Professional Documents
Culture Documents
The conformational flexibility of the tetrapyrrolic phytochromobilin (PΦB) chromophore of the bacterio-
phytochrome Deinococcus radiodurans (DrCBD) in the Pr state has been investigated by molecular dynamics
simulations. Because these simulations require accurate force field parameters for the prosthetic group, in the
present work we developed new empirical force field parameters for the PΦB molecule that are compatible
with the CHARMM22 force field for proteins. For this reason, the new force field parameters for the nonbonded
(partial atomic charges) and bonded (bonds, angles, dihedrals, improper) energy terms were derived by
reproducing ab initio target data following the methodology used in the development of the CHARMM22
force field. This new set of parameters was employed to analyze structural and dynamical features of PΦB
inside DrCBD. The 45 ns all-atom molecular dynamics (MD) simulation reveals the existence of two stable
conformational states of the chromophore characterized by distinct torsional angles around the C-C bond at
the methine bridge connecting rings A and B of the tetrapyrrole. This result supports experimental observations
derived from NMR and resonance Raman spectroscopy. Furthermore, statistical analysis of H-bonding events
allowed us to identify (a) important H-bonds between the propionic side chains of the chromophore and the
apoprotein which may be relevant for the signal transduction step during the photoinduced cycle and (b) a
network of eight water molecules which remain in the vicinity of the chromophore during the entire 45 ns
production run.
Figure 1. 3-D structure of DrCBD (left) and structure of model compound 2(R),3(E)-PΦB and fragmentation scheme for optimization of torsion
angle parameters (right). New atom types are written in bold letters.
the structural and dynamical properties of the entire phyto- Ab initio calculations required for the generation of target
chrome protein is the development of adequate force field data were performed with the Gaussian 03 program,14 whereas
parameters for the bilin chromophore which are compatible with molecular mechanics calculations and molecular dynamics
the already existing CHARMM all-atom force field.10,11 simulations were done using the CHARMM15 version 32b2 and
Starting from the available X-ray structure of DrCBD in the NAMD 2.616 codes.
Pr state, we employed empirical molecular dynamics to identify 2.1. Definition of the Potential Energy Function. The
and characterize possible conformers of the PΦB chromophore, CHARMM potential energy function is defined as the sum of
which may be relevant for the interpretation of the experimental various internal (bonded) and nonbonded terms in the following
data.6,7 The present paper is separated in two parts: the first way:
part (section 2) describes the development and validation of
new force field parameters for the PΦB chromophore, whereas
the second part (section 3) reports the results of molecular U(rj) ) ∑ Kb(b - b0)2 + ∑ KUB(S - S0)2 +
dynamic simulation of DrCBD in solution. Special emphasis is bonds UB
given in elucidating the conformational flexibility of the
chromophore and the stability of the hydrogen bond and water
∑ Kθ(θ - θ0)2 + ∑ Kχ(1 + cos(nχ - δ)) +
[( ) ( ) ]
angle dihedral
network in the protein binding pocket during the MD simulation. Rij 12 Rij 6
∑ Kφ(φ - φ0)2 + ∑ εij
rij
-2
rij
+
improper nonbond
2. Development and Validation of Force Field qiqj
Parameters (1)
4πεrij
The optimization and evaluation of new force field parameters
for phytochromobilin was performed following the iterative
procedure used for developing the CHARMM22 force field.10 where K denotes force constants for bonds (b), Urey-Bradley
This procedure is based on the reproduction of target data at a 1,3-distances (S), bond angles (θ), dihedral angles (χ) and
molecular mechanics level. The target data include structural improper dihedral angles (φ) and b0, S0, θ0, χ0 and φ0 are the
data, energies (e.g., rotational barriers) and intermolecular corresponding equilibrium values of the individual terms. The
interaction energies of model compounds estimated from ab nonbonded interactions are described as the sum of Lennard-
initio calculations.10,12,13The model compounds are simply Jones 6-12 and Coulomb terms. The first depends on the
isolated molecules containing the structural units (bond lengths, Lennard-Jones parameters, εij and Rij, for the atom pair i and j
bond angles and dihedral angles) associated with those param- and the distance rij between them. The Coulomb terms are a
eters that are to be optimized. The parametrization strategy is function of the atomic partial charges qi and qj and their relative
then to maximize the agreement between the target data and distance rij.
the corresponding molecular properties of the model compounds 2.2. Model Compounds. The model compound for the
computed using the empirical force field. protein-bound chromophore consists of an isolated protonated
The use of this parametrization approach guaranties consis- 2(R),3(E)-phytochromobilin (PΦB) molecule in a ZZZssa
tency with the available force fields CHARMM2210 and conformation/configuration of the methine bridges as found in
CHARMM2711 employed to model the phytochrome apoprotein. the X-ray structure of DrCBD shown in Figure 1. The propionic
Details about the parametrization procedure are described in side chains were protonated (neutral) and the linkage of the
the paragraphs below. chromophore to the protein was saturated by a hydrogen atom.
MD Simulations of the Chromophore Binding Site of DrCBD J. Phys. Chem. B, Vol. 113, No. 4, 2009 947
TABLE 1: Atom Types, Labels and Optimized Partial Charges of the PΦB Moleculea
atom atom type partial charge atom atom type partial charge atom atom type partial charge
CAC CPM -0.082 C3B CPY6 -0.091 HO3 HA 0.09
HAC HA 0.123 O_B O -0.407 HO4 HA 0.09
C1C C 0.289 CHB CPY2 -0.448 CGD CC 0.62
H2C HA 0.057 HHB HA 0.219 O2D OC -0.76
N_C NR1 -0.555 CAB CE1 -0.332 O1D OC -0.76
H_C H 0.332 HAB HE1 0.307 CBA CT2 -0.28
C4C CA 0.312 CBB CE2 -0.269 HO1 HA 0.09
C3C CPY1 -0.160 HV1 HE1 0.189 HO2 HA 0.09
C2C CT1 0.315 HV2 HE2 0.153 CGA CC 0.62
O_C O -0.445 CBC CT2 0.043 O2A OC -0.76
CHD CPY3 -0.501 HL1 HA 0.058 O1A OC -0.76
HHD HA 0.214 HL2 HA 0.058 CAA CT2 -0.18
C1D CPA 0.375 CMC CT3 -0.119 HO5 HA 0.09
N_D NR1 -0.543 HE1 HA 0.038 HO6 HA 0.09
H_D H 0.306 HE2 HA 0.038 CAD CT2 -0.18
C4D CPA 0.364 HE3 HA 0.038 HO7 HA 0.09
C3D CPB -0.091 CMD CT3 -0.024 HO8 HA 0.09
C2D CPB -0.319 HD1 HA 0.022
C1A CPA 0.119 HD2 HA 0.022
N_A NR1 -0.654 HD3 HA 0.022
H_A H 0.355 CMA CT3 -0.075
C2A CPY4 -0.038 HA1 HA 0.050
C3A CPB -0.155 HA2 HA 0.050
C4A CPA 0.487 HA3 HA 0.050
C4B C 0.476 CMB CT3 -0.018
N_B NR1 -0.560 HB1 HA 0.049
H_B H 0.425 HB2 HA 0.049
C1B CA 0.452 HB3 HA 0.049
C2B CPY5 -0.115 CBD CT2 -0.28
a
Values related to the propionate chains are written in italic letters. New atom types are in bold letters.
To account for electron delocalization effects, the entire PΦB previous section. All ab initio calculations were performed in
molecule was used for the optimization of the internal bond vacuo using Becke’s three parameters functional (B3LYP)18 and
lengths, bond angles and improper dihedral parameters and the the 6-31G(d) basis set.
nonbonded electrostatic parameters. Figure 1 (right) shows the 2.4. Nonbonded Parameter. The optimization of nonbonded
chromophore structure together with the atomic labels for force field parameters involves electrostatic and van der Waals
the heavy atoms. The list of all atoms belonging to the PΦB interactions. The main goal is to obtain a set of parameters that
chromophore and their corresponding atom type is given in can reproduce, in a reliable way, the water-solute interaction,
Table 1. Six new atom types were specifically created for the where water is described through a TIP3P model.17
PΦB molecule to be able to define new torsion potentials and Following MacKerell’s procedure,10 the partial atomic charges
consequently improve the accordance with the internal properties for the PΦB chromophore were obtained by computing, at a
of the target structure, as will be shown later. quantum mechanical level, minimum interaction energies and
For the evaluation of torsion parameters for the chromophore, distances between the model compound and water molecules.
the PΦB molecule was divided into smaller fragments (see We performed the calculations using fourteen water molecules
Figure 1 (right)) to reduce computational cost. The hexameth- located at various sites as shown in Figure 2. The PΦB geometry
ylpyrromethene (HMPM) and pentamethylpyrrolone (PMPO) was previously optimized in vacuo using the hybrid density
molecules were used to model the torsions at the methine bridge functional method B3LYP18 with the 6-31G(d) basis set whereas
between inner rings B and C and the methine bridges between the water molecules were built according to the TIP3P water
inner rings B/C and the outer ring A/D, respectively. The torsion model.17 The minimum interaction distances were estimated
parameters at the chromophore-protein linkage site were evalu- through single point energy calculations of the various
ated using the [(aminosulfanyl)oxo]pyrrolidine acid (ASOP) and PΦB-water complexes at the HF/6-31G(d) level of theory.
for the parameter associated with the vinyl function at ring D These complexes were built by systematically varying the
we used a methylenevinylpyrrolone (MVPO) molecule a as interaction distances in steps of 0.01 Å. Interaction energies
model compound. In this way all torsion angle parameters which were then estimated by subtracting the sum of the monomer
are relevant for the chromophore flexibility were accurately and energies from the energy of the entire complex. Because the
efficiently determined. chromophore is a charged compound, the target HF/6-31G(d)
2.3. Target Data. The target data for optimizing internal interaction energies where not scaled.10 Initial partial atomic
force field parameters such as bond, angle and improper torsion charges were obtained from fitting the electrostatic potential
parameters, was derived from the DFT optimized geometries using the CHELPG19 method implemented in Gaussian03 while
of the entire PΦB chromophore. The same structure has been simultaneously reproducing the overall dipole moment of the
used for the evaluation of partial charge. The use of a complete chromophore.
PΦB molecule as the target structure guaranties the correct The partial charges of atoms belonging to the propionate side
description of the electron delocalization effects. However, to chains were not optimized. These values were taken form the
reduce computational cost, the dihedral parameters were de- already existing parameters for the heme cofactor20 without
veloped using the molecular fragments mentioned in the further adjustment.
948 J. Phys. Chem. B, Vol. 113, No. 4, 2009 Kaminski et al.
Figure 3. Potential energy surfaces for partially constrained model compounds. The surfaces were scanned in steps of 10°. The black curves
denote the target data derived using ab initio methods (DFT/B3LYP/6-31G(d)), and the gray dashed-curves result from optimized torsion parameters
(see Table 2). Potential curves were shifted by setting the lowest energy conformations for each curve equal to zero.
performing constraint energy minimizations with a rms force analogue molecular mechanics calculations. In each step only
criterion of 10-4 for the ab initio calculations and 10-6 for the torsional angle in question was kept fixed to a certain value
950 J. Phys. Chem. B, Vol. 113, No. 4, 2009 Kaminski et al.
TABLE 3: Optimized Parameters for All Newly Defined Internal Coordinates (Except Torsions) of PΦB
bond Kb b0 bond Kb b0
parameter [(kcal/mol)/Å2] (Å) parameter [(kcal/mol)/Å2] (Å)
NR1-CPA 377.2 1.3817 CPY3-CA 360.0 1.3676
NR1-C 260.0 1.4190 CPY3-CPA 360.0 1.4225
CT2-CPM 230.0 1.4830 CPY4-CPA 299.8 1.4052
NR1-CA 377.2 1.4117 CPY4-CPB 340.7 1.4018
CPY1-CA 305.0 1.4772 CT2-CPY4 230.0 1.4946
CT1-CPY1 230.0 1.4907 CPY5-CA 305.0 1.4620
CPM-CPY1 360.0 1.3266 CT3-CPY5 230.0 1.4850
CPY2-CA 360.0 1.3800 CPY6-C 250.0 1.4711
HA-CPY2 367.6 1.0902 CPY6-CE1 450.0 1.4290
CPY2-CPA 360.0 1.4455 CPY6-CPY5 305.0 1.3470
angle Kθ θ0 angle Kθ θ0
parameter [(kcal/mol)/rad2] (deg) parameter [(kcal/mol)/rad2] (deg)
NR1-CPA-CPB 122.0 110.0 CPY1-CA-CPY3 61.6 127.57
CPA-NR1-CPA 139.3 116.3 CA-CPY3-CPA 94.2 122.80
CPM-CPA-NR1 88.0 131.8 CPY3-CPA-CPB 61.6 125.07
NR1-C-O 80.0 118.0 CPM-CPA-CPY4 61.6 132.50
HA-CT2-CPM 49.3 107.5 CPB-CPY4-CPA 30.8 136.01
C-NR1-CA 50.0 133.5 CPY4-CPB-CPA 30.8 145.01
HA-CPM-CPY1 12.7 117.44 CT2-CT2-CPY4 70.0 114.7
CT2-CPM-CPY1 45.8 117.49 NR1-CA-CPY5 122.0 112.4
NR1-CA-CPY1 122.0 115.0 CPY5-CA-CPY2 61.60 124.97
CT1-CPY1-CPM 45.8 116.6 CT3-CPY5-CA 45.8 115.9
CT1-CPY1-CA 45.8 128.0 NR1-C-CPY6 20.0 109.5
CPY1-CT1-C 52.0 113.7 C-CPY6-CPY5 52.0 116.0
CPM-CPY1-CA 61.6 124.1 C-CPY6-CE1 70.0 114.49
CPY2-CPA-NR1 88.0 112.39 CE2-CE1-CPY6 40.0 117.6
CPY2-CA-NR1 88.0 129.0 CA-CPY5-CPY6 40.0 116.5
CA-CPY2-CPA 94.2 127.0
improper Kφ φ0 improper Kφ φ0
parameter [(kcal/mol)/rad2] (deg) parameter [(kcal/mol)/rad2] (deg)
CPA-CPB-NR1-CPM 200.0 0.0 CPA-CPM-NR1CPY4 61.0 0.0
CPA-CPB-NR1CPY2 140.0 0.0 CPY4-CPA-CPB-CT2 64.0 0.0
CPA-CPB-NR1CPY3 145.0 0.0 CPY5-NR1-CA-CPY2 7.0 180.0
CPY1-NR1-CACPY3 140.0 180.0 CPY6-CE1-C-CPY5 180.0 0.0
while the rest of the structure was allowed to relax. Because of heights of rotational barrier as well as the overall shape of the
the chemical similarity between the outer rings A-B and C-D potential surfaces (Figure 3). Most problematic in this respect
(see Figure 3C) one set of target data (here the structure of ring was the potential describing single bond rotations of the linking
C-D was chosen) was sufficient to model the corresponding fragment to the protein (Figure 3f,g), because of the high
methine bridge torsions. Due to electron conjugation at the flexibility of this structure. Because we are not interested in
central methine bridge, the ab initio potentials for the two bond predicting Z/E isomerization energy barriers at double bonds,
torsions are practically identical. only a certain region of the potential energy curve around a
We now can summarize the iterative procedure for the local minimum was scanned. Within these limited dihedral-angle
parametrization of the nonbonded parameters. At first, partial intervals the torsion potentials were in general very well
charges and internal parameters (no torsions) have been derived reproduced. However, for the rotation around the CdC bond
for the entire chromophore. These parameters have been at the AB methine bridge (Figure 3a) although the overall
transferred to the smaller fragments to evaluate torsion param- curvature and local minimum at around 195° could be well
eters that were transferred back to the chromophore to refine predicted, the energy barrier and the second local minimum
the residual internal parameters and partial charges until self- slightly differ from the QM target.
consistency within all force field parameters for the chromophore 2.6. Validation of Force Field Parameters. Validation of
was achieved. The optimized bond length, bond angle and the final set of all internal and nonbonded parameters optimized
improper torsion parameters for the protein bound chromophore for the PΦB cofactor was done by comparing the energy
are listed in Table 3 whereas the corresponding optimized minimized structures of the isolated PΦB molecule and the
dihedral parameters are given in Table 2. model fragments (see section 2.2) obtained at the QM and MM
Furthermore, it is important to mention that, for the MM in levels. Energy minimizations of the PΦB molecule were
vacuo calculations, the force field parameters of the PΦB- performed starting from the crystal structure (the protein-bound
propionic side chain in its neutral form were taken from the PΦB in DrCBD). To avoid the formation of hydrogen bonds
acetic acid included in the CHARMM force field (see between the propionate side chains and the PΦB backbone
toppar_all22_prot_model.str/RESI ACEH). In the condensed during minimizations in vacuo, which may distort the validation
phase, molecular dynamics simulations of PΦB with charged of the new force field parameters, the PΦB-acidic groups were
propionate side chain were performed using the force field removed.
parameters derived for the heme cofactor. Figure 4 (bottom) shows the resulting MM- and QM-
During the parametrization of the dihedral torsions emphasis minimized structure of PΦB and the corresponding root-mean-
was placed on the accurate reproduction of the local minima, square deviations (rmsd) for all heavy atoms after optimal
MD Simulations of the Chromophore Binding Site of DrCBD J. Phys. Chem. B, Vol. 113, No. 4, 2009 951
TABLE 4: Interaction Energies and Distances for Chromophore-Water Complexes Evaluated at the ab Initio (HF/6-31G(d))
and Empirical Levels
HF/6-31G(d) empirical/CHARMM deviations
interaction
Emin (kcal/mol) Rmin (Å) Emin (kcal/mol) Rmin (Å) ∆E (kcal/mol) ∆R (Å)
1 -2.549 3.39 -2.762 3.38 -0.213 -0.01
2 -3.789 3.27 -3.883 3.43 -0.094 0.16
3 -5.418 2.63 -5.386 2.58 0.032 -0.05
4 -6.154 2.60 -6.036 2.55 0.118 -0.05
5 -3.059 3.39 -3.0427 3.39 0.016 0.0
6 -3.095 4.09 -4.479 3.40 -1.384 -0.69
7 -2.329 3.33 -2.497 3.33 -0.168 0.0
8 -2.672 3.80 -2.583 3.51 0.082 -0.29
9 -2.909 2.06 -3.887 1.82 -0.98 -0.24
10 -3.692 2.61 -4.298 2.54 -0.61 -0.07
11 -2.450 3.41 -2.597 3.34 -0.147 -0.07
12 -0.739 3.25 -0.977 3.26 -0.238 0.01
13 -5.489 2.51 -5.648 2.36 -0.159 -0.15
14 -4.966 3.20 -5.019 3.14 -0.053 -0.06
average differential energy: 0.307 kcal/mol
average difference distance: 0.184 Å
dipole moment (Debye)/HF-6-31G(d): total ) 13.370, x ) -13.369, y ) 0.104, z ) -0.162
dipole moment (Debye)/empirical: total ) 14.646, x ) -14.556, y ) -1.426, z ) 0.772
Figure 5. Root-mean-square deviations (rmsd) for the protein (light gray), its backbone atoms (dark gray) and the chromophore heavy atoms
(black) and as a function of time.
This gives a strongly negatively charged protein with a total strained to their minimum energy values by applying the
charge of -10 e. SHAKE algorithm.25 All molecular structures were drawn using
Afterward, the protein was solvated in a cubic box of water the visual Molecular Dynamics Software VMD 1.8.6.26
with an initial volume of 884736 Å3 (26307 solvent water 3.2. Dynamic Properties of the Protein. In Figure 5 the
molecules). The shortest initial distance between protein and root-mean-square deviations (rmsd) for all heavy atoms of the
water box walls was approximately 15 Å. This distance is protein and chromophore were plotted over the simulation time
deliberately larger than the chosen van der Waals long-range using the first frame of the simulation as the reference struc-
cutoff (12 Å) to avoid long-range interactions between the ture (conformations were stored each 1 ps). Because the rmsd
protein and its images. The TIP3P potential was used for the increased progressively during the first 5 ns of the MD
water molecules.17 The protein charge was neutralized by simulation, they were considered as part of the equilibration
inserting chloride and sodium counter-ions in the following phase and were not used for further analysis. The rmsd during
manner: each charged residue not involved in salt bridges was the remaining 45 ns of the simulation, on the other hand,
individually neutralized by placing a counterion within a 3 Å remained stable and were therefore used for the statistical
radius from it, whereas the remaining ions were placed randomly evaluation of the trajectory. The average of the rmsd evaluated
in the water box. over the 45 ns of the production run yield 2.4 Å for the entire
All subsequent calculations concerning heating, equilibration protein, 1.9 Å for the protein backbone and 1.3 Å for the
and production were performed with the NAMD program chromophore (excluding hydrogen atoms). These values are
(version 2.6)16 using the CHARMM27 force field. At first the similar to those previously calculated for cytochrome c oxi-
solvent waters were heated to 300 K and equilibrated for 40 ps dase.13 The slightly higher average rmsd for the protein results
while keeping all other parts of the system fixed. Subsequently, from the larger mobility of the protein side chains. During the
several energy minimization runs (2000 steps of conjugate simulation the overall shape of the protein is conserved as
gradient) were performed in which the initial harmonic restraints indicated by the small fluctuations (0.12 Å) of its calculated
(10 kcal/(mol Å2)) applied to the chromophore and all heavy average radius of gyration of 20.33 Å. The rmsd involving all
backbone atoms of the protein were gradually released until heavy atoms of the average protein structure resulting from the
the entire system was free. These harmonic positional restrains MD simulation and the crystal structure of DrCDB is 1.82 Å.
were described with a quadratic potential energy function. The mayor structural differences between these two structures
Afterward, the system was heated up to 300 K during 60 ps are found in the unstructured coil region near the N-terminus
using Langevin dynamics with restraint chromophore and (residues 4-11) as well as in the three R helices defined by the
protein backbone atoms. These steps were followed by 100 ps residue sequences 58-62, 66-69 and 81-88. Structurally very
molecular dynamics simulation under constant pressure, constant conserved on the other hand, are the six β strands surrounding
temperature (NPT) conditions using a combination of the the chromophore.
Langevin Piston Nose-Hoover method,22,23 as implemented in To analyze the mobility of the molecular system, the atomic
NAMD. Here again, the harmonic quadratic restraints were root-mean-square fluctuations (rmsf) were calculated from the
stepwise released until the system was totally unrestrained. MD simulation. Under the assumption of isotropic atomic
After the system was equilibrated, we proceeded with the fluctuations, the rmsf are related to the experimental B-factors
production run consisting in 50 ns molecular dynamics simula- (Debye-Waller or temperature factor) from the X-ray structure
tion using a reduced Langevin damping factor (varying from via the following expression27
5.0 to 1.0) to more closely approximate free dynamics (NPT
conditions at 300 K). All simulations were done under periodic 3
boundary conditions with the long-range electrostatic interaction 〈∆ri2〉 ) Bi (2)
described via the particle-mesh-Ewald method.24 For the van 8π2
der Waals interactions a cutoff (12 Å) was used in combination
with a switching function. To use a 2 fs time step, all bond where Bi and the square root of 〈∆ri2〉 represent the temperature
lengths between heavy atoms and hydrogen have been con- factor and the root-mean-square fluctuations associated with
MD Simulations of the Chromophore Binding Site of DrCBD J. Phys. Chem. B, Vol. 113, No. 4, 2009 953
Figure 6. Per residue averaged root-mean-square fluctuations over all atoms derived from of experimental B-factor (black) and calculated from
MD simulation (red).
atom i, respectively. The 〈∆ri2〉 of each atom was calculated calculated over the time intervals (each 4 ns long) indicated in
with respect to its average position during the MD simulation. Figure 7 with dashed boxes. The results are presented in Figure
Figure 6 shows the experimental and the calculated rmsf 8, where the average structures for state 1 (left) and for state 2
averaged over the individual amino acids. The theoretical values (right) are superimposed with the crystallographic structure. For
are systematically larger than the experimental ones by almost state 1, we obtain an overall good agreement concerning the
a factor of 2. Therefore, one should compare these values in a chromophore and linkage to the protein (rmsd for PΦB of 0.58
qualitative manner28 to determine whether the relative mobilities Å). The distance between ring A of the chromophore and HIS
predicted by the MD simulation for different regions of the 260 (5.23 Å) is comparable to the X-ray structure (4.99 Å).
system are in accordance with the experiment. In fact, the results The average structure of state 2 reveals, on the other hand,
plotted in Figure 6 indicate that for DrCBD that is, in general, significant distortions of the chromophore structure (rmsd for
the case. For example, we see that for one of the most flexible PΦB: 0.92 Å) that mainly involve ring A and ring C. Unlike
parts in the protein, a 3-10 helix (residues 130-134), simula- the crystal structure, ring A moves out of the plane formed by
tion and experiment are in a qualitative good agreement. the two coplanar inner rings. The dihedral angles at the A-B
However, large deviation in flexibility can be observed for the methine bridge of the average structures predicted for these two
loop region between two β-strands comprising residues 279-282 chromophore states are 13.41° (single bond)/18.44° (double
as well as for the loop region between residues 104-108. In bond) in state 1 and 18.48° (single bond)/17.21° (double bond)
particular, the predicted mobility of residues lying in the vicinity in state 2. These torsional angles are significantly larger
of the chromophore, such as CYS24, ASP207, MET 259, HIS compared to those measured in the X-ray structure: 7.55° (single
260, SER 272, SER 274 and HIS290, agrees very well with bond)/6.773° (double bond). In addition, from an energetic point
the corresponding B-factors. of view, we cannot distinguish one state from the other. The
3.3. Dynamic Properties of the Chromophore. When energy plots along the simulation show that the protein and the
analyzing the 45 ns production run of the MD trajectory, we chromophore remain stable. This would be in agreement with
noticed a particularly interesting behavior of the chromophore the fact that at room temperature and during 45 ns simulation
and its interaction with residue HIS 260. During the simulation, both states are almost equally occupied (see Figure 7). Other
the system seems to jump between two states characterized by regions of the chromophore are structurally similar for both
the presence (state 1) or absence (state 2) of a water molecule states, although slight deviations also exist in the vicinity of
localized between rings B and C, the so-called pyrrole water ring D.
illustrated in Figure 11. In state 1, the pyrrole water is hydrogen The conformational flexibility of the chromophore was further
bonded to HIS260 and to the NH group on ring A as observed investigated by analyzing the statistical distribution of all
in the crystal structure. In state 2, on the other hand, this pyrrole dihedral angles at the methine bridges. Except for the single
water moves away from the binding pocket and HIS260 comes bond torsion between rings A and B (A-B methine bridge) all
closer to ring A strengthening its interaction with the chro- distributions could be perfectly fitted with a single Gaussian
mophore. Each state is therefore characterized by distinct N_C function with widths oscillating between 8° to 10°. As an
(ring A in PΦB)-ND1 (H260) distances as shown in Figure 7. example we plotted in Figure 9 (bottom, left) the statistical
To evaluate in more detail the structural properties of the distribution of the dihedral angle at the double bond of the A-B
chromophore in each of these states, average structures were methine bridge. In this case a single Gaussian function with
954 J. Phys. Chem. B, Vol. 113, No. 4, 2009 Kaminski et al.
Figure 7. Time dependent distance between ring A (N_C) of the chromophore and HIS 260 (ND1). Dashed boxes indicate the time interval from
which average structures of state 1 and state 2 were determined.
Figure 8. Chromophore binding pocket: Crystallographic structure (dark gray) superimposed with average structures (light gray) resulting from a
4 ns MD simulation of state 1 (left) and state 2 (right). rmsd values for the heavy atoms of PΦB are 0.58 Å (state 1) and 0.92 Å (state 2).
maximum at 18.7° and a bandwidth of 14° is sufficient for a experimental X-ray B-factors for the chromophore which are
satisfactory description of the dihedral angle distribution. The larger for ring A than for the rest of the PΦB chromophore. In
distribution of the A-B torsion around the single bond, addition, these results would also support experimental observa-
however, displays an asymmetric shape (see Figure 9 bottom tions derived from NMR9 and resonance Raman spectroscopy.6
right), which can only be fitted if at least two Gaussian By plotting the N_C-ND1(HIS260) distance as a function
components are considered. The maxima of these two compo- of the dihedral angles at the A-B methine bridge (Figure 9, top
nents are located at 9.5° and 16.2° with bandwidths of 12° and left and right) we could extract some information about how
9°, respectively, which are slightly larger than those required these quantities correlate. The scatterplots for the two dihedral
for fitting the angular distribution of the other torsions. Because angles show basically two cluster of points located at the average
the band widths of the distribution curves reflect the degree of N_C-ND1 distances characterizing the protein’s state 1 and
fluctuation associated with each torsional angle, we conclude state 2 described above. In this respect, the upper cluster
that the A-B methine bridge is more flexible than the B-C and corresponds to state 1 whereas the lower cluster stands for state
C-D methine bridges. This conclusion is in agreement with the 2. For the scatter plot associated with the double bond torsion
MD Simulations of the Chromophore Binding Site of DrCBD J. Phys. Chem. B, Vol. 113, No. 4, 2009 955
Figure 9. Dihedral angle distributions (bottom) for the double (left) and single (right) bond torsion between rings A and B estimated over the 45
ns MD simulation. Scatter plots (top) show the correlation between these dihedral angles and the N_C (ring A)-ND1 (HIS 260) distance distribution.
TABLE 8: Statistics of H-Bond Events between Important bridge reduces the mobility of ring D in the Pr state. In addition,
Water Molecules Surrounding the Chromophore W3, W5, W6 and W7 establish relatively stable H-bonds with
water water average lifetime (ps) events occupancy (%) the two PΦB propionate side chains. The corresponding
hydrogen bond occupancies are larger than 50%. Although water
W1 W2 39.1 905 79
W2 W4 6.9 2077 32
W4 is not H-bonded neither to the chromophore nor to the
W3 W4 93.1 291 60 apoprotein, its presence seems to be of fundamental importance
W4 W5 7.6 1816 31 for the stability of the entire water network. This water is located
W5 W6 33.0 738 54 between the propionate side chains and interacts with W3, W2
W7 W8 16.8 1626 61 and W5 with hydrogen bond occupancies larger than 30%. These
H-bonds are constantly formed and broken during the entire
production run; however, the relative positions of these water
equal than a 2.4 Å.31 The stability of H-bonds formed between molecules do not significantly change. The most stable
the protein and the chromophore can be efficiently described water-water interaction is that predicted between W1 and W2
in terms of H-bond occupancy as the product between average with 79% H-bond occupancy. As shown in Figure 11, these
lifetime and number of events divided by the temporal length two water molecules do not directly interact with the chromophore.
of the MD simulation. The results presented in Table 6 show Noticeable is the dynamical behavior of the crystallographic
that stable H-bonds are those formed between the ARG254 and pyrrole water. This water is located in the center of the
SER274 residues and the oxygen atoms at the propionic side chromophore cavity forming H-bonds with the NH groups of
chain bound to ring B, with occupancies higher than 65%. This rings A, B and C. Our molecular dynamics simulations show
salt bridge between the ARG254 and PΦB seem to be that these H-bonds are not strong enough to trap the pyrrole
responsible for rigidity of chromophore’s B ring. In a similar water in the cavity for a long time. In particular, we observe
way, the propionic side chain attached to ring C forms a stable that the pyrrole water moves away from its initial position, and
H-bond with the HIS260 and with the SER274, characterized after a while the empty space is filled up with a solvent water.
by long average lifetime of 405 and 512 ps, respectively, and This process is repeated several times during the 45 ns
low number of events. The H-bonds formed between the production run. As mentioned in the previous section, the
ASP207 and the NH groups of rings B and C, on the other movement of the pyrrole water out of the chromophore cavity
hand, do not seem to be very stable. They are constantly broken significantly alters the chromophore structure.
and formed during the simulation, as indicated by the large
number of events, but each time they are formed they only exist
for a very short time (lifetime ∼ 10 ps). Thus the predicted 4. Conclusions
occupancies are fairly low (36% and 21%). In the present work, we could successfully extend the
3.5. Water Network in the Chromophore Binding Pocket. CHARMM force field by developing parameters for phytocho-
To investigate the formation of a water network in the mobilin (PΦB), a member of the biologically important family
chromophore binding pocket, we selected eight water molecules of linear tetrapyrroles. The set of parameters reproduces in an
which remained during the production run within a 5 Å distance accurate manner the structural properties, rotational barriers and
from the chromophore. An illustration of the most stable water intermolecular interactions of the chromophore as compared to
network we identified is given in Figure 11, showing the ab initio target data.
geometric average over the last 15 ns of the MD simulation. In addition, the quality of the new force field parameters was
Among these water molecules, only W3 is already observed in tested by performing molecular dynamics simulations of the
the X-ray structure; W5, W7 and W8 are solvent waters that at entire bacteriophytochrome DrCDB. As a result, both structural
an early stage of the simulation occupy the position of a (average structure) and dynamic properties (root-mean-square-
crystallographic water, and the remaining water molecules are fluctuations) of the chromophore are in satisfactory agreement
solvent water that find a stable position in the vicinity of the with the available experimental data. In contrast to the X-ray
chromophore. As described in the previous section, the dynami- structure where the chromophore’s rings A, B and C lie in a
cal behavior of this group of water molecules was analyzed by coplanar arrangement, the MD simulations favors a twisted
performing a statistical analysis of the H-bonding events conformation of the A-B methine bridge.
involving the selected waters and calculating the so-called Furthermore, the MD simulations reveal the existence of
H-bond occupancy. The results of the statistical analysis are structural heterogeneity of PΦB chromophore in DrCBD. We
listed in Tables 7 and 8. could identify two conformational states of the chromophore
Very stable interactions are the H-bonds formed between W2 characterized by distinct torsional angles around the C-C single
and the ARG 254 and THR256. These H-bonds are character- bond at the A-B methine bridge: 13.8° for state 1 and 18.5°
ized by a large number of events (1127 and 1538) with average for state 2. The formation of these two states is associated with
lifetimes of ca. 25 ps and, consequently, large occupation factors the formation of hydrogen bonds between the residue HIS260
(79% and 84%). Similar results are found for the H-bond and the PΦB cofactor, as well as with the presence/ absence or
interactions formed between W6 and the PΦB-propionic side the pyrrole water in the chromophore cavity. The existence of
chain on ring C and between W7 and the SER272 residue. An conformational heterogeneity involving ring A is in qualitative
also very important interaction is that between THR224 and agreement with NMR and RR spectroscopic data.
W5. This water forms only few H-Bonds with the threonine, A statistical analysis of H-bonding events was performed to
but each of them is stable for 1.3 ns. study the dynamical behavior of H-bonds in the chromophore
Furthermore, water W8 seems to be an important water bridge binding pocket. Among the ten H-bond donor and acceptor pairs
connecting the chromophore’s ring D with the propionic side considered in our analysis, those formed between the PΦB
chain attached to ring C. The H-bonds involving W8 and the chromophore and HIS260, ARG254 and SER274 residues are
PΦB chromophore are very stable as indicated by the high in particular very stable. These H-bonds are responsible for the
occupancy values (75% for the W8-PΦB(H_B) hydrogen bond rigidity of the propionic side chains at rings B and C and may
and 51% for the W8-PΦB(O2A) hydrogen bond). This water therefore of relevance for understanding the signal transduction
958 J. Phys. Chem. B, Vol. 113, No. 4, 2009 Kaminski et al.
mechanism between chromophore and protein at an intermediate (10) MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack, R. L.;
stage of the photocycle. Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-
McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos, C.;
Finally, the MD simulations allowed us to identify a stable Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux,
water network comprising at least eight water molecules in the B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.;
close vicinity of the chromophore. The presence of this water Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M. All-atom empirical potential
for molecular modeling and dynamics studies of proteins. J. Phys. Chem.
network has a strong influence on the chromophore’s structural B 1998, 102 (18), 3586–3616.
properties. In particular, the propionic chains and ring D gain (11) Foloppe, N.; MacKerell, A. D. All-atom empirical force field for
rigidity from this network. nucleic acids: I. Parameter optimization based on small molecule and
condensed phase macromolecular target data. J. Comput. Chem. 2000, 21
(2), 86–104.
Additional Information (12) Reuter, N.; Lin, R.; Thiel, W. Green fluorescent proteins: Empirical
The parameter and topology files for PΦB are available from force field for the neutral and deprotonated forms of the chromophore.
Molecular dynamics simulation’s of the wild type and S65T mutant. J. Phys.
the authors upon request. Chem. B 2002, 106 (24), 6310–6321.
(13) Olkhova, E.; Hutter, M. C.; Lill, M. A.; Helms, V.; Michel, H.
Acknowledgment. We acknowledge the help from the Dynamic water networks in cytochrome c oxidase from Paracoccus
Konrad-Zuse-Zentrum in Berlin for supplying us with compu- denitrificans investigated by molecular dynamics simulations. Biophys. J.
2004, 86 (4), 1873–1889.
tational resources and we would like to thank Dr. Bernd Kallies (14) Gaussian 03, revision C.02; Gaussian, Inc.: Wallingford, CT, 2004.
for technical support, as well as Prof. Peter Hildebrandt for (15) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.;
helpful discussions. The financial support by the Deutsche Swaminathan, S.; Karplus, M. Charmm - A Program for Macromolecular
Energy, Minimization, and Dynamics Calculations. J. Comput. Chem. 1983,
Forschungsgemeinschaft, SfB498 to M.A.M and G.D. and the 4 (2), 187–217.
financial support by the Investitionsbank Berlin and the Excel- (16) Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.;
lence Cluster “Unifying concepts in catalysis” to S.K. and Villa, E.; Chipot, C.; Skeel, R. D.; Kale, L.; Schulten, K. Scalable molecular
M.A.M. is gratefully acknowledged. dynamics with NAMD. J. Comput. Chem. 2005, 26 (16), 1781–1802.
(17) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.;
Klein, M. L. Comparison of Simple Potential Functions for Simulating
Abbreviations Liquid Water. J. Chem. Phys. 1983, 79 (2), 926–935.
(18) Becke, A. D. A New Mixing of Hartree-Fock and Local Density-
RR: resonance Raman Functional Theories. J. Chem. Phys. 1993, 98 (2), 1372–1377.
DrCB: Deinococcus radiodurans chromophore binding domain (19) Breneman, C. M.; Wiberg, K. B. Determining Atom-Centered
PΦB: Phytochromobilin Monopoles from Molecular Electrostatic Potentials - the Need for High
Sampling Density in Formamide Conformational-Analysis. J. Comput.
QM: quantum mechanical Chem. 1990, 11 (3), 361–373.
MM: molecular mechanical (20) Autenrieth, F.; Tajkhorshid, E.; Baudry, J.; Luthey-Schulten, Z.
MD: molecular dynamics Classical force field parameters for the heme prosthetic group of cytochrome
c. J. Comput. Chem. 2004, 25 (13), 1613–1622.
References and Notes (21) Brunger, A. T.; Karplus, M. Polar Hydrogen Positions in Proteins
- Empirical Energy Placement and Neutron-Diffraction Comparison. Proteins-
(1) Rockwell, N. C.; Su, Y. S.; Lagarias, J. C. Phytochrome structure Struct. Funct. Genet. 1988, 4 (2), 148–156.
and signaling mechanisms. Annu. ReV. Plant Biol. 2006, 57, 837–858. (22) Feller, S. E.; Zhang, Y. H.; Pastor, R. W.; Brooks, B. R. Constant-
(2) Quail, P. H. Phytochrome photosensory signalling networks. Nature Pressure Molecular-Dynamics Simulation - the Langevin Piston Method.
ReV. Mol. Cell Biol. 2002, 3 (2), 85–93. J. Chem. Phys. 1995, 103 (11), 4613–4621.
(3) Briggs, W. R.; Spudich, J. L. Handbook of Photosensory Receptors; (23) Martyna, G. J.; Tobias, D. J.; Klein, M. L. Constant-Pressure
Wiley Verlag: Weinheim, Germany, 2005. Molecular-Dynamics Algorithms. J. Chem. Phys. 1994, 101 (5), 4177–4189.
(4) Wagner, J. R.; Brunzelle, J. S.; Forest, K. T.; Vierstra, R. D. A (24) Darden, T.; York, D.; Pedersen, L. Particle Mesh Ewald - An
light-sensing knot revealed by the structure of the chromophore-binding N.Log(N) Method for Ewald Sums in Large Systems. J. Chem. Phys. 1993,
domain of phytochrome. Nature 2005, 438 (7066), 325–331. 98 (12), 10089–10092.
(5) Wagner, J. R.; Zhang, J. R.; Brunzelle, J. S.; Vierstra, R. D.; Forest, (25) van Gunsteren, W. F.; Berendsen, H. J. C. Algorithms for
K. T. High resolution structure of Deinococcus bacteriophytochrome yields Macromolecular Dynamics and Constraint Dynamics, 34th ed.; 1977; pp
new insights into phytochrome architecture and evolution. J. Biol. Chem. 1311-1327.
2007, 282 (16), 12298–12309. (26) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular
(6) von Stetten, D.; Günter, M.; Scheerer, P.; Murgida, D. H.; dynamics. J. Mol. Graph. 1996, 14 (1), 33.
Mroginski, M. A.; Krauβ, N.; Lamparter, T.; Zhang, J.; Anstrom, D. M.; (27) Willis, B. T. M.; Prior, A. W. Thermal Vibrations in Crystal-
Vierstra, R. D.; Forest, K. T.; Hildebrandt, P. Chromophore heterogeneity lography; Cambridge University Press: New York, 1975.
and the photoconversion in phytochrome crystals and solution studied by (28) Brooks, C. L.; Karplus, M.; Pettitt, B. M. A. Theoretical Perspective
resonance Raman spectroscopy. Angew. Chem. Int. Ed. 2008, 47, 1–4. Of Dynamics, Structure, and Thermodynamics. AdV. Chem. Phys. 1988,
(7) Wagner, J. R.; Zhang, J.; von Stetten, D.; Günter, M.; Murgida, 71, 1–200.
D. H.; Mroginski, M. A.; Walker, J. M.; Forest, K. T.; Hildebrandt, P.; (29) Essen, L. O.; Hughes, J.; Mailliet, J.; The structure of a complete
Vierstra, R. D. Mutational analysis of Deinococcus Radiodurans Bacterio- phytochrome sensory module in the Pr ground stateProc. Natl. Acad. Sci.
phytochrome reveals key amino acids necessary for structural integrity and U.S.A. 2008, 105, 14709–14714.
the proton exchange cycle during phoroconversion. J. Biol. Chem. 2008, (30) Yang, X.; Stojkovic, E. A.; Kuk, J.; Moffatt, K. Crystal structure
283, 12212–12226. of the chromophore binding domain of an unusual bacteriophytochrome,
(8) Schmidt, P.; Gensch, T.; Remberg, A.; Gartner, W.; Braslavsky, RpBphP3, reveals residues that modulate photoconversion. Proc. Natl. Acad.
S. E.; Schaffner, K. The complexity of the P-r to P-fr phototransformation Sci. U.S.A. 2007, 104 (30), 12571–12576.
kinetics is an intrinsic property of native phytochrome. Photochem. (31) Deloof, H.; Nilsson, L.; Rigler, R. Molecular-Dynamics Simulation
Photobiol. 1998, 68 (5), 754–761. of Galanin in Aqueous and Nonaqueous Solution. J. Am. Chem. Soc. 1992,
(9) van Thor, J. J.; Mackeen, M.; Kuprov, I.; Dwek, R. A.; Wormald, 114 (11), 4028–4035.
M. R. Chromophore structure in the photocycle of the cyanobacterial
phytochrome Cph1. Biophys. J. 2006, 91 (5), 1811–1822. JP8047532