Subscriber access provided by University of Texas Libraries
Article
Mechanistic Elements of Protein Cold Denaturation
†
Carlos F. Lopez, Richard K. Darst, and Peter J. Rossky
J. Phys. Chem. B, 2008, 112 (19), 5961-5967• DOI: 10.1021/jp075928t • Publication Date (Web): 09 January 2008
Downloaded from http://pubs.acs.org on February 22, 2009
More About This Article
Additional resources and features associated with this article are available within the HTML version:
•
•
•
•
•
Supporting Information
Links to the 5 articles that cite this article, as of the time of this article download
Access to high resolution figures
Links to articles and content related to this article
Copyright permission to reproduce figures and/or text from this article
The Journal of Physical Chemistry B is published by the American Chemical Society.
1155 Sixteenth Street N.W., Washington, DC 20036
J. Phys. Chem. B 2008, 112, 5961-5967
5961
Mechanistic Elements of Protein Cold Denaturation†
Carlos F. Lopez, Richard K. Darst, and Peter J. Rossky*
Center for Computational Molecular Sciences, Institute for Computational Engineering and Science, and
Department of Chemistry and Biochemistry, 1 UniVersity Station, A5300, The UniVersity of Texas at Austin,
Austin, Texas 78712-1167
ReceiVed: July 26, 2007; In Final Form: NoVember 1, 2007
Globular proteins undergo structural transitions to denatured states when sufficient thermodynamic state or
chemical perturbations are introduced to their native environment. Cold denaturation is a somewhat
counterintuitive phenomenon whereby proteins lose their compact folded structure as a result of a temperature
drop. The currently accepted explanation for cold denaturation is based on an associated favorable change in
the contact free energy between water and nonpolar groups at colder temperatures which would weaken the
hydrophobic interaction and is thought to eventually allow polymer entropy to disrupt protein tertiary structure.
In this paper we explore how this environmental perturbation leads to changes in the protein hydration and
local motions in apomyoglobin. We do this by analyzing changes in protein hydration and protein motion
from molecular dynamics simulation trajectories initially at 310 K, followed by a temperature drop to 278 K.
We observe an increase in the number of solvent contacts around the protein and, in particular, distinctly
around nonpolar atoms. Further analysis shows that the fluctuations of some protein atoms increase with
decreasing temperature. This is accompanied by an observed increase in the isothermal compressibility of the
protein, indicating an increase in the protein interior interstitial space. Closer inspection reveals that atoms
with increased compressibility and larger-than-expected fluctuations are localized within the protein core
regions. These results provide insight into a description of the mechanism of cold denaturation. That is, the
lower temperature leads to solvent-induced packing defects at the protein surface, and this more favorable
water-protein interaction in turn destabilizes the overall protein structure.
I. Introduction
Globular proteins represent a complex type of condensed
matter. They contain a core with solid-like packing and a surface
with liquid-like characteristics, while displaying a rich array of
motions that are crucial for protein function.1,2 Sufficient
perturbation of the native protein environment through heating,
cooling, pressurization, pH changes, or chemical denaturants
will lead to a loss of the native fold and protein function.3,4
The mechanisms behind denaturation are reasonably expected
to differ in these cases, and hence one can speculate that the
unfolded ensembles differ as well. For example, the ensemble
of structures just below the cold-denaturation temperature may
not closely resemble those above the heat-denaturation point,5
and pressure-induced denaturation may differ from these.6 Cold
denaturation is thought to proceed largely as a result of changes
in the nature of the interaction between water and nonpolar
groups.5,7,8 As temperature is decreased, the free energy penalty
of the entropically unfavorable interaction between water and
hydrophobes becomes smaller, leading to increased nonpolar
group hydration.4,5 Hence, the global loss of tertiary structure
stability reflects a weakening of hydrophobic associations.
Although the thermodynamic nature of this process is very
reasonable, the associated mechanical processes have not been
elucidated. These processes are the focus of the present study.
A large body of experimental evidence shows that globular
proteins generically possess a cold unfolding temperature.3,4,5
It has been suggested that the cold unfolding of globular proteins
†
Part of the “Attila Szabo Festschrift”.
* Corresponding author. E-mail: [email protected].
can be well described by a two-state process between a native
state and a denatured state, thus implying an all-or-nothing,
single-step mechanism for the collapse of protein tertiary
structure, a notion that has been supported by experiments.5 For
example, Gruebele and co-workers used temperature jump
experiments to show that the cold unfolding of apomyoglobin
(apoMB) is tightly related to the stability of the hydrophobic
core formed by the A-, G-, and H-helices. At temperatures below
the cold-denaturation point, the A-helix detaches from the Gand H-helices, and unfolding of the protein follows. These
authors have also suggested the appearance of a “pre-transition”
prior to unfolding, corresponding to the creation of a relatively
conformationally loosened native state.5 A recent and important
low-temperature NMR study by Wand and co-workers expands
on the all-or-nothing mechanism that has been accepted for cold
denaturation.7 Using micelle encapsulation to allow supercooling, Wand and co-workers showed that the cold unfolding of
ubiquitin is not completely cooperative and can take place in
steps. Residual structure was detected even after the cold
unfolding steps had taken place, thus suggesting that the cold
unfolding process does not imply complete structural loss in
ubiquitin. The aforementioned experimental studies stress the
importance of understanding cold denaturation as a means to
extract the cooperative substructures of more complex protein
systems and as a way of understanding the possible pathways
leading to protein folding and stability.
From a theoretical perspective, the cold-denaturation process
is thought to be tightly linked to the change in affinity between
nonpolar groups and water at low temperatures. At cold
temperatures, the solvation of nonpolar groups in water occurs
10.1021/jp075928t CCC: $40.75 © 2008 American Chemical Society
Published on Web 01/09/2008
5962 J. Phys. Chem. B, Vol. 112, No. 19, 2008
more readily.9 This change in the nonpolar-water interaction
is thought to cause packing changes and the eventual break up
of hydrophobic associations. Theoretical work on cold denaturation has been carried out with protein-like and reduced
representation techniques using constraint models,10 mixed
neighbor water bonding models,11 and lattice Monte Carlo
methods12 among others. Other recent work has aimed to
understand changes in hydrophobic hydration at different
temperatures, with a focus on the general aspects of the
phenomenon.13 These previous studies have focused on explaining the origin of the thermodynamic aspects of the cold
unfolding process rather than on the actual mechanism of protein
cold denaturation. Here we explore how the changes in
hydrophobic hydration with temperature and protein cold
denaturation are linked mechanistically at the molecular level.
Extensive experimental14 and theoretical15,16 data exists
describing the structural properties of apoMB, making it a wellunderstood and suitable protein for simulations such as that
carried out in this work. The availability of experimental cold
unfolding data from Gruebele and co-workers5 further support
this choice. Given intrinsic simulation time limitations, we do
not attempt to study the complete unfolding process here. Rather,
we aim to probe the initial stages of the cold-denaturation
process in hopes that we can observe a mechanistic description
of the early events in the cold unfolding process. To this end,
molecular dynamics (MD) simulations of sperm whale (Physeter
catodon) apoMB were carried out at 310 K and 278 K. The
former temperature lies within the experimental temperature
range of the protein within its native environment, while the
latter is chosen to be slightly above the experimental cold
unfolding temperature range of 263-276 K.5
The paper is organized as follows. The Methods section
contains detailed information about our simulation protocol and
analysis methodology. This is followed by the Results and
Discussion section in which we examine our results and further
comment on their significance. The Concluding Remarks section
follows, where a summary of the findings and their possible
impact is discussed.
II. Methods
In this section we describe the details of the calculations
carried out, including elements of the structural preparation,
simulation protocols, and analysis tools.
A. Apomyoglobin Structure. The structure of apoMB and
holomyoglobin (holoMB) are very similar. Besides the obvious
absence of the prosthetic group, apoMB has a somewhat lower
helical content17 and a slightly increased radius of gyration
relative to the holoprotein.18 However, the tertiary structure of
the apoprotein appears to be largely preserved when compared
to the holoprotein.14 The apoprotein differs from the holoprotein
by an EF-F-FG region that is not well resolved via NMR
measurements and is thus thought to exhibit large spatiotemporal
variations in solution.14 Decreased secondary structure near the
C-terminus of the H-helix is also observed in the apoprotein.
Since no three-dimensional structure exists for apoMB in the
protein data bank (PDB), it was necessary to carry out some
preliminary MD simulation to allow the protein to evolve to a
structure representative of its apo state. The holoprotein neutron
structure19 (PDB ID: 1CQ2; pH 6.2) was downloaded from the
PDB. The pH, and the corresponding protonation state of
histidine in particular, is important to the unfolding behavior
of this protein, and this pH value is close to that (pH 5.9) where
the bulk of the cold unfolding experiments in ref 5 were carried
out. The protonation state of residues was fixed throughout the
Lopez et al.
simulations, as that reported with this neutron structure. A table
of the precise protonation states and corresponding force field
designations for these residues is provided as part of the
Supporting Information. The heme was removed, and the protein
was solvated by 31 060 water molecules in a periodic box with
edges measuring 100 Å. Seven water molecules at least 15 Å
away from the protein were replaced by negatively charged
chloride ions using the “autoionize” utility of the VMD
molecular manipulation package to balance out the protein
charges. The protein was then held fixed, and a short minimization was carried out to relieve bad contacts between molecules.
All constraints were then removed, and the system was run for
almost 2 ns at a temperature of 510 K to accelerate relaxation
of the system toward the apoprotein conformation. The protein
was then allowed to relax at 310 K for another 1 ns before the
production simulations were run. The final structure of apoMB
used in the present study has a disrupted secondary structure in
the EF-F-FG region, exhibits a loss of secondary structure near
the C-terminus of the H-helix, and has a ∼55% helical content,
as calculated with the DSSP20 program. The helical content of
the apoprotein for the present simulation runs compares favorably with the measured experimental helical content of ∼55%
for apoMB.17 For reference, the holoprotein has been measured
to be composed of an ∼78% helical structure.17 We also
calculate the root-mean-square deviations (RMSDs) of the
protein and of individual secondary structure protein segments
relative to its average structure during the run to measure the
structural fluctuations of the protein structure during simulation
(data not shown). All secondary structure segments of the
protein, excluding the EF-F-FG region (5 Å), the C-terminus
region of the H-helix (7 Å), and the B-helix (3.5 Å), exhibit an
RMSD value of 2 Å or less. The experimental radius of
gyration18 is 19 Å. This value compares satisfactorily with the
simulation average radius of gyration of 16 Å, when one
considers that experiments appear to overestimate the value of
the radius of gyration by about 2 Å.18 The distance between
fluorophores in the A- and G-helices has been reported21 to be
24 Å for horse apoMB and has been shown to be of similar
magnitude for the holo and apo forms of the protein. We
measure the average A-helix to G-helix distance at 22 Å for
the present apoMB structures. This value compares favorably
with the PDB myoglobin distance of 22.7 Å.19 The present data
indicates that the apoMB structure employed in the present
simulation study compares well with that reported in previous
experiments for apoMB.
B. Simulations. B.1. Production Runs. Two production
simulations were run for the apoprotein. The first was run at 1
atm pressure and 310K for 1.1 ns (following the relaxation
process described above). The last 1 ns of the run was used for
analysis with structures collected every 1 ps for a total of ∼103
structures. A second simulation, with a starting structure taken
from the end of the relaxation simulation described above, was
run at 278 K and 1 atm pressure for 2 ns. The last 1 ns (∼103
structures) was used for analysis. We provide, as three figures
in the Supporting Information, results for the energy and local
structure in each of these production runs, demonstrating the
stability of these simulations and showing that the analysis
represents that of a state that is sufficiently steady on this time
scale.
B.2. Simulation Protocols. All the protein simulations were
performed with the NAMD (v. 2.5)22 program and using the
CHARMM (v. 27)23 force field for the protein and SPC/E24
model for water. Protein manipulations and illustrations were
prepared with the VMD (v. 1.8.3) program, and with our in-
Mechanistic Elements of Protein Cold Denaturation
J. Phys. Chem. B, Vol. 112, No. 19, 2008 5963
house mp3 analysis package.25 The temperature and pressure
were controlled using the Langevin thermostat and the NoseHoover Langevin piston, respectively, as implemented in
NAMD. Nonbond short-range terms were included using a 12
Å cutoff with a smooth switching function starting at 10 Å.
The particle mesh Ewald method was used to include longrange electrostatic interactions. The mesh spacing was kept near
1 Å in each direction throughout the simulation runs. To increase
the efficiency of the simulations, the RESPA multiple time step
integrator scheme was used,26 with an inner step of 1 fs, a
nonbonded time step of 2 fs, and an outer time step of 4 fs.
The SHAKE algorithm was used to constrain the geometry of
water molecules. The total system for the production runs
contained 95 630 atoms.
C. Simulation Analysis. C.1. Isothermal Compressibility. The
isothermal compressibility (IC) is directly linked to density
fluctuations in a material. Correspondingly, the rigidity of
packing in a material, including a protein, is reflected in the
compressibility, as exploited by Dadarlat and Post.31 A suitable
definition of a local compressibility can be constructed as well.
The IC was calculated using the formalism described in ref 31,
with the difference that the volumes associated with each atom
were calculated using a Voronoi tessellation procedure as
described by Gerstein27 and employing “method 2” in the
program distributed from the Gerstein laboratory. The IC in the
isothermal-isobaric ensemble is given by31
IC ) 〈∆V2〉NPT/kBT〈V〉NPT
(1)
In order to calculate the IC, the average and variance of the
protein (or protein segment) volumes must be obtained. To
calculate the protein volumes, the volume of each protein atom
and neighboring water (within 15 Å from the protein) was
estimated by using one Voronoi cell per atom. The volume
associated with protein atoms was then summed to estimate the
total volume of the protein segment. This procedure was
repeated for each simulation frame. The volume average and
variance were then estimated from the obtained volume data.
A local isothermal compressibility difference (ICD) for clusters
is calculated as follows: A 3 Å cluster of atoms is defined
around each atom in the protein at time zero of the analysis
trajectory. The IC of each cluster is then calculated both at 310
K and at 278 K. The difference yields ICD ) IC278K - IC310K,
where the subscripts indicate temperature. This allows us to
estimate the local change in IC in the protein in going from the
high temperature to the low temperature.
C.2. SolVent Contacts. A solvent contact is defined by a water
molecule that is within a specified distance cutoff (rcut). We
use rcut ) σO + 0.5σA, where σO, denotes the Lennard-Jones
sigma value for the SPC/E water oxygen,28 and σA denotes the
sigma value from the CHARMM force field29 for the protein
atom type being tested. Atoms are further classified into polar
or nonpolar on the basis of their force field charge value. Atoms
with |q| > 0.29 are classified as polar, while the remainder are
classified as nonpolar. All values reported are averaged over
the last 1 ns of the production runs at 278 K and 310 K.
C.3 Root-Mean-Square DeViation. The RMSD is calculated
in the usual manner, relative to the average structure of the
protein during the simulation run. The average structure for each
simulation run is calculated by aligning all protein atoms to a
reference structure (here, the first structure in the trajectory).
The alignment is performed by minimizing the RMSD score of
each protein structure in the trajectory using the simplex
minimization method.30 The protein is translated and rotated
while keeping the overall protein structure rigid until the RMSD
Figure 1. (A) Sum of SCD by atom type (atoms with charge, |q| >
0.29e are defined as polar and marked with “*”); (B) rRMSDd by
protein atom (see text; the red line at -0.053 indicates the expected
decrease for a harmonic oscillator).
between the reference structure and the aligned structure has
reached a minimum. Once all the frames in the trajectory have
been aligned, the average structure is found by determining the
average position for each atom in the protein.
C.4. RelatiVe RMSD Difference. The relative RMSD difference (rRMSDd) is defined as follows: the RMSD, 〈∆x2〉1/2, is
calculated for each atom as described above at 278 K and 310
K for each atom. The ratio of the difference between the two
temperatures and the initial temperature is then evaluated to
give
rRMSDd ) (〈∆x2〉278K1/2 - 〈∆x2〉310K1/2)/〈∆x2〉310K1/2
(2)
where the subscripts indicate the corresponding temperatures
for the calculation. These mean square displacements per atom
are calculated from the last 1 ns of data from the 278 K and
310 K simulations. The relationship in eq 2 has the convenient
property that, for a classical harmonic system, it is only a
function of temperatures, namely ∆T/T.
III. Results and Discussion
A. Solvation. To test the extent of solvation change, we
proceed to calculate the protein-water solvent contacts at both
278 K and 310 K as discussed in the Methods section. We
5964 J. Phys. Chem. B, Vol. 112, No. 19, 2008
Lopez et al.
Figure 2. Representative snapshots indicative of the observed increase in the number of solvent contacts around the hydrophobic side chain
moieties of Arg45 (top row) and Val114 (pink, bottom row) when changing the temperature from 310 K (left column) to 278 K (right column). Atom
colors: carbon, cyan; oxygen, red; nitrogen, blue; water oxygen, yellow. Hydrogen not shown for clarity. A cavity that could accommodate three
water oxygen atoms at 310 K (top left) can accommodate five water oxygen atoms at 278 K (top right). A previously buried Val moiety at 310 K
(bottom left) becomes solvent exposed at 278 K (bottom right).
consider the solvent contact with individual protein atoms first.
To quantify the change, we group these by atom type, and plot
the solvent contact difference (SCD) as shown in Figure 1A.
The value is the sum over all examples of each atom type in
the protein. The nonpolar hydrogen (HA), as well as the aliphatic
type carbon atoms (CT1, CT2, CT3), and the aromatic carbon
(CA) exhibit an increase in the number of solvent contacts
ranging from 28 to almost 100. Peptide bond atoms (C, NH1,
and O) show an increase of 16-20 solvent contacts per atom
type. It is worth noting that the observed changes in solvent
contacts appear to occur very rapidly, during the first ∼50 ps
following the temperature drop. This provides numerical
evidence that, as anticipated, water molecules have an increased
probability of contact with the nonpolar protein moieties at lower
temperature, in contrast to polar side chain moieties.
The total number of solvent molecules hydrating the protein
exhibits only a modest increase (∼15 gained) upon cooling, but
distinct changes are evident for individual protein moieties. An
example is shown for Arg45 in Figure 2. At 310 K, the
guanidinium group of Arg45 is fully solvated, but the aliphatic
chain of Arg45 is buried within a hydrophobic pocket. At the
lower temperature, a snapshot from the end of the simulation
manifests water molecules contacting atoms previously inaccessible within this hydrophobic pocket. As a second example,
the extent of hydrophobic hydration of Val114 at the low
temperature is depicted in Figure 2. The hydrophobic amino
acid Val114 resides within a hydrophobic pocket at the higher
temperature and is almost completely inaccessible to water
contact. At the lower temperature, Val114 is no longer sequestered in the hydrophobic pocket and becomes solvated by water
molecules. Water molecules therefore evidently modify contacts
between surface hydrophobic moieties at the protein-water
interface at the low temperature.
B. Structural Fluctuations. Given the observed changes at
the interface, we now aim to understand whether and how these
changes affect the protein structural fluctuations. Following the
reasoning of Post and co-workers,31 the relative stabilization of
the protein hydrophobic group-water interfacial interaction
could lead to a lowered surface tension and increased protein
spatial fluctuations. Given that there is a relation between the
change in interstitial space (i.e., in the number density) and the
compressibility and, as evidenced by Post and co-workers,31 a
correlation between protein structural stability and compressibility, we shift our attention to examine these aspects below.
To understand these mechanical aspects, we first consider
the rRMSDd as described in the Methods section. The rRMSDd
is the difference between the time averaged root-mean-square
displacement per atom at the low temperature and the high
temperature, normalized by the high-temperature simulation.
This quantity provides the fractional change in the atomic
RMSD. As a point of reference, we include this quantity for a
harmonic oscillator; the rRMSDd for such an oscillator is
-0.053. That is, we might expect a 5.3% decrease in the RMSD
in going from 310 K to 278 K. The calculation of the rRMSDd
Mechanistic Elements of Protein Cold Denaturation
J. Phys. Chem. B, Vol. 112, No. 19, 2008 5965
Figure 3. The IC of myoglobin and its secondary structural components
at 310 K (red dots) and 278 K (black dots). The E-helix, G-helix,
A-core, G-core, and AGH-core sections, along with the whole protein
exhibit an increase in IC upon cooling.
per atom is shown in Figure 1b. Atoms with rRMSDd values
above the harmonic oscillator reference line (∼130 atoms in
total) exhibit larger-than-expected fluctuations at the lower
temperature. Given the thermal energy decrease in going to the
low-temperature simulation, an increase in some protein atom
fluctuations is particularly notable.
We further characterize the internal motions of the protein
by measuring the Lindemann criterion for apoMB at the high
and low temperatures. The Lindemann criterion (∆L) was
originally defined for crystals by comparing the atomic fluctuation amplitudes to the lattice constant of the crystal. The ratio
can then be used to characterize the structure as solid-like or
liquid-like.1 A ∆L less than 0.1 is typical of solid crystal
structures, while a ∆L value greater than 0.1 is characteristic of
melting. The approach has been used to characterize the state
of a protein.1 We follow the procedure described in previous
work by Zhou et al.1 to calculate ∆L for apoMB from both the
high- and low-temperature simulations. We find that, even at
the low temperature, the protein core is still solid-like (∆L )
0.09 < 0.1), and the exterior is liquid-like (∆L ) 0.14 g 0.1).
The ∆L values obtained in the present simulations are in good
agreement with those obtained in ref 6 for several proteins, even
at the low temperature. We do not observe, in this simulation,
indications of protein core melting based on the calculation of
∆L, and one should not expect to if the simulated temperature
corresponds to experiment.
The previous analysis leads us to consider the changes in
the protein interstitial space. Unless motions are highly concerted, an increase in interior atomic fluctuation amplitudes must
necessarily be accompanied by an increase in free space.
Previous work31 has postulated that favorable interactions at the
interface will result in a lowered barrier to protein volume
fluctuations and therefore an observed increase in density
fluctuations or IC.31 Given the observed more favorable contact
between water and nonpolar groups at the low temperature, we
quantify the change in protein interstitial space more closely.
We report the IC for the 310 K apoMB simulation and the
278 K apoMB simulation in Figure 3 for the whole protein and
different components of the protein secondary structure, including the AGH-core region and its secondary structure components. The volume of apoMB remains essentially constant with
an ∼1% volume decrease upon cooling. When considering the
IC of individual secondary structure elements of apoMB, only
the G-helix exhibits an increase in compressibility. However
the changes in the compressibility for the whole protein, for
the AGH-core as a whole, and for the core segments of the A-
Figure 4. RDFs for atoms grouped by (A) their fractional RMSD
difference relative to the COM; (B) their change in IC; (C) distribution
of atoms in groups relative to each other.
and G-helices do exhibit an increase. The overall change in the
protein compressibility confirms that there is an increase in the
internal atomic fluctuations. This increase in the core region is
suggestive of a developing mechanical instability in this region.
The observed IC increase in the apoMB core region upon
cooling appears to correlate well with the observations by
Gruebele and co-workers,5 suggesting that a breakdown of the
AGH-core region is a key step in cold denaturation of apoMB.
We clearly see indications of reduced stability in this core region
in our results. To visualize the local changes within the protein
more clearly we consider the IC for spherical volume elements
of 3 Å radius centered about each heavy atom. We observe that
∼90 volume elements display increased IC upon cooling.
More quantitatively, we first explore the spatial distribution
of elements with increased IC and atoms with increased
rRMSDd upon cooling through radial distribution functions
(RDFs) relative to the protein center of mass (COM) as shown
in Figure 4A,B. As shown, atoms with increased rRMSDd upon
cooling tend to be located closer to the COM of the protein
than an average atom, indicating prevalence in the core region
of the protein. Figure 4B depicts RDFs for atom centers
5966 J. Phys. Chem. B, Vol. 112, No. 19, 2008
Lopez et al.
loosened, more highly fluctuating native state,5 is consistent with
the mechanism outlined here.
A more thorough study of cold denaturation including a
sequence of temperatures spanning the transition point would
clearly be of interest. Work along these lines is being pursued
in our laboratory. Particularly interesting parallel investigations
are the examination of similarities and differences between
thermal (cold- or heat-induced) denaturation and pressureinduced denaturation. For the latter, the leading hypothesis for
the mechanism6 involves, as an initial step, solvent penetration
between native hydrophobic contacts. Denaturation could then
proceed mechanistically similarly to that described here for cold
denaturation. Work along these lines is being pursued in our
laboratory.
Figure 5. Front view (A) and side view (B) of the averaged apoMB
structure from MD simulations at 278 K. The yellow dots represent
atoms (hydrogen not shown for clarity) that have increased fluctuation
amplitudes upon cooling. The semi-transparent blue spheres are centered
on 3 Å radius regions of the protein with increased local IC upon
cooling (see text). Ovals highlight core regions of the protein that
possess increased compressibility and increased motions upon cooling.
classified based on their ICD upon cooling. A pattern similar
to that seen in panel A is seen; atoms with increased ICD are
located near the COM of the protein, again indicating prevalence
in core regions. Figure 4C depicts the RDF for atoms with
increased ICD with respect to those with increased rRMSDd.
As shown, these atoms tend to exist near each other, thus
indicating that atoms with increased positional fluctuations are
in fact near the loci of increased compressibility. To complement
the RDF data, we depict atoms with increased ICD and increased
rRMSDd on the protein structure in Figure 5. As shown by the
RDFs, these atoms tend to reside both near the core of the
protein and in close proximity to each other. We see clear
evidence of increased fluctuations developing between the Gand H-helices, near the AGH-core, and near the C-terminus end
of the H-helix where it comes in contact with other helices and
the heme group in the holoprotein. The images clearly show
that regions with increased compressibility and regions containing atoms with increased positional fluctuations are in close
proximity. On the basis of the graphic, it is also possible to
visually confirm that atoms with increased fluctuations do not
lie near the solvent-accessible surface.
IV. Concluding Remarks
We conclude by discussing the mechanistic elements for cold
unfolding of globular proteins that can be inferred from the data
presented in this work. When the temperature is lowered, there
is an enhancement of hydrophobic group solvation. This causes
packing changes at the water-protein interface, as water
penetrates interstitial spaces at the interface. A lowered free
energy penalty for protein interior structural fluctuations is
observed in concert. We infer that internal fluctuations are a
direct result of the peripheral penetration of water and can lead
to the development of mechanical instability within the protein
core and an eventual breakdown of tertiary associations. As such,
the breakdown of the protein would occur in steps, depending
on the relative stability of individual hydrophobic protein core
contacts. In the case of a small protein with a single core region,
such as apoMB, the unfolding would take place in a single step.
Previous observations of single5 or multiple step7 denaturation
events are therefore compatible with this interpretation of a
denaturation mechanism. The proposal of a “pre-transition” prior
to unfolding, corresponding to the creation of a conformationally
Acknowledgment. We are grateful for generous grants from
the NSF Collaborative Research in Chemistry Program (CHE0404695), the R. A. Welch Foundation (F-0019), as well as for
computational support of the Texas Advanced Computing
Center. C.F.L. acknowledges the useful insight of Profs.
Alexander D. Mackerell and Steve O. Nielsen during the course
of this work.
Supporting Information Available: Total, potential and
kinetic energies of the 1 ns production segments of the apoMB
simulations; RMSDs of segments of the apoprotein vs time;
protonation state of HIS in the simulated apoMB. This material
is available free of charge via the Internet at http://pubs.acs.org.
References and Notes
(1) Zhou, Y.; Vitkup, D.; Karplus, M. J. Mol. Biol. 1999, 285, 1371.
(2) Frauenfelder, H.; Petsko, G. A.; Tsernoglou, D. Nature 1979, 280,
558.
(3) Smeller, L. Biochim. Biophys. Acta 2002, 1595, 11.
(4) Privalov, P. L. Crit. ReV. Biochem. Mol. Biol. 1990, 25, 281.
(5) Sabelko, J.; Ervin, J.; Gruebele, M. J. Phys. Chem. B 1998 102,
1806. Ervin, J.; Larios, E.; Osvath, S.; Schulten, K.; Gruebele, M. Biophys.
J. 2002, 83, 473.
(6) Hummer, G.; Garde, S.; Garcia, A. E.; Paulaitis, M. E.; Pratt, L.
R. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 1552.
(7) Babu, C. R.; Hilser, V. J.; Wand, A. J. Nat. Struct. Mol. Biol. 2004,
4, 352.
(8) Frauenfelder, H.; MacMahon, B. Proc. Natl. Acad. Sci. U.S.A. 1998,
95, 4795.
(9) Widom, B.; Bhimalapuram, P.; Koga, K. Phys. Chem. Chem. Phys.
2003, 5, 3085.
(10) Jacobs, D. J.; Wood, G. G. Biopolymers 2004, 75, 1.
(11) Robinson, G. W.; Cho, C. H. Biophys. J. 1999, 77, 3311.
(12) Choi, H. S.; Huh, J.; Jo, W. H. Biomacromolecules 2004, 5, 2289.
(13) Paschek, D. J. Chem. Phys. 2004, 120, 6674.
(14) Eliezer, D.; Wright, P. E. J. Mol. Biol. 1996, 263, 531.
(15) Onufriev, A.; Case, D. A.; Bashford, D. J. Mol. Biol. 2003, 325,
555.
(16) Tirado-Rives, J.; Jorgensen, W. L. 1993, 32, 4175.
(17) Hughson, F. M.; Wright, P. E.; Baldwin, R. L. Science 1990, 249,
1544.
(18) Eliezer, D.; Jennings, P. A.; Wright, P. E. Science 1995, 270, 487.
(19) Shu, F.; Ramakrishnan, V.; Schoenborn, B. P. Proc. Natl. Acad.
Sci. U.S.A. 2000, 97, 3872.
(20) Kabsch, W.; Sander, C. Biopolymers 1983, 22, 2577.
(21) Tcherkasskaya, O.; Ptitsyn, O. B. FEBS Lett. 1999, 455, 325.
(22) Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.;
Villa, E.; Chipot, C.; Skeel, R. D.; Kale, L.; Schulten, K. J. Comput. Chem.
2005, 26, 1781.
(23) MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack, R. L.;
Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S. J. Phys.
Chem. B 1998, 102, 3586.
(24) Berendsen H. J. C.; Grigera, J. R.; Straatsma, T. P. J. Phys. Chem.
1987, 91, 6269.
(25) Darst, R. K.; Dowdle, J. R.; Lopez, C. F.; Rossky, P. J. MP3: A
Package for Data Analysis, 2006.
(26) Tuckerman, M. E. J. Phys. Chem. B 2000, 104, 159.
(27) Gerstein, M.; Tsai, J.; Levitt, M. J. Mol. Biol. 1995, 249, 955.
Mechanistic Elements of Protein Cold Denaturation
(28) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. J. Phys. Chem.
1987, 91, 6269.
(29) MacKerell, A. D., Jr.; Bashford, D.; Bellott, M.; Dunbrack, Jr., R.
L.; 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.;
Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.; Roux,
J. Phys. Chem. B, Vol. 112, No. 19, 2008 5967
B., III; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub, J.; Watanabe, M.;
Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M. J. Phys. Chem. B 1998,
102, 3586.
(30) Nelder, J. A.; Mead, R. Comput. J. 1965, 7, 308.
(31) Dadarlat, V. M.; Post, C. B. Proc. Natl. Acad. Sci. U.S.A. 2003,
100, 14778.