunnels (rep-
resented as yellow van der Waals spheres) usually cross the
low-energy isosurface, indicating these tunnels display
energetic preferences. Nevertheless, the central portion of
the tunnel in subunit C apparently deviates from the low-
energy area. This result also suggests that diffusion of
oxygen is more difficult in subunit C than in other subunits.
In the case of tunnel A, it can be easily accessed by O
2
because this route has the lowest energy barrier in the
region close to the active center and moderately low energy
barriers along this pathway. Although a local minimum is
located at approximately 1.0 nm in tunnel B, it seems
difficult for O
2
to overcome the energy barrier and migrate
into the active center. This energy-favorable basin may act
as the anteroom for O
2
diffusion. Tunnel D also possesses a
wide energy-minimum region but much closer to the iron.
O
2
diffusion seems much easier in this tunnel, because the
radii of cavities along this path are much larger than in the
others (Fig. 4). However, once O
2
migrates into the above
basin, it can be easily trapped there, because of the high
barrier around the basin. We also note that the bottleneck
radii of one branch, i.e. tunnel D1, are much smaller than
those of the other branch, tunnel D2 (Fig. 4), and the
corresponding PMF values are much lower than for tunnel
D2. As a consequence, tunnel D is characterized as a
geometrically favorable but energetically unfavorable O
2
diffusion pathway. However, after 10-ns structural relaxa-
tion, no branch is observed in tunnel D (Fig. S3).
An additional feature of Fig. 5, e.g., at the beginning of
pathway (\0.5 nm), is that the radii of the tunnel become
so narrow that collision of O
2
with the protein is more
frequent and, consequently, the interaction energy is high.
The energy profiles indicate that the O
2
pathways in the
same four subunit are characterized by different binding
potential for O
2
(A [B, D [C) and relate well to the fact
that the catalytic product is found in subunit A, the unstable
alkylperoxo intermediate in subunits B and D, and the
highly unstable superoxide intermediate in subunit C
(Kovaleva and Lipscomb 2007).
Recent experimental studies have revealed that the
diameter of the tunnel’s bottleneck partially correlates with
the observed rates of CO and H
2
diffusion in hydrogenase
(Leroux et al. 2008). In our analysis of the lowest-energy
tunnels, the most important bottleneck radii approximately
0.38 nm from the iron are 0.11, 0.060, 0.042, and 0.095 nm
for tunnels A, B, C and D, respectively. This radius at
0.38 nm was used, because it is common to all tunnels and
located outside the volume occupied by the ligand (Daigle
et al. 2009). Interestingly, this order is accord with the
order of oxygen affinity of the four subunits. The bottle-
neck of each tunnel provides more evidence for the dif-
ferent rate of O
2
transport in each subunit. As a result, the
rate of catalytic reaction could be affected accordingly.
It has been realized that implicit ligand sampling can
only describe the gas diffusion pathways defined by the
protein’s thermal motion. Slow conformational changes
Fig. 5 Energy profiles for O
2
tunnels in subunit A–D, relative to the
solvent (8.2 kJ/mol). Only the result of the first 10-ns MD simulation
is given. The result of the second 10-ns MD simulation is given in
Fig. S4
332 Eur Biophys J (2010) 39:327–336
123

such as the opening of a channel gate are beyond the
simulation time scales (Cohen et al. 2008). The PMF map
constructed from the current 10-ns MD simulation should
be adequate to allow us to draw reliable conclusions,
because the conformational changes among the four sub-
units are not significant in the first 10-ns MD simulation.
The electrostatic effect on the PMF map is omitted in
this present study, as is usually done (Cohen et al. 2008).
However, it may be of interest to investigate the electro-
static potential along these tunnels (Fig. S5 in Supple-
mentary Material). Electrostatic analysis of the trajectory
taken at 10 ns reveals the presence of positive electrostatic
potential around the four tunnels that seems to affect O
2
diffusion, binding, and activation. For example, the oxygen
activation and insertion steps are sensitive to the protein
environment, thus the positive electrostatic potential could
stabilize the superoxo intermediate.
Inspecting the tunnel-surrounding residues, we find that
the first local energy minimum located at 0.5–0.7 nm from
the iron is formed by the residues bound to the iron, i.e.,
His155, His 214, Glu267, and the ligand 4NC. The second
minimum centered at approximately 1.0 nm is between
Arg293, His213, Trp304, Tyr305, and Thr205 just below
the protein surface. However, the same region in tunnel C
cannot be easily accessed by O
2
, as mentioned above.
To assess the effect of side-chain flexibility of lining
residues on tunnel formation, the v
1
(N–C
a
–C
b
–C
c
) and v
2
(C
a
–C
b
–C
c
–C
d1
) dihedral angles of these amino acids were
examined. It was found that except for Arg293, especially
the v
2
angle, the side-chain conformations of other amino
acids are very similar (data not shown). The v
2
angles of
Arg293 were extracted from the 20-ns MD simulation and
are plotted in Fig. 6. This shows that the side-chain of the
Arg293 in subunit A is in the extended conformation
(v
2
=-162.4°) for most of the time whereas in the other
subunits Arg293 prefers to be in twisted conformation
(v
2
=-42.1°). The collective movements of these amino
acids were also examined by calculating the distance
between different residues. It was found that the cavity
composed by Thr205, His213, and Trp304 undergoes the
so-called breathing motion (Tomita et al. 2009), as shown
in Fig. 7. The expansion and contraction of the cavity may
play specific role in controlling the passage of O
2
.This
result suggests that the increase of the distances between
these residues concomitantly causes expansion of the
cavity, leading to opening of the O
2
pathway. It clearly
shows that the frequency of the breathing motion of these
residues in subunit C is far less than those in the other
subunits, indicating the smallest chance of O
2
access to the
active site of subunit C.
With regard to the relevant residues lining the tunnels,
as shown in Fig. 8,O
2
might also interact with Asn157 and
His213 when migrating to the iron in tunnel A. The addi-
tional interaction between O
2
and the aromatic residue
Trp192 in tunnel B could increase the energy barrier in the
region adjacent to the binding site. Tunnel D is generally
Fig. 6 Arg293 (R293) v
2
as
function of time for subunits
A–D in the 20-ns MD
simulation
Eur Biophys J (2010) 39:327–336 333
123

broader (Fig. 4) and the fluctuation of Arg293 (Fig. 6d)
along with His213 could create an alternative O
2
diffusion
pathway (tunnel D2) when approaching the active center.
It has been confirmed by experimental approaches that
the rate of inhibition could be limited by CO or H
2
binding
at the active sites in hydrogenase (Leroux et al. 2008).
In particular, slowing of the transport of H
2
,D
2
, and HD in
the channel reduces the probability that the intermediate
produced, HD, exits the enzyme before it further reacts to
give H
2
(Leroux et al. 2008). This fact shows that the
simultaneous presence of three different intermediates in a
simple enzyme with four independent active sites could
partially be attributed to the distinct O
2
pathways in 2,3-
HPCD. Subunit A contains the most preferred O
2
pathway
and, accordingly, the final product is found there. In the
case of subunit C, the O
2
pathway is the least favorable and
therefore, the highly unstable superoxo intermediate should
be found with the highest probability. Subunits B and D
Fig. 7 Top snapshots of the
expansion and contraction of the
cavity (length in angstrom, A
˚).
Bottom, distance between
Thr205 C
c2
and His213 C
d2
(black), Thr205 C
c2
and Trp304
C
h2
(red), His213 C
d2
and
Trp304 C
h2
(blue) observed in
the 20-ns MD simulation
334 Eur Biophys J (2010) 39:327–336
123

contain different O
2
tunnels and the alkylperoxo interme-
diate was found to be simultaneously present in them
(Kovaleva and Lipscomb 2007). Moreover, it should be
noted that the rate of catalysis is not only limited by O
2
access to the active site but also limited by its reaction at
the active site (Leroux et al. 2008). In this way, charac-
terization of O
2
pathways alone could not be expected to
explain all of the experimental outcomes. Recent experi-
ments have shown that use of an alternative substrate,
4-sulfonylcatechol, and the Glu323Leu variant of HPCD,
have led to formation of a new intermediate that seems to
occur between the alkylperoxo intermediate and the
product complexes in the reaction cycle (Kovaleva and
Lipscomb 2008).
Conclusions
To address the question of the different enzyme reactivity
of each monomer in a homotetrameric 2,3-HPCD, implicit
ligand sampling and tunnel detection were conducted on
the trajectory of a 20-ns molecular dynamics simulation.
Specific O
2
migration tunnels that link the bulk solvent and
the cavity of the enzyme have been probed. In contrast with
randomized behavior, O
2
diffusion in the four subunits is
clearly limited to specific regions located within the con-
served active site domain. A major finding of our study is
that nominally identical subunits have distinct O
2
affinity
tunnels. In connection with recent experimental results for
NiFe hydrogenase, the outcome of our computations sug-
gests that the rate of dioxygenase-catalyzed reaction is
generally correlated with the rate of access of O
2
to the
active center. The determination of the dynamic behavior
of some key residues should aid understanding of the
exquisite control over O
2
binding for each subunit. To this
end, this work described herein will be valuable for engi-
neering non-heme iron dioxygenases to find intermediates
by charactering O
2
migration pathways.
Acknowledgments This work was supported by the Youth Founda-
tion of DLUT (893103), National Natural Science Foundation
(10772042),and the National Basic Research Program (2009CB918501)
of China.
References
Abu-Omar MM, Loaiza A, Hontzeas N (2005) Reaction mechanisms
of mononuclear non-heme iron oxygenases. Chem Rev
105:2227–2252
Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H,
Shindyalov IN, Bourne PE (2000) The protein data bank.
Nucleic Acids Res 28:235–242
Bugg TDH (2003) Dioxygenase enzymes: catalytic mechanisms and
chemical models. Tetrahedron 59:7075–7101
Bugg TDH, Lin G (2001) Solving the riddle of the intradiol and
extradiol catechol dioxygenases: how do enzymes control
hydroperoxide rearrangements? Chem Commun 941–952
Fig. 8 Snapshots of optimal routes for O
2
movement in the four
monomers (A–D). The relevant residues lining each tunnel are shown
in licorice representation. The Fe-bound ligands are shown in CPK
representation. The tunnels are represented as gray solid surface
Eur Biophys J (2010) 39:327–336 335
123

Cohen J, Schulten K (2007) O
2
migration pathways are not conserved
across proteins of a similar fold. Biophys J 93:3591–3600
Cohen J, Kim K, King P, Seibert M, Schulten K (2005) Finding gas
diffusion pathways in proteins: application to O
2
and H
2
transport in CpI [FeFe]-hydrogenase and the role of packing
defects. Structure 13:1321–1329
Cohen J, Arkhipov A, Braun R, Schulten K (2006) Imaging the
migration pathways for O
2
, CO, NO, and Xe inside myoglobin.
Biophys J 91:1844–1857
Cohen J, Olsen KW, Schulten K (2008) Finding gas migration
pathways in proteins using implicit ligand sampling. Methods
Enzymol 437:439–457
Costas M, Mehn MP, Jensen MP, Que LJr (2004) Dioxygen activation
at mononuclear nonheme iron active sites: enzymes, models, and
intermediates. Chem Rev 104:939–986
Daigle R, Guertin M, Lagu
¨e P (2009) Structural characterization of
the tunnels of Mycobacterium tuberculosis truncated hemoglobin
N from molecular dynamics simulations. Proteins 75:735–747
Damborsky
´J, Petr
ˇek M, Bana
´s
ˇP, Otyepka M (2007) Identification of
tunnels in proteins, nucleic acids, inorganic materials and
molecular ensembles. Biotechnol J 2:62–67
Deeth RJ, Bugg TDH (2003) A density functional investigation of the
extradiol cleavage mechanism in non-heme iron catechol
dioxygenases. J Biol Inorg Chem 8:409–418
Fiorucci S, Golebiowski J, Cabrol-Bass D, Antonczak S (2006)
Molecular simulations reveal a new entry site in quercetin 2,3-
dioxygenase. A pathway for dioxygen? Proteins 64:845–850
Georgiev V, Borowski T, Blomberg MRA, Siegbahn PEM (2008)
A comparison of the reaction mechanism of iron- and manga-
nese-containing 2,3-HPCD: an important spin transition for
manganese. J Biol Inorg Chem 13:929–940
Humphrey W, Dalke A, Schulten K (1996) VMD—visual molecular
dynamics. J Mol Graph 14:33–38
Johnson BJ, Cohen J, Welford RW, Pearson AR, Schulten K,
Klinman JP, Wilmot CM (2007) Exploring molecular oxygen
pathways in Hansenula polymorpha copper-containing amine
oxidase. J Biol Chem 282:17767–17776
Kita A, Kita S, Fujisawa I, Inaka K, Ishida T, Horiike K, Nozaki M,
Miki K (1999) An archetypical extradiol-cleaving catecholic
dioxygenase: the crystal structure of catechol 2,3-dioxygenase
(metapyrocatechase) from Pseudomonas putida mt-2. Struc-
ture 17:25–34
Kovaleva EG, Lipscomb JD (2007) Crystal structures of Fe
2?
dioxygenase superoxo, alkylperoxo, and bound product inter-
mediates. Science 316:453–457
Kovaleva EG, Lipscomb JD (2008) Intermediate in the O–O bond
cleavage reaction of an extradiol dioxygenase. Biochemistry
47:11168–11170
Kovaleva EG, Neibergall MB, Chakrabarty S, Lipscomb JD (2007)
Finding intermediates in the O
2
activation pathways of non-heme
iron oxygenases. Acc Chem Res 40:475–483
Kryatov SV, Rybak-Akimova EV (2005) Kinetics and mechanisms of
formation and reactivity of non-heme iron oxygen intermediates.
Chem Rev 105:2175–2226
Leroux F, Dementin S, Burlat B, Cournac L, Volbeda A, Champ S,
Martin L, Guigliarelli B, Bertrand P, Fontecilla-Camps J,
Rousset M, Le
´ger C (2008) Experimental approaches to kinetics
of gas diffusion in hydrogenase. Proc Natl Acad Sci USA
105:11188–11193
Lipscomb JD (2008) Mechanism of extradiol aromatic ring-cleaving
dioxygenases. Curr Opin Struct Biol 18:644–649
MacKerell AD, Bashford D, Bellott M, Dunbrack RL, Evanseck JD,
Field MJ, Fischer S, Gao J, Guo H, Ha S, Joseph-McCarthy D,
Kuchnir L, Kuczera K, Lau FTK, Mattos C, Michnick S, Ngo T,
Nguyen DT, Prodhom B, Reiher WE, Roux B, Schlenkrich M,
Smith JC, Stote R, Straub J, Watanabe M, Wiorkiewicz-Kuczera
J, Yin D, Karplus M (1998) All-atom empirical potential for
molecular modeling and dynamics studies of proteins. J Phys
Chem B 102:3586–3616
Morris GM, Goodsell DS, Halliday RS, Huey R, Hart WE, Belew RK,
Olson AJ (1998) Automated docking using a lamarckian genetic
algorithm and empirical binding free energy function. J Comput
Chem 19:1639–1662
Orlowski S, Nowak W (2007) Oxygen diffusion in minihemoglobin
from Cerebratulus lacteus: a locally enhanced sampling study.
Theor Chem Acc 117:253–258
Petr
ˇek M, Otyepka M, Bana
´s
ˇP, Kos
ˇinova
´P, Koc
ˇa J, Damborsky
´J
(2006) CAVER: a new tool to explore routes from protein clefts,
pockets and cavities. BMC Bioinformatics 7:316
Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, Villa E,
Chipot C, Skeel RD, Kale L, Schulten K (2005) Scalable
molecular dynamics with NAMD. J Comput Chem 26:
1781–1802
Ruscio JZ, Kumar D, Shukla M, Prisant MG, Murali TM, Onufriev
AV (2008) Atomic level computational identification of ligand
migration pathways between solvent and binding site in myo-
globin. Proc Natl Acad Sci USA 105:9204–9209
Saam J, Ivanov I, Walther M, Holzhu
¨tter HG, Kuhn H (2007)
Molecular dioxygen enters the active site of 12/15-lipoxygenase
via dynamic oxygen access channels. Proc Natl Acad Sci USA
104:13319–13324
Siegbahn PEM, Haeffner F (2004) Mechanism for catechol ring-
cleavage by non-heme iron extradiol dioxygenases. J Am Chem
Soc 126:8919–8932
Solomon EI, Brunold TC, Davis MI, Kemsley JN, Lee SK, Lehnert N,
Neese F, Skulan AJ, Yang YS, Zhou J (2000) Geometric and
electronic structure/function correlations in non-heme iron
enzymes. Chem Rev 100:235–349
Solomon EI, Decker A, Lehnert N (2003) Non-heme iron enzymes:
contrasts to heme catalysis. Proc Natl Acad Sci USA 100:
3589–3594
Teeter MM (2004) Myoglobin cavities provide interior ligand
pathway. Protein Sci 13:313–318
Teixeira VH, Baptista AM, Soares CM (2006) Pathways of H
2
toward
the active site of [NiFe]-hydrogenase. Biophys J 91:2035–2045
Tomita A, Sato T, Ichiyanagi K, Nozawa S, Ichikawa H, Chollet M,
Kawai F, Park SY, Tsuduki T, Yamato T, Koshihara SY, Adachi
S (2009) Visualizing breathing motion of internal cavities in
concert with ligand migration in myoglobin. Proc Natl Acad Sci
USA 106:2612–2616
van den Bosch M, Swart M, van Gunsteren WF, Canters GW (2004)
Simulation of the substrate cavity dynamics of quercetinase.
J Mol Biol 344:725–738
Wallar BJ, Lipscomb JD (1996) Dioxygen activation by enzymes
containing binuclear non-heme iron clusters. Chem Rev
96:2625–2657
336 Eur Biophys J (2010) 39:327–336
123