Molecular Dynamics Simulations of p53 DNA-Binding Domain

Abstract

We have studied room-temperature structural and dynamic properties of the p53 DNA-binding domain in both DNA-bound and DNA-free states. A cumulative 55 ns of explicit solvent molecular dynamics simulations with the particle mesh Ewald treatment of electrostatics was performed. It was found that the mean structures in the production portions of the trajectories agree well with the crystal structure: backbone root-mean-square deviations are in the range of 1.6 and 2.0 A. In both simulations, noticeable backbone deviations from the crystal structure are observed only in loop L6, due to the lack of crystal packing in the simulations. More deviations are observed in the DNA-free simulation, apparently due to the absence of DNA. Computed backbone B-factor is also in qualitative agreement with the crystal structure. Interestingly, little backbone structural change is observed between the mean simulated DNA-bound and DNA-free structures. A notable difference is observed only at the DNA-binding interface. The correlation between native contacts and inactivation mechanisms of tumor mutations is also discussed. In the H2 region, tumor mutations at sites D281, R282, E285, and E286 may weaken five key interactions that stabilize H2, indicating that their inactivation mechanisms may be related to the loss of local structure around H2, which in turn may reduce the overall stability to a measurable amount. In the L2 region, tumor mutations at sites Y163, K164, E171, V173, L194, R249, I251, and E271 are likely to be responsible for the loss of stability in the protein. In addition to apparent DNA contacts that are related to DNA binding, interactions R175/S183, S183/R196, and E198/N235 are highly occupied only in the DNA-bound form, indicating that they are more likely to be responsible for DNA binding.

Full text

Molecular Dynamics Simulations of p53 DNA-Binding Domain
Qiang Lu1, Yu-Hong Tan1, and Ray Luo*
Department of Molecular Biology and Biochemistry, University of California, Irvine, CA 92697-3900
Abstract
We have studied room-temperature structural and dynamic properties of the p53 DNA-binding
domain in both DNA-bound and DNA-free states. A cumulative 55ns of explicit solvent molecular
dynamics simulations with the Particle Mesh Ewald treatment of electrostatics were performed. It is
found that the mean structures in the production portions of the trajectories agree well with the crystal
structure: backbone root-mean squared deviations are in the range of 1.6Å and 2.0Å. In both
simulations, noticeable backbone deviations from the crystal structure are observed only in loop L6,
due to the lack of crystal packing in the simulations. More deviations are observed in the DNA-free
simulation, apparently due to the absence of DNA. Computed backbone B-factor is also in qualitative
agreement with the crystal structure. Interestingly little backbone structural change was observed
between the mean simulated DNA-bound and DNA-free structures. Notable difference is only
observed at the DNA-binding interface. The correlation between native contacts and inactivation
mechanisms of tumor mutations is also discussed. In the H2 region, tumor mutations at sites D281,
R282, E285, and E286 may weaken five key interactions that stabilize H2, indicating that their
inactivation mechanisms may be related to the loss of local structure around H2, which in turn may
reduce the overall stability to a measurable amount. In the L2 region, tumor mutations at sites Y163,
K164, E171, V173, L194, R249, I251 and E271 are likely to be responsible for the loss of stability
in the protein. In addition to apparent DNA contacts that are related to DNA binding, interactions
R175/S183, S183/R196, and E198/N235 are highly occupied only in the DNA-bound form,
indicating that they are more likely to be responsible for DNA binding.
INTRODUCTION
p53 is a transcription factor that binds to specific DNA sequences1–4 to activate gene
expression.5–9 In response to activation of certain oncogenes, it can stop cell cycle, prevent
genetic alterations, and induce apoptosis or programmed cell death.10–12 Mutation in the p53
gene is one of the most frequent events in the process of oncogenesis. It is estimated that
approximately 50% of human tumors contain mutations in this pivotal gene.13 The in vivo
importance of p53 transactivation is underscored by the fact that 95% of all known tumor
mutations occur in the DNA-binding domain of the protein.13,14 More interestingly, 75% of
these mutations occur as single missense mutations.13,14 Therefore, the oncogenic form of
p53 is predominantly a full-length protein with a single amino-acid substitution in its DNA-
binding domain. Tumors with inactive p53 mutants are aggressive and often resistant to
ionizing radiation and chemotherapy. Clearly, p53 mutants present one of the most important
clinical targets for drug intervention of tumors. Many experimental approaches are in progress
to develop p53-based therapies.15,16
p53 bears the usual hallmarks of a transcription factor, with an amino-terminal transactivation
domain, a DNA-binding domain, and carboxy-terminal tetramerization and regulatory
* Please send correspondence to: R. Luo. E-mail: rluo@uci.edu. FAX: (949) 856-8551.
1These authors contributed equally to this work.
NIH Public Access
Author Manuscript
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
Published in final edited form as:
J Phys Chem B. 2007 October 4; 111(39): 11538–11545. doi:10.1021/jp0742261.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

domains. The crystal structure of the DNA-binding domain, solved in 1994, provides a starting
point for understanding the nature of mutant p53 (Fig. 1).17 The structure consists of a β-
sandwich scaffold and a DNA-binding surface, including a loop-sheet-helix motif (LSH) and
two loops (L2 and L3) tethered by a single Zinc atom. The loops and the LSH motif form the
DNA-binding surface of p53 and provide contacts to the DNA backbone and bases.
Due to the large number of mutations occurring at the p53 DNA-binding domain, it is
instructive to classify these mutations based on their locations for better understanding of their
effects for drug intervention. It appears that 30% of all mutations fall on six mutation hot spots,
and these hot spots cluster to the DNA-binding surface: two contact DNA directly, R248 (L3)
and R273 (LSH), and four stabilize the surrounding structure, R175 (L2), G245 (L3), R249
(L3) and R282 (LSH).17 These give rise to two classes of mutations, DNA-contact and
structural. At the physiological condition, all six hot spots fail to bind target DNA sequences.
Further classifications of mutants are possible after considering their thermodynamic
properties. Overall, three broad phenotypes are proposed: DNA-contact mutations that have
little or no effect on stability, mutations that disrupt local structure, and mutations that cause
denaturation.17,18 To date only a small fraction of tumor mutations are analyzed by
experimental methods. The inactivation mechanisms of many tumor mutations are still
unknown.17,18
This article reports our effort in the computational analysis of this important protein.
Specifically, molecular dynamics simulations in explicit solvent have been used to study the
dynamic properties of the wild type protein to infer the inactivation mechanisms of tumor
mutations.
METHODS
There are two computational difficulties in molecular dynamics simulations of the p53 DNA-
binding domain (core domain, p53c). The first difficulty is its large size: a monomer p53c is
already around 50Å in diameter, but p53 functions as a tetramer in cell. Simulation of the
complete tetramer p53c in explicit water would be extremely difficult considering the number
of water molecules needed to solvate the protein. Fortunately, it has been shown that the
monomer p53c still binds to DNA, though weaker.19 This indicates that the monomer p53c
can be used a model to study this important interaction. Further, all existing biophysical
analysis of the wild type and mutants of p53c show that the monomer p53c is stable in vitro,
so that the monomer model is also suitable for its stability analysis.18
The second difficulty is accurate modeling of the Zinc-binding interface in p53c. Previous
experiments with metal chelators showed that p53 depends upon the coordination of Zinc for
both correct folding and correct binding to specific DNA in intact cells, especially for binding
in the DNA minor groove.21,22 Therefore the Zinc-binding interface is important for
simulation of p53c and cannot be neglected. Two approaches have been developed for
modeling Zinc coordination interfaces in proteins. The non-bonded approach uses electrostatic
and van der Waals interactions to maintain four-ligand or five-ligand coordinations of Zinc in
proteins.23–26 It is suitable to simulate proteins in which Zinc is required for catalytic function,
though it is quite challenging to maintain the interface stability at room temperature during
long-time simulations. The bonded approach uses covalent bonds between Zinc and its
coordinating ligands to maintain desired geometries.27–29 The advantage of the bonded
approach is that it is easier to preserve desired geometries as in crystal structures and maintain
stability at room temperature by choosing suitable bond and angle parameters. The
disadvantage of this approach is that it cannot model bond breaking between Zinc and ligands.
In this study the bonded approach is chosen in our computational simulations of p53c.
Lu et al. Page 2
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Quantum mechanical analysis of the Zinc interface
The Zinc interface in p53c, [Zn(CYS)3(HIS)1], is chopped out from the crystal structure 1TSR
B (Fig. 1). To minimize perturbation to the interface, four main chain Cα atoms are included
in the model compound for optimization of force field parameters. Since there is no hydrogen
in the crystal structure to determine the total charge, all possible net charges, 2, 1, 0 and −1,
−2, are investigated (Table 1).
The geometries of all tested protonation states were optimized with tight option using rational
function optimization (RFO) method.30 After the RFO method, direct inversion in the iterative
subspace (GDIIS)31,32 was used to reach a minimum. B3LYP density functional method33,
34 with the 6-311+G** basis set was employed in the optimization. Finally B3LYP/6-311+G**
was chosen for frequency analysis because it was reported to be successful in reproducing the
hydration energy of Zinc divalent cation35,36 and in studying the farnesyltransferase
inhibitors.37 A scaling factor 0.989 was used for frequency calculation with the basis set 6-311
+G**.38 B3LYP/6-311+G** was also used to obtain ESP potential39–41 in the atomic charge
derivation. All quantum mechanical calculations were performed using Gaussian03(G03)42
at NCSA. It is found that only the protonation state with −1 net charge keeps the Zinc interface
structure intact. The structures with net charges of 2, 1, 0 and −2 lose the tetrahedral structure
after 20 steps of optimization. The final optimized structure of −1 charge is shown in Fig. 2.
Interestingly, the same protonation state with three thiolate cystine residues coordinated with
Zinc were found to facilitate the fourth thiolate cyctine residue to be a chemically active
nucleophilic species in the methyl transfer reactions.20
Molecular mechanics analysis of the Zinc interface
Molecular mechanical optimizations were performed with NMODE43 in Amber944 in order
to obtain normal mode frequencies. The initial geometry is the optimized structure from
G03.42 The Newton-Raphson method was applied to locate the minimum. The convergence
criterion is 10−6 kcal/mol-Å.
Two bonds and five angle parameters for the Zinc four-ligand coordination interface in p53c
[Zn(CYS)3(HIS)1] were empirically optimized based on the comparison of frequencies
between Amber and G03. We have set all Zinc-related dihedral terms to zero to simplify
parameterization. The van der Walls parameters for the Zinc atom were taken from the
literature: σ = 1.10Å and ε = 0.0125kcal/mol.45 RESP charges46–48 were derived from the
G03 ESP potential with ANTECHAMBER in Amber9.44 Partial charges of CYS (CA,
−0.0396; HA1, −0.002; HA2, −0.002; HA3, −0.002; CB, 0.2061; HB1, −0.0231; HB2,
−0.0231; SG, −0.7380) and partial charges of HIS (CA, −0.1523; HA1, 0.0249; HA2, 0.0249;
HA3, 0.0249; CB, 0.0693; HB1, 0.0694; HB2, 0.0694; CG, −0.1183; ND1, −0.0151; ZN,
0.8024; CE1, 0.0741; HE1, 0.0950; NE2, −0.3378; HE2, 0.3347; CD2, −0.1782; HD2, 0.1611)
were obtained. The total charge is −1 according to the QM analysis.
Molecular dynamics simulation of DNA-free and DNA-bound p53 core domain
As reviewed, the monomer p53c can bind DNA with approximately one-fifth the affinity of
intact p53.19 This gives us an opportunity to study the DNA binding of p53c with a monomer
model. In our molecular dynamics simulation, the initial complex structure is derived from the
p53c (1TSR chain B) in complex with DNA (5′-TAGACTTGCCCA-3′).17 1TSR chain B is
chosen as the initial DNA-free p53c structure to study the difference between DNA-bound and
DNA-free structures. Hydrogen atoms were added using LEAP of Amber9.44 In both
simulations the solute molecules were solvated in a TIP3P water box with a 10 Å buffer.49
Both systems were neutralized by adding counter ions. The detailed information of the two
simulation systems are listed in Table 2.
Lu et al. Page 3
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

A revised parm99 force field was used for intramolecular interactions.50,51 Particle Mesh
Ewald was used to treat long-range electrostatics with the default setting in Amber9.52 All
bonds involving hydrogens were constrained by SHAKE,53 so that a time step of 2fs was used
for dynamics integration. Before simulation, the initial coordinates were relaxed by 1000 steps
steepest-descent energy minimization. After minimization the two systems were heated up to
80K in order to adjust the water configurations, while the backbone atoms of protein and DNA
were restrained by weak harmonic forces. Finally the systems were heated up to 293K within
20ps. In the complex simulation, the backbone of DNA was weakly restrained by a weak force
of 1kcal/mol-Å2 throughout the simulation because it is truncated. Constant temperature
(293K) and constant pressure (1 bar) were maintained by the Berendsen’s method.54 A total
of 55ns trajectories at 293K were collected for both systems.
RESULTS AND DISCUSSION
Quality of covalent force field terms
Projections of normal modes onto internal coordinates agree well between G03 and Amber.
The distributions of normal modes over Zinc-related bonds and angles are listed in Supporting
Material (Fig. S-1). Note that variation of Zinc-related parameters only influences normal
modes less than 500 cm−1.
Table 3 lists all optimized force parameters of Zinc-related bonds and angles. Literature values
mainly come from Human Carbonic Anhydrase II (3HIS, 1H2O)27,28 and Alcohol
Dehydrogenase (2CYS, 1HIS, 1H2O).55 NA-ZN bond force constant, 40 kcal/mol-Å2, comes
from experimental data.56–58
Similar structures are obtained by Amber and G03 minimization as shown in Fig. 3. After
parameterization, comparison of frequencies over normal modes is performed and shown in
Fig. 4. Frequencies from 0 to 500cm−1 correspond to the Zinc-related covalent terms, whose
RMSD is 14.4cm−1 between G03 and Amber. The RMSD is 34.50cm−1 for frequencies below
1,000 cm−1. Overall an excellent agreement can be observed between Amber and G03.
Agreement between simulation and crystal structures
Molecular dynamics simulations of both DNA-bound and DNA-free proteins were run up to
30ns and 25ns, respectively. Their accumulate average potential energies show that the DNA-
bound trajectory has equilibrated after 22.5ns and the DNA-free trajectory has equilibrated
after 5ns, respectively (Fig. 5). Stable structures are observed in both trajectories, as reflected
in their backbone RMSD fluctuations (Fig. 6a).
The backbone RMSD, including all loop residues, of the DNA-bound protein with respect to
chain B of 1TSR (the one in complex with DNA) is about 1.6Å (Fig. 6a). While the backbone
RMSD of the DNA-free protein is around 2.0Å (Fig. 6a), which is higher than that of the
complex, apparently due to the absence of binding to DNA. The mean structures in the
production portion of the trajectories (7.5ns for DNA-bound and 20ns for DNA-free) also agree
with chain B of 1TSR well: the backbone RMSD of the DNA-bound and DNA-free proteins
is 1.61Å and 2.07Å, respectively (Fig. 7). This should be compared with the RMSD differences
among the three chains in 1TSR, around 0.85Å. Considering that the simulations were
performed at room temperature and without crystal packing, the agreement between simulation
and experiment is very good. In both simulations, noticeable backbone deviations from the
crystal structure are observed only in loop L6. This is due to the lack of crystal packing in the
monomer simulations. In the DNA-free simulation, deviations are also observed in L1, H1,
and L5′, apparently due to the absence of DNA. In Fig. 7 one can find that the L1 DNA-binding
interface has noticeably deformed in the DNA-free simulation.
Lu et al. Page 4
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Computed backbone B-factor of the DNA-bound protein is in qualitative agreement with that
of chain B in 1TSR (Fig. 8). Major differences between simulation and experiment are all due
to the existence of crystal packing with other chains in 1TSR (Fig. 8). Interestingly, computed
backbone B-factor of the DNA-free protein agrees better with chain A that does not have
specific DNA contacts (Fig. 9). This indicates that chain A might be used as a model of the
DNA-free protein for dynamics analysis. However, it should be pointed out that chain A is a
less accurate structural model than chain B.
RMSD fluctuation of the heavy atoms at the Zinc interface indicates that the tetrahedron
structure is maintained as in 1TSR (Fig. 6b). Further, relative B-factor of the side-chain heavy
atoms with respect to those of the backbone heavy atoms show that the bonded model adopted
here can reproduce the dynamic properties of the Zinc interface in p53c at least for equilibrium
properties at the native state (Table 4).
Differences between DNA-bound and DNA-free room-temperature structures
One of our interests is to understand the binding-induced conformational change in p53c at
room temperature. As shown in Fig. 7, overall there is little change in the backbone between
the mean DNA-bound and DNA-free structures, similar to that observed between crystal
structures 1TSR B and A chains at 98K. Notable difference in the backbone is only observed
in L1, H1, and L3 at the DNA-binding interface.
A detailed comparison of the arrangement of secondary structures in DNA-bound and DNA-
free structures is shown in Fig. 10. The major differences are only observed in places close to
the DNA-binding interface. On the H2 side of the binding interface, the only difference is at
L1 that moves away from H2 in the DNA-free structure. On the L3 side of the binding interface,
L3 is slightly distorted when compared with that in the DNA-bound structure. Another
difference is observed in loop L2, a central structure connecting L3 with L2′, H1, NT, L5′, and
portion of the β sandwich (Fig. 7). In fact, L2 has most contacts with other secondary structural
units in the DNA-binding region (Table 5). In the DNA-bound structure L2 is closer to L2′ and
L3, but it is farther away from L5′ and the β sandwich. In addition, the contacts between L2/
NT and L5′ are weaker when compared with those in the DNA-free structure (Table 5),
corresponding to the observed higher B-factor on NT and L5′ in the complex (Fig. 8). A similar
difference on NT and L5′ is also observed in the crystal structure 1TSR. NT is packed well in
chain A but not in chain B, and the B-factor of NT is also lower in chain A than in chain B
(Fig. 11). These differences indicate that contacts involving L2 may be important to DNA
binding even if they are not in direct contact with DNA.
The difference of native contacts between the DNA-bound and DNA-free protein is also
compared to understand the binding-induced conformational changes. This gives us the basis
for the following analysis related to common tumor mutations. In the DNA-binding region, all
side-chain/side-chain or side-chain/main-chain contacts are monitored. Those contacts with
percentage occupancies higher than 50% in the DNA-bond structure or the DNA-free structure
are listed in Table 5 and Table 6. Comparison between the contacts in the DNA-bound and
DNA-free structures reveals many differences. Interestingly, there are ten contacts
significantly weakened by binding to DNA (V97/M169, S116/C124, M169/R213, V173/A161,
R174/E180, R175/D184, D184/R196, D186/R196, L188/Y201, L194/Y236), but only five
contacts are significantly strengthened by binding to DNA (T118/R283, M169/I162, R175/
S183, S183/R196, L188/V203) outside the β sandwich (Table 5). In contrast, few differences
are found in the β-sandwich region as shown in Table 6. Specifically, four contacts (T140/
N235, V143/I255 and E198/H233. I251/V272) are weakened, and two contacts are
strengthened (V143/F270 and E198/N235) by DNA binding.
Lu et al. Page 5
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Implications to common tumor mutations
A more detailed analysis of the native contacts indicates that there are additional differences
between the two room-temperature simulations. These differences may shed lights in
understanding the correlation between tumor mutations and DNA binding or folding stability
in p53c. In the following, we discuss the correlation of native contacts with inactivation
mechanisms of certain tumor mutations in four different regions: β sandwich, H2 region, L2
region, and DNA-binding region.
β sandwich—It is apparent that mutations in the β-sandwich region (hydrophobic core) are
more likely to be responsible for the protein’s overall stability: most native contacts are
hydrophobic in nature. Instead of analyzing all contacts generically, we have chosen to focus
on two centers of the hydrophobic core: I195 and V143. Tumor mutations targeting these two
centers can severely destabilize the p53 core domain (mutation I195T and V143A destabilize
p53c by 4.12 kcal/mol and 3.50 kcal/mol, respectively).59 I195 forms hydrophobic contacts
with A161, Y234, and Y236. None of these change upon DNA binding (Table 6). V143 forms
hydrophobic contacts with L111, Y234, I255, and F270 (Table 6). Most of these hydrophobic
contacts are also well preserved in both simulations.
H2 Region—The inactivation mechanisms of tumor mutations in other parts of p53c are less
clear. Here we have monitored all side-chain contacts to track their occupancies at room-
temperature. In the H2 region, tumor mutations at sites D281, R282, E285, and E286 may
weaken five key interactions that stabilize H2 (Table 5). Our data indicates that their
inactivation mechanism may be related to the loss of local structure around H2, which in turn
may reduce the overall stability to a measurable amount. Our analysis on contacts related to
H2 is in part supported by the structure and stability data from Fersht and co-workers: mutation
R282W affects the packing of Loop-Sheet-Helix,60 and destabilizes p53c by 3.3 kcal/mol.59
L2 Region—In the L2 region, contacts K164/E271, V173/Y163, V173/L194, V173/I251,
and R249/E171 are highly occupied in both DNA-bound and DNA-free simulations. Our
contact analysis thus implies that mutations at sites Y163, K164, E171, V173, L194, R249,
I251 and E271 are likely to be responsible for the loss of stability in p53c. The finding for
R249 is in agreement with available structure and stability data from Fersht and co-workers:
mutation R249S was found to induce substantial structural perturbation around mutation site
in loop L3,61 and to destabilize p53c by 1.92 kcal/mol.59
DNA binding—Many interactions are in direct contact with DNA (Table 7), indicating that
they are responsible for DNA binding. Except N239 the results are in agreement with the crystal
structure from Cho et al: residue K120, S241, R248, R273, R280, and R283 are direct in contact
with DNA.17 In addition, the crystal structure from Fersht and co-workers also shows that
R273C and R273H mutations simply remove DNA contacts without perturbing the
conformation of nearby residues.60,61
In addition to apparent DNA contacts that are related to DNA binding, many interactions are
highly occupied only in the DNA-bound form, indicating that they are more likely to be
responsible for DNA binding. Three such contacts that are not in direct contact with DNA are
found: R175/S183, S183/R196 (Table 5), and E198/N235 (Table 6).
CONCLUDING REMARKS
With optimized force field parameters for the Zinc-binding interface of p53c, a cumulative
55ns of room-temperature molecular dynamics simulations in explicit solvent with the PME
treatment of electrostatics were performed for both DNA-bound and DNA-free p53c. It is found
that the backbone RMSDs, including all loop residues, of both DNA-bound and DNA-free
Lu et al. Page 6
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

p53c with respect to chain B in 1TSR are from 1.6Å to 2.0Å. The mean structures in the
production portions of the trajectories also agree with chain B of 1TSR well. In both
simulations, noticeable backbone deviations from the crystal structure are observed only in
loop L6. This is due to the lack of crystal packing in both simulations. In the DNA-free
simulation, deviations are also observed in L1, H1, and L5′, apparently due to the absence of
DNA. Computed backbone B-factor of the DNA-bound protein is also in qualitative agreement
with that of chain B in 1TSR. Major differences between simulation and experiment are due
to the existence of crystal packing with other chains in 1TSR. RMSD fluctuation of the heavy
atoms at the Zinc-binding interface indicates that the tetrahedron structure is maintained as in
1TSR. Relative B-factor of the side-chain heavy atoms with respect to those of the backbone
heavy atoms at the Zinc-binding interface also agrees well between simulation and crystal
structures, indicating that the bonded model adopted here may reproduce equilibrium dynamic
properties of the Zinc-binding interface in p53c.
Our room-temperature simulation shows little backbone change between the mean DNA-
bound and DNA-free structures, similar to that observed between crystal structures 1TSR B
and A chains at 98K. Notable difference in the backbone is only observed on the DNA-binding
interface. On the H2 side of the binding interface, the only difference is L1 that moves away
from H2 in the DNA-free structure. On the L3 side of the binding interface, L3 is slightly
distorted when compared with that in the DNA-bound structure. Another difference is observed
in loop L2. In the DNA-bound structure L2 is closer to L2′ and L3, but it is farther away from
L5′ and the β sandwich. In addition, the contacts between L2/NT and L5 are weaker when
compared with those in the DNA-free structure, corresponding to the observed higher B-factor
on NT and L5′ in the complex. A similar difference on NT and L5′ is also observed in the
crystal structure 1TSR. NT is packed well in chain A but not in chain B, and the B-factor of
NT is also lower in chain A than in chain B. These differences indicate that contacts that involve
L2 may be important to DNA-binding even if they are not in direct contact with DNA.
The correlation between native contacts and inactivation mechanisms of certain tumor
mutations is also discussed. It is apparent that mutations in the β-sandwich region (hydrophobic
core) are more likely to be responsible for the protein’s overall stability: most native contacts
are hydrophobic in nature. The inactivation mechanisms of tumor mutations in other parts of
p53c are less clear. Here we have monitored all side-chain contacts to track their occupancies
at room-temperature. In the H2 region, tumor mutations at sites D281, R282, E285, and E286
may weaken five key interactions that stabilize H2. Our data indicates that their inactivation
mechanism may be related to the loss of local structure around H2.
In the L2 region, contacts K164/E271, V173/Y163, V173/L194, V173/I251, and R249/E171
are highly occupied in both DNA-bound and DNA-free simulations. Our contact analysis thus
implies that mutations at sites Y163, K164, E171, V173, L194, R249, I251 and E271 are likely
to be responsible for the loss of stability in p53c.
Many interactions are in direct contact with DNA, indicating that they are responsible for DNA
binding. In addition to apparent DNA contacts that are related to DNA binding, many
interactions are highly occupied only in the DNA-bound form, indicating that they are more
likely to be responsible for DNA binding. Three such contacts that are not in direct contact
with DNA are found: R175/S183, S183/R196, and E198/N235.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Lu et al. Page 7
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Acknowledgements
This work is supported in part by NIH (GM069620) and the State of California (CRCC).
References
1. Bargonetti J, Friedman PN, Kern SE, Vogelstein B, Prives C. Cell 1991;65:1083. [PubMed: 1646078]
2. Kern SE, Kinzler KW, Bruskin A, Jarosz D, Friedman P, Prives C, Vogelstein B. Science
1991;252:1708. [PubMed: 2047879]
3. Funk WD, Pak DT, Karas RH, Wright WE, Shay JW. Molecular and Cellular Biology 1992;12:2866.
[PubMed: 1588974]
4. Eldeiry WS, Kern SE, Pietenpol JA, Kinzler KW, Vogelstein B. Nature Genetics 1992;1:45. [PubMed:
1301998]
5. Fields S, Jang SK. Science 1990;249:1046. [PubMed: 2144363]
6. Raycroft L, Wu H, Lozano G. Science 1990;249:1049. [PubMed: 2144364]
7. Farmer G, Bargonetti J, Zhu H, Friedman P, Prywes R, Prives C. Nature 1992;358:83. [PubMed:
1614538]
8. Kern SE, Pietenpol JA, Thiagalingam S, Seymour A, Kinzler KW, Vogelstein B. Science
1992;256:827. [PubMed: 1589764]
9. Unger T, Nau MM, Segal S, Minna JD. Embo Journal 1992;11:1383. [PubMed: 1314165]
10. Harris CC. Journal of the National Cancer Institute 1996;88:1442. [PubMed: 8841019]
11. Ko LJ, Prives C. Genes & Development 1996;10:1054. [PubMed: 8654922]
12. Prives C, Hall PA. Journal of Pathology 1999;187:112. [PubMed: 10341712]
13. Martin ACR, Facchiano AM, Cuff AL, Hernandez-Boussard T, Olivier M, Hainaut P, Thornton JM.
Human Mutation 2002;19:149. [PubMed: 11793474]
14. Olivier M, Eeles R, Hollstein M, Khan MA, Harris CC, Hainaut P. Human Mutation 2002;19:607.
[PubMed: 12007217]
15. Nikolova PV, Wong KB, DeDecker B, Henckel J, Fersht AR. Embo Journal 2000;19:370. [PubMed:
10654936]
16. Hupp TR, Sparks A, Lane DP. Cell 1995;83:237. [PubMed: 7585941]
17. Cho YJ, Gorina S, Jeffrey PD, Pavletich NP. Science 1994;265:346. [PubMed: 8023157]
18. Bullock AN, Fersh AR. Nature Reviews Cancer 2001;1:68.
19. Pavletich NP, Chambers KA, Pabo CO. Genes & Development 1993;7:2556. [PubMed: 8276238]
20. Min DH, Liu YS, Carbone I, Yang W. Journal of Chemical Physics 2007;126
21. Hainaut P, Mann K. Antioxidants & Redox Signaling 2001;3:611. [PubMed: 11554448]
22. Meplan C, Richard MJ, Hainaut P. Biochemical Pharmacology 2000;59:25. [PubMed: 10605931]
23. Vedani A, Huhta DW. Journal of the American Chemical Society 1990;112:4759.
24. Aqvist J, Warshel A. Journal of the American Chemical Society 1990;112:2860.
25. Stote RH, Karplus M. Proteins-Structure Function and Genetics 1995;23:12.
26. Wasserman ZR, Hodge CN. Proteins-Structure Function and Genetics 1996;24:227.
27. Hoops SC, Anderson KW, Merz KM. Journal of the American Chemical Society 1991;113:8262.
28. Lu DS, Voth GA. Proteins-Structure Function and Genetics 1998;33:119.
29. Pang YP. Journal of Molecular Modeling 1999;5:196.
30. Simons J, Jorgensen P, Taylor H, Ozment J. Journal of Physical Chemistry 1983;87:2745.
31. Csaszar P, Pulay P. Journal of Molecular Structure 1984;114:31.
32. Farkas O, Schlegel HB. Journal of Chemical Physics 1999;111:10806.
33. Lee CT, Yang WT, Parr RG. Physical Review B 1988;37:785.
34. Becke AD. Journal of Chemical Physics 1993;98:5648.
35. Hehre, WJ.; Radom, L.; Schleyer, PVR.; Pople, J. Ab initio molecular orbital theory. New York:
1986.
36. Pavlov M, Siegbahn PEM, Sandstrom M. Journal of Physical Chemistry A 1998;102:219.
Lu et al. Page 8
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

37. Pang YP, Xu K, Yazal JE, Prendergast FG. Protein Science 2000;9:2583.
38. Foresman, AFJB. Exploring Chemistry with Electronic Structure Methods. 2. Gaussian; 1996.
39. Momany FA. Journal of Physical Chemistry 1978;82:592.
40. Cox SR, Williams DE. Journal of Computational Chemistry 1981;2:304.
41. Singh UC, Kollman PA. Journal of Computational Chemistry 1984;5:129.
42. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, Cheeseman JR, Montgomery JA,
TVreven J, Kudin KN, Burant JC, Millam JM, Iyengar SS, Tomasi J, et al. Gaussian 03. 2003
43. Kottalam J, Case DA. Biopolymers 1990;29:1409. [PubMed: 2361153]
44. Case DA, Darden TA, Cheatham TEI, Simmerling CL, Wang J, Duke RE, Luo R, Merz KM, Pearlman
DA, Crowley M, Walker RC, Zhang W, Wang B, Hayik S, Roitberg A, Seabra G, Wong FK, Paesani
F, Wu X, Brozell S, Tsui V, Gohlke H, Yang L, Tan C, Mongan J, Hornak V, Cui G, Beroza P,
Mathews DH, Schafmeister C, Ross WS, Kollman aPA. AMBER 9. 2006
45. Merz KM. Journal of the American Chemical Society 1991;113:406.
46. Bayly CI, Cieplak P, Cornell WD, Kollman PA. Journal of Physical Chemistry 1993;97:10269.
47. Cornell WD, Cieplak P, Bayly CI, Kollman PA. Journal of the American Chemical Society
1993;115:9620.
48. Cieplak P, Cornell WD, Bayly C, Kollman PA. Journal of Computational Chemistry 1995;16:1357.
49. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. Journal of Chemical Physics
1983;79:926.
50. Wang JM, Cieplak P, Kollman PA. Journal of Computational Chemistry 2000;21:1049.
51. Lwin TZ, Luo R. Protein Science 2006;15:2642. [PubMed: 17075138]
52. Darden T, York D, Pedersen L. Journal of Chemical Physics 1993;98:10089.
53. Ryckaert JP, Ciccotti G, Berendsen HJC. Journal of Computational Physics 1977;23:327.
54. Berendsen HJC, Postma JPM, Vangunsteren WF, Dinola A, Haak JR. Journal of Chemical Physics
1984;81:3684.
55. Ryde U. Proteins-Structure Function and Genetics 1995;21:40.
56. Nakamoto K, Takemoto J, Chow TL. Applied Spectroscopy 1971;25:352.
57. Sacconi L, Gans P, Sabatini A. Inorganic Chemistry 1964;3:1772.
58. Nakamoto, K. Infrared and Raman Spectra of Inorganic and Coordination Compounds. New Youk:
1978.
59. Bullock AN, Henckel J, Fersht AR. Oncogene 2000;19:1245. [PubMed: 10713666]
60. Joerger AC, Ang HC, Fersht AR. Proceedings of the National Academy of Sciences of the United
States of America 2006;103:15056. [PubMed: 17015838]
61. Joerger AC, Ang HC, Veprintsev DB, Blair CM, Fersht AR. Journal of Biological Chemistry
2005;280:16030. [PubMed: 15703170]
Lu et al. Page 9
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 1.
Ribbon representation of chain B of 1TSR with the Zinc interface.
Lu et al. Page 10
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 2.
Optimized Zinc interface [Zn(CYS)3(HIS)1] in B3LYP/6-311+G**.
Lu et al. Page 11
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 3.
Superposition of optimized Zinc-binding interface structure between Amber and Gaussian.
Black is for Gaussian and white is for Amber. Root mean square deviation for all heavy atoms
is 0.94 Å.
Lu et al. Page 12
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 4.
Comparison of normal mode frequencies of the Zinc interface between Gaussian and Amber.
Lu et al. Page 13
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 5.
Cumulative average potential energies for DNA-free and DNA-bound p53c. The moving
window is 1ns.
Lu et al. Page 14
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 6.
(a) The black and grey lines are backbone RMSD of all the residues of DNA-bound and DNA-
free simulations respectively. (b) The black and grey lines are RMSD of the five atoms (ZN,
3S, N) at the Zinc-binding interface.
Lu et al. Page 15
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 7.
Superposition of mean simulated structures and 1TSR B chain. Right: DNA-free p53c; left:
DNA-bound p53c. Gray: crystal; black: simulation.
Lu et al. Page 16
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 8.
B-factors for DNA-bound p53c between simulation and crystal structure (chain B). P: for
crystal packing position. The four vertical lines mark the four residues at the Zinc-binding
interface.
Lu et al. Page 17
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 9.
B-factors for DNA-free p53c between simulation and crystal structure (chain A).
Lu et al. Page 18
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 10.
Diagrams of secondary structure interaction networks. Upper: DNA-free; lower: DNA-bound.
Lu et al. Page 19
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

Figure 11.
(a) B-factors between DNA-free and DNA-bound simulations. (b) B-factors in crystal structure
(chains A and B).
Lu et al. Page 20
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.
NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript

NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Lu et al. Page 21
Table 1
Possible total charges (a. u.) and their corresponding total energies (Hartree).
Total charge +2 +1 0 −1−2
HIS 0 0 0 0 −1
CYS 0 0 0 −1−1
CYS 0 0 −1−1−1
CYS 0 −1−1−1−1
Total energy −3518.046193 −3517.767790 −3517.386853 −3516.865598 −3516.231373
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.

NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Lu et al. Page 22
Table 2
Information of the p53 and p53 complex MD simulations.
With DNA Without DNA
Total Atoms 43058 24993
Protein & DNA Residues 218 194
Protein & DNA Atoms 3769 3007
Water Molecules 13329 7328
Counter Ions 22 Na+ 2 Cl−
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.

NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Lu et al. Page 23
Table 3
Empirically optimized force constants for ZN related bonds and angles.
This work Literature
Kb/Kθ, beq/θeq Kb/Kθ, beq/θeq
ZN-SG 60.0, 2.360 81.0, 2.29055
ZN-ND1 26.0, 2.270 40.0, 2.10056–58
SG-ZN-SG 32.0, 116.6 50.0, 146.555
SG-ZN-ND1 36.0, 100.8 30.4, 103.055
ZN-SG-CB 30.0, 103.5 19.0, 111.655
ZN-ND1-CE1 30.0, 113.0 20.0, 111.645
ZN-ND1-CG 30.0, 121.7 20.0, 111.645
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.

NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Lu et al. Page 24
Table 4
Ratios of B-factors between side chain and main chain heavy atoms at the Zn interface.
Residues Simulation Crystal
176 0.947 0.949
179 0.994 0.944
238 1.355 1.231
242 0.888 0.940
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.

NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Lu et al. Page 25
Table 5
Percentage occupancies of native contacts outside the β-sandwich region. The table only lists percentage occupancies that are higher than
50%. w DNA: DNA-bound simulation; w/o DNA: DNA-free simulation; IP: ion pair; HP: hydrophobic pair.
Secondary Structures Residues w/o DNA w DNA Type
H2 S2 E286 S127 97.1 98.0 IP
H2 S2′E285 K132 97.1 100.0 IP
H2 S2 R282 Y126 99.0 98.5 IP
H2 S2 R282 T125 100.0 93.5 IP
H2 S10 D281 R273 99.5 100.0 IP
L3 L2 R249 E171 100.0 99.5 IP
L3 S4 R249 Y163 99.2 95.1 IP
L3 S5 M246 L194 74.9 99.5 HP
L3 S9 M246 I251 100.0 100.0 HP
L3 ZN G245 C242 97.5 92.5 HB
L3 S10 S240 V274 99.5 97.5 HB
L3 S3 M237 A138 99.6 100.0 HP
L2′S8 L194 Y236 74.1 0.0 HP
L2′S6 P190 Y205 100.0 98.4 HP
L2′S6 A189 Y205 100.0 100.0 HP
L2′S6 L188 Y201 71.3 9.7 HP
L2′L5 L188 V203 53.0 92.0 HP
L2′S6 L188 Y205 99.2 100.0 HP
L2′S5 D186 R196 88.7 1.0 IP
L2′S5 D184 R196 91.5 1.0 IP
H1 S5 S183 R196 0.0 77.3 IP
L2 L2′R175 S183 0.0 87.9 IP
L2 L2′R175 D184 100.0 0.0 IP
L2 H1 R174 E180 100.0 1.0 IP
L2 S4 V173 A161 99.2 65.2 HP
L2 S4 V173 Y163 100.0 100.0 HP
L2 S5 V173 L194 100.0 100.0 HP
L2 S9 V173 I251 98.7 99.2 HP
L2 L3 V173 M246 91.5 83.0 HP
L2 L5′T170 R213 50.0 49.5 IP
L2 L5′M169 R213 98.5 1.0 IP
L2 S4 M169 I162 64.0 97.6 IP
L2 S10 K164 E271 100.0 98.4 IP
S2′S10 K132 E271 100.0 99.2 IP
S2 S2′Y126 N131 100.0 100.0 HB
L1 H2 T118 R283 27.0 100.0 IP
L1 S2 G117 T125 100.0 79.0 HB
L1 S2 S116 C124 100.0 42.5 HB
NT L5′V97 M169 91.5 43.0 HP
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.

NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Lu et al. Page 26
Table 6
Same as Table 5, but for native contacts inside the β-sandwich region.
Secondary Structure Residues w/o DNA w DNA Type
S3 S1 V143 L111 78.0 89.5 HP
S3 S8 V143 Y234 94.5 68.5 HP
S3 S9 V143 I255 66.5 4.0 HP
S3 S10 V143 F270 45.0 93.5 HP
S3 S7 V147 Y220 100.0 98.5 HP
S4 S3 R156 E258 100.0 100.0 IP
S4 S3 V157 L145 99.5 98.0 HP
S4 S9 R158 E258 100.0 100.0 IP
S5 S4 I195 A161 95.0 95.0 HP
S5 S8 I195 Y236 57.0 47.5 HP
S5 S8 I195 Y234 94.0 99.0 HP
S5 S8 E198 N235 0.0 99.6 HB
S3 S8 T140 N235 53.4 0.0 HB
S5 S3 E198 T140 97.6 100.0 IP
S5 S8 E198 H233 84.2 0.0 IP
L5 L6 R202 E221 91.9 92.7 IP
S6 L5 Y205 V203 98.0 100.0 HP
S6 S7 Y205 V216 100.0 99.0 HP
S9 S10 I251 V272 93.5 65.5 HP
S9 S1 I255 F109 94.5 100.0 HP
S9 S8 I255 Y234 100.0 100.0 HP
S9 L4 L257 P151 99.5 99.0 HP
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.

NIH-PA Author Manuscript NIH-PA Author Manuscript NIH-PA Author Manuscript
Lu et al. Page 27
Table 7
Same as Table 5, but for native contacts with DNA.
Secondary Structures Residues w DNA Type
L1 DNA K120 G7 96.8 IP
L1 DNA K120 G8 100.0 IP
L3 DNA N239 G13 100.0 HB
L3 DNA S241 T12 93.5 HB
L3 DNA S241 G13 89.1 HB
L3 DNA R248 T11 100.0 IP
L3 DNA R248 T12 81.4 IP
L3 DNA R248 T26 100.0 IP
L3 DNA R248 G25 65.2 IP
S10 DNA R273 T12 100.0 IP
H2 DNA R280 G13 100.0 IP
H2 DNA R283 G19 100.0 IP
J Phys Chem B. Author manuscript; available in PMC 2008 October 4.