Protein–Protein Interactions Affect Alpha Helix Stability in Crowded

Recently, it has been recognized that attractive protein–crowder interactions play .... Each residue is assigned a hydrophobicity ε (0 ≤ ε ≤ 1...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/JPCB

Protein−Protein Interactions Affect Alpha Helix Stability in Crowded Environments Bryanne Macdonald,† Shannon McCarley,† Sundus Noeen,† and Alan E. van Giessen* Department of Chemistry, Mount Holyoke College, 50 College Street, South Hadley, Massachusetts 01075, United States S Supporting Information *

ABSTRACT: The dense, heterogeneous cellular environment is known to affect protein stability through interactions with other biomacromolecules. The effect of excluded volume due to these biomolecules, also known as crowding agents, on a protein of interest, or test protein, has long been known to increase the stability of a test protein. Recently, it has been recognized that attractive protein−crowder interactions play an important role. These interactions affect protein stability and can destabilize the test protein. However, most computational work investigating the role of attractive interactions has used spherical crowding agents and has neglected the specific roles of crowding agent hydrophobicity and hydrogen bonding. Here we use multicanonical molecular dynamics and a coarse-grained protein model to study the folding thermodynamics of a small helical test protein in the presence of crowding agents that are themselves proteins. Our results show that the stability of the test protein depends on the hydrophobicity of the crowding agents. For low values of crowding agent hydrophobicity, the excluded volume effect is dominant, and the test protein is stabilized relative to the dilute solution. For intermediate values of the crowding agent hydrophobicity, the test protein is destabilized by favorable side chain−side chain interactions stabilizing the unfolded states. For high values of the crowding agent hydrophobicity, the native state is stabilized by the strong intermolecular attractions, causing the formation of a packed structure that increases the stability of the test protein through favorable side chain−side chain interactions. In addition, increasing crowding agent hydrophobicity increases the “foldability” of the test protein and alters the potential energy landscape by simultaneously deepening the basins corresponding to the folded and unfolded states and increasing the energy barrier between them.



INTRODUCTION The cellular environment in which proteins fold and function is crowded with biomacromolecules and is known to affect protein stability relative to dilute solution.1 Much of the work investigating the effect of a crowded environment on the stability of proteins has focused on macromolecular crowding by using large, inert macromolecules to reproduce the crowded cellular milieu.2 Experimental investigations of macromolecular crowding frequently use large, neutral, and inert molecules such as Ficoll or dextran as crowding agents.3 In theoretical and computational studies, these are usually represented as inert spheres that have no interactions with the protein other than excluded volume interactions.4 It is now widely accepted that it is essential to include the attractive interactions between the protein of interest, the “test protein,” and other cellular macromolecules, referred to as “crowding agents”.5−11 Of the computational work that includes attractive interactions between the test protein and crowding agents, most continue to use large unphysical spheres as crowding agents, with either single7,12 or multiple13 interaction sites per crowder with a fixed attractive interaction. The only work to vary the strength of the attraction between the test protein and the (spherical) crowders7,12 focuses on protein association and not stability. © XXXX American Chemical Society

Recently, efforts have been made to include both attractive interactions and excluded volume effects by studying protein stability in a more realistic cellular-like environment in silico,9,11,14−19 in vitro,6,10 and in vivo.5,20−25 In addition, it has been shown that the hydrophobic nature of a chaperonin cavity can affect the stability of the enclosed protein.26−28 These studies have shown that the stability of a protein is affected by the chemical nature of its environment, not just by the presence of crowding agents. This fact gains importance when one takes into account the heterogeneity of the cellular environment, in terms of both the molecular composition of the cell and its spatial heterogeneity. The molecular composition of the cytoplasm is not uniform but instead varies at different areas in the cell.29 A protein that is stable in one region of the cell may be unstable in another. Introducing chemical specificity by using proteins as crowding agents is a necessary step toward a more accurate representation of the cellular environment. However, there are as yet no systematic Received: December 18, 2014 Revised: January 13, 2015

A

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B investigations into how variances in the strength of protein− crowder interactions affect protein stability. To close this critical gap, we focused on the stability of αhelices, a common structural motif, in crowded environments by simulating a short helical test protein in the presence of crowding agents that were themselves proteins. The test protein interacted with the crowding agents though hydrophobic side chain−side chain interactions, hydrogen bonding, and van der Waals interactions. We varied the strength of the protein−protein hydrophobic attraction by systematically changing the crowder sequence. The test protein, (AAQAA)3, is a small helical protein for which the folding transition has been studied both experimentally and computationally, so its thermodynamic and structural properties in dilute solution are known.30−34 Its small size and known characteristics make it an attractive test protein for evaluating the effect of molecular crowding on the stability of α-helices. We examine the stability of the test protein through determination of its thermal stability, as measured by its folding temperature, and through its thermodynamic stability, the change in free energy of unfolding. To understand how molecular crowding affects the stability of a test protein, it is necessary to disentangle the various contributions due to attractive interactions, crowder size and shape, and solvent effects. This work focuses on the role played by the crowding agent hydrophobicity. Other effects, such as those due to sequence heterogeneity, crowder size polydispersity, and solvent effects, are not present. As discussed in detail later, the crowding agents used in these simulations are homopeptides of uniform length and shape. (Each crowder is fixed in an α-helical configuration.) While this represents a significant advance over the use of physically unrealistic spherical crowders, it still deviates from the true cellular environment in its uniformity. However, it is that very uniformity that is necessary to isolate the role played by the crowding agent hydrophobicity in affecting the stability of the test protein. Consequently, the results of this work are intended to shed light on one component of why protein crowders can destabilize test proteins while synthetic crowders often stabilize them6 in terms of the strength of the crowder−test protein interactions.

Figure 1. Schematic representation of the protein model. Each residue is represented by four interaction sites: three backbone sites and one side-chain site. All side chains except glycine have identical sizes.

have the same size and shape, when changing the sequence of the crowding agents, we change only the strength of the hydrophobic interaction and need not worry about the combined effects of size, shape, and hydrophobicity. The simulations were run using the replica exchange statistical temperature molecular dynamics (RESTMD) algorithm.39,40 This algorithm is a member of the Wang−Landau family of “flat histogram” methods employing a non-Boltzmann weight.41 The density of states is determined during the course of the simulation by a continual and dynamic refinement of the statistical temperature, T(U). The statistical temperature is related to the density of states via ⎡ ∂S(U ) ⎤−1 ⎡ ∂ ln Ω(U ) ⎤−1 T (U ) = ⎢ ⎥ =⎢ ⎥ ⎣ ∂U ⎦ ⎣ ⎦ ∂U

(1)

where we have taken kB = 1. The RESTMD algorithm divides the overall temperature range of interest into M subranges, with one replica α (1 ≤ α ≤ M) per subrange, and determines Tα(U) within that subrange. As the simulation progresses, Tα(U) is updated on a progressively finer scale, the magnitude of which is related to the rescaling parameter f. Rescaling occurs through the operation f → √f every 2.5 × 107 timesteps. The initial value was f = 1.0001 for all simulations, resulting in a total of 2.5 × 108 timesteps of equilibration. Once converged, each replica samples its entire temperature subrange with a generalized weight Wα = exp{−Sα(U)}. The microcanonical entropy Sα(U) can be determined at any time by integrating eq 1 as Sα(U) = ln Ω = ∫ U 1/Tα(z) dz. In our simulations, M = 32 for the isolated monomer (temperature range from 150 to 450 K) and M = 48 for each system with crowding agents (temperature range from 240 to 420 K). The geometric temperature distribution was used with overlap parameter of κ = 0.1. Every 2000 timesteps, a replica exchange is attempted between two neighboring temperature subranges. The combination of multicanonical sampling within a temperature subrange and replica exchange between subranges efficiently overcomes any energy barriers and results in ergodic sampling. Once the scaling factor decreased below f − 1 < 1 × 10−7, configurations were saved every 2000 timesteps for later analysis. A total of 50 000 configurations per replica were saved. Once the simulation is complete, the optimized density of states is determined through the statistical temperature weighted histogram analysis method (ST-WHAM).42 Canonical thermodynamic averages for any temperature T can then be determined through reweighting



COMPUTATIONAL METHODS The large size and long time scale necessary for simulations of crowded environments lend themselves to the use of a coarsegrained protein model. We use the model of Bereau and Deserno,35 which accounts for all 21 naturally occurring amino acids. The structure of the protein is shown schematically in Figure 1. This intermediate resolution model features a knowledge-based implicit solvent scheme based on an analysis of the Protein Data Bank (PDB).36 Charges are not explicitly represented but are implicit in the statistics-based interaction potential. Each residue is assigned a hydrophobicity ε (0 ≤ ε ≤ 1), which determines the strength of the residue−residue hydrophobic attraction. On this scale, hydrophobicity increases as ε increases. Importantly, this model is not biased toward any particular secondary structure. One additional feature of this model is that all side chains have the same size. While not physically correct (the strength of the hydrophobic interaction depends on the size of the hydrophobe),37 this is an advantageous feature for this study. It is known that crowding agent size and shape influence how strongly they affect the stability of the test protein.38 Because our crowding agents all B

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B Table 1. Details of the Test Protein and the Seven Crowding Agent Systemsa ε

crowding agent

Tf (K)

ΔΔG

ΔS

Ω

σ

89.1 (0.1)

116.5 (0.1)

0.0016 (0.0004)

50.0 49.3 61.8 123.8 380.6 330.6 283.9

0.0030 0.0030 0.0024 0.0013 0.0000 0.0000 0.0000

dilute 322.4 (0.1) solution (ExV)15 (Lys)15 (Ser)15 (His)15 (Tyr)15 (Met)15 (Leu)15

0.000 0.110 0.260 0.490 0.670 1.000

361.3 359.6 332.0 319.6 320.7 329.8 337.1

(0.5) (0.4) (0.7) (0.4) (0.1) (0.2) (0.2)

0.59 0.58 0.23 −0.01 −0.02 0.51 0.53

(0.07) (0.07) (0.05) (0.04) (0.04) (0.04) (0.04)

16.0 17.7 15.1 73.8 152.3 148.6 123.9

(0.1) (0.2) (0.2) (0.2) (0.1) (0.1) (0.1)

(0.2) (0.2) (0.4) (0.3) (0.2) (0.2) (0.2)

(0.0019) (0.0016) (0.0029) (0.0017) (0.0004) (0.0008) (0.0004)

a Hydrophobicity of the crowding agents, ε, the folding temperature of the test protein, and the change in the free energy of unfolding due to crowding (in kcal/mol) at T = 322.4 K are given along with entropy of unfolding (in cal/mol·K), the cooperativity, Ω, and the foldability, σ. The standard deviation is in parentheses.

Figure 2. Fluctuations in the overlap parameter χ as a function of the temperature. Data have been shifted vertically for clarity in order of increasing crowding agent hydrophobicity. The vertical dashed line is the folding temperature in dilute solution.

A (T ) =

∑U A(U )eS *−βU ∑U eS *−βU

configuration. Details of each system, including the folding temperature of the test protein and the hydrophobicity of the crowding agents, are given in Table 1. To eliminate complications due to different crowding agent shapes, each crowding agent was fixed in an α-helical configuration through additional harmonic interactions between all α-carbon pairs and through each side chain particle and the particles connected to it by three bonds. Consequently, the only difference from one system to another was strength of the hydrophobic interaction and, for (ExV)15, the lack of intermolecular hydrogen bonding. Crowding agents interact with each other only through excluded volume interaction.

(2)

where the asterisk indicates the ST-WHAM optimized microcanonical entropy. We simulated the (AAQAA)3 test protein by itself and in the presence of seven different homopeptide crowding agents, each of which consisted of 15 residues. Six of the crowding agents were chosen to span the range from very hydrophilic (Lys)15, (ε = 0.00) to very hydrophobic (Leu)15 (ε = 1.00). Although the amino acid lysine can present hydrophobic behavior due to its hydrocarbon chain, in this model35 it is assigned a hydrophobicity of ε = 0 and is therefore considered to the least hydrophobic residue. The remaining crowding agent, (ExV)15, interacted with the test protein only through excluded volume interactions and serves as a macromolecular crowding control, as there are no attractive interactions between the test protein and crowding agents. (Lys)15 differs from (ExV)15 only though the presence of hydrogen bonding between the test protein and the crowding agents. The simulation box had side lengths of 45 Å and periodic boundary conditions. Each system consisted of one test protein and nine identical crowding agents with a total density of 300 mg/mL, comparable to that found in the cell. All initial configurations had the test protein in a helical



RESULTS Configuration data were analyzed using the ST-WHAM algorithm. A measure of how similar a configuration is relative to the native state is given by the overlap parameter χ χ=

Nα − 3 2 Σ Nα2 − 5Nα + 6 i = 1



∑ j=i+3

Θ(σ − |rij − rijN |) (3)

When the test protein is in the α-helical native state, χ = 1. Here Nα is the number of α-carbons in the test protein, rij is the distance between α carbons i and j, and rNij is the same distance C

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 3. Folding temperature of (AAQAA)3 plotted as a function of crowding agent hydrophobicity. The dashed horizontal line is the folding temperature in dilute solution. Error bars are smaller than the symbols.

in the native state. Θ is the Heaviside function and is equal to 1 when its argument is positive and is zero otherwise. Thus, pair ij only contributes to the sum in eq 3 when their distance is within σ = 0.5 Å of that in the native state. The folding temperature, Tf, is taken to be the temperature with the maximum fluctuations in the structure of the test protein, as defined by fluctuations in χ. The average fluctuations in the overlap parameter χ are shown as a function of temperature in Figure 2 for all seven crowding agent systems as well as for the dilute solution (data have been shifted vertically for clarity). The folding temperature for the isolated test protein (dilute solution) was found to be 322.4 K. As is clear from Figure 2, the folding temperature of the test protein depends on the hydrophobicity of the crowding agents. The dashed vertical line is the folding temperature of (AAQAA)3 in dilute solution. The folding temperature for each system is reported in Table 1 along with the hydrophobicity of the crowding agents. Uncertainty values were determined by randomly assigning each saved configuration to one of four groups and averaging within each group. The standard deviation for all four averages was taken to be the uncertainty. In comparison with the folding temperature of the dilute solution, the histidine and tyrosine systems are destabilized and unfold at a lower temperature, while the excluded volume and lysine (which are less hydrophobic) and serine, methionine, and leucine (which are more hydrophobic) systems are stabilized and unfold at a higher temperature. The dependence of Tf on crowding agent hydrophobicity is shown in Figure 3. As the hydrophobicity of the crowding agents increases from ε = 0.000 (lysine) to ε = 1.000 (leucine), the thermal stability of the native state of the test protein initially decreases, passes through a minimum, and then increases. In the presence of histidine and tyrosine crowding agents, the decrease in thermal stability is large for the test protein to unfold at a temperature lower than that in dilute solution. Three factors contribute to this dependence of Tf on ε: (1) excluded volume effects, (2) enthalpic hydrophobic attractions, and (3) a packed arrangement of crowding agents around the test protein that becomes increasingly significant as the crowder hydrophobicity increases. For crowding agents with low hydrophobicities, the excluded volume effect

dominates and the test protein is stabilized relative to the dilute solution. For intermediate values of the crowding agent hydrophobicity (0.2 < ε < 0.6), the test protein is destabilized due to favorable test protein−crowder interactions stabilizing its unfolded states. For high values of crowding agent hydrophobicity, the native state is stabilized by crowding agents packing around the folded test protein, increasing its stability through favorable side chain−side chain interactions. We compared the free energy surfaces (FES) for the test protein in dilute solution to that of the test protein in the presence of the seven crowding agents, shown in Figure 4. Here the free energy of the system is plotted as a function of the radius of gyration of the test protein and its number of native contacts. A “contact” is defined as any two particles within two Angstroms of their cross diameter. The minimum free energy path across each surface is shown in Figure 5. The FESs were calculated at T = 322.4 K, the folding temperature of the monomer in dilute solution. The related potential energy surfaces are shown in Figure S1 in the Supporting Information. Given the coarse-grained nature of our model, the free energies we calculate differ from what would be measured experimentally. However, we expect the trends in their dependence on the crowding agent properties to be general to both experiments and more detailed computational models. In calculating ΔG, the folded state was taken to be the minimum of the native basin, which corresponds to a partially folded state for all systems except for those with the methionine and leucine crowding agents, where it corresponds to a fully folded state. Values for the change in the free energy of unfolding at T = 322.4 K are given in Table 1 and are between −0.05 and 0.7 kcal/mol. By comparison, Wang et al.10 reported values of ΔG for ubiquitin, a larger protein than our test protein, of 0 ± 4 kcal/mol in Bovine Serum Albumin and 4 ± 2 kcal/mol in lysozyme (both at concentration of 100 mg/mL) at its folding temperature of T = 378 K. As the hydrophobicity of the crowding agents increases from the excluded volume crowders to tyrosine crowders, the free energy of the unfolded states decreases relative to that of the folded state. For the histidine and tyrosine systems, which have a folding temperature less than that of the dilute solution, the free energy of the extended states is less than that of the folded D

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 4. Free-energy landscape ΔG (in kcal/mol) at T = 322.4 K for (AAQAA)3 in dilute solution and in the presence of the seven different crowders, plotted as a function of the number of native contacts (abscissa) and the radius of gyration (ordinate). Representative configurations of the test protein in the minimum of the native state basin are shown for the histidine and tyrosine crowder systems. Protein images were generated using VMD.43

Figure 5. Free energy of unfolding along the minimum free energy path, ξ. The colors are the same as in Figure 2.

state. At T = 322.4 K, both of these systems are just above their folding transition temperatures and the fluctuations in χ are near their peak. Consequently, the free energy of the folded state is only slightly higher than that of the unfolded state. For the methionine and leucine crowders, the extended states are again higher in free energy than the native state. As the crowding agent hydrophobicity increases from histidine crowders to leucine crowders, the number of intermediate, partially folded states between the folded and the unfolded states also decreases. The significant decrease in the size of the folded basin and its corresponding increase in how sharply it is defined are due to the effects of a packed arrangement of crowding agents discussed later. Note that the FES for the test protein in the presence of histidine crowders is very similar to that in dilute solution. In Figure 5, ΔG(ξ = 0) is the minimum free energy at the smallest number of native contacts and ΔG(ξ = 1) is the minimum free energy at the greatest number of native contacts.

As the hydrophobicity of the crowders increases, the free energy of the unfolded states first decreases, passes through a minimum below the free energy of the folded state, and then increases. In addition, as the hydrophobicity of the crowders increases, the free energy barrier between the folded and unfolded states also decreases, vanishes altogether, and then increases again. For the excluded volume and lysine crowders, the barrier is asymmetric and ∼0.6 kcal/mol for folding and 1.2 kcal/mol for unfolding. In the presence of serine crowding agents, the minimum free energy path across the FES slopes from downward from the unfolded states to the folded states without a barrier. In the presence of histidine crowders, the free energy path is essentially flat between the folded and unfolded states, with only a small barrier of ∼0.05 kcal/mol between them. For the tyrosine crowders, the free energy of the two basins is nearly equal, and there is a barrier of 0.8 kcal/mol between them. When the crowding agents are methionine and leucine crowders, however, the free energy barrier increases E

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 6. Radius of gyration of the test protein as a function of the temperature. Colors are the same as in Figure 2.

Figure 7. Fractional helicity as a function of the temperature. When the test protein is fully helical, θ = 1. Colors are the same as in Figure 2.

partially folded states is reflected in the behavior of the fractional helicity, discussed later. Figure 6 shows the temperature dependence of the radius of gyration for the test protein in each system. The excluded volume and lysine systems, where the excluded volume effect is clearly dominant, show a smaller radius of gyration relative to the dilute solution for any given temperature. This is consistent with the predictions of macromolecular crowding.2 For the serine, histidine, tyrosine, methionine, and leucine systems, in which hydrophobic interactions are present and play a more significant role, the unfolded states of the test protein are more extended than those of the dilute solution. Hydrophobic interactions between the test protein and the crowding agents are favorable; therefore, the unfolded states of the test protein are more extended in the systems with hydrophobic crowding agents to maximize the number of possible interactions. However, in systems with highly hydrophobic crowding agents,

dramatically and is again asymmetric: it is 0.8 and 0.9 kcal/mol, respectively, for folding but 1.3 and 1.4 kcal/mol for unfolding. The increase in the free energy barrier correlates with the presence of the packed configuration, which strongly favors the folded state discussed later. The folding transitions of the test protein indicated by the peaks of Δχ in Figure 2 are between the unfolded state and a partially folded state for all eight systems. Because the histidine and tyrosine crowding agent systems are very near their folding temperature, this can be seen in their FESs in Figure 4 as well as in the minimum free energy path in Figure 5. Representative configurations of these partially folded states for these two systems are shown in Figure 4. These partially folded states are characterized by frayed ends of the helix. These ends can be stabilized by hydrogen bonding with the crowding agents. (See Figure S2 in the SI for examples.) The presence of these F

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

energy of the test protein due solely to intraprotein interactions. We note that when interactions with the crowding agents are included, the maximum in the heat capacity can occur at a temperature below the folding temperature, Tf. Values of σf for each system are given in Table 1 and are all very close to zero. The behavior of σf parallels that of the cooperativity, indicating that as the crowding agent hydrophobicity increases, the test protein becomes a better folder. The presence of hydrophilic crowders (excluded volume, lysine, and serine) causes the foldability to increase compared with that in the dilute solution, while more hydrophobic crowders cause it to decrease, becoming indistinguishable from zero for the three most hydrophobic crowding agents. As with the cooperativity, histidine has only a small effect on σf. As the hydrophobicity increases, the potential energy basins on the PES become deeper and more clearly defined, creating a higher energy barrier between the multiple stabilized unfolded states and the fewer stabilized folded states. (See Figure S1 in the SI.) A sharper variation in the depth of the basins from the unfolded to the folded states, for the tyrosine, methionine, and leucine systems, in particular, results in a more first-order-like folding transition and a decrease in σf.49 A consequence of this is that the test protein becomes a better folder as the hydrophobicty of the crowding agents increases.

methionine and leucine, the unfolded state of the test protein is not as extended as those in the less hydrophobic systems. This is due to the strong attractions between the crowding agents and the test protein in the leucine and methionine systems, which cause the crowding agents to be in closer proximity to the unfolded test protein. Indeed, at temperatures above the folding transition, the test protein can be wrapped around one of the crowding agent molecules. The overall effect is a decrease in the average value of the radius of gyration. The temperature dependence of the fractional helicity, θ, a measure of the amount of secondary structure, is shown in Figure 7. Crowding is known to increase the amount of secondary structure.38,44,45 At all temperatures below the folding temperature, the test protein has a greater amount of helical structure in the systems with crowding agents than in dilute solution. Those crowding agents that form the “packed” structure discussed later (tyrosine, methionine, and leucine) show the greatest increase in secondary structure. This increase in secondary structure corresponds to the native basin in the FES becoming deeper and more sharply defined. The packed crowding agents effectively confine the test protein and prevent fluctuations, resulting in an increase in the amount of secondary structure. The packing is most efficient when the test protein is fully helical. Finally, as can be seen in Figure 2, the shape of the curve of Δχ as a function of T, as well as the folding temperature, depends on the crowding agent hydrophobicity. The narrowness of the folding transition is measured by the cooperativity. One convenient measure of the cooperativity is given by the dimensionless cooperativity index46 Ω=−

2 ⎛ ⎡ d⟨χ ⟩ ⎤⎞ Tmax ⎟ ⎜max⎢ ⎣ dT ⎥⎦⎠ ΔT ⎝



DISCUSSION In the presence of crowding agents that interact solely through excluded volume (black diamond in Figure 3), the test protein has a folding transition at 361.3 K, 38.9 K above that of the dilute solution. Crowding agents decrease the number of available microstates and consequently decrease the entropy of the test protein. Because the crowding agents are largely unaffected by the folding transition of the test protein (as evidenced by the distance between the centers of mass, see the Supporting Information), we can compare the ΔS of unfolding of the test protein in dilute solution, ΔS = 89.1 cal/mol·K, to that in the presence of excluded volume crowders, ΔS = 16.0 cal/mol·K. The presence of crowding agents decreases the entropy of the unfolded states by approximately ΔΔS = 70 cal/ mol·K. This is in qualitative agreement with values of ΔΔS of 300−500 cal/mol·K reported by Feig and Sugita for chymotrypsin inhibitor 2 in the presence of bovine serum albumin and lysozyme.9 Because these systems are larger and have more degrees of freedom than the ones considered here, they are expected to have larger values of ΔΔS. The unfolded states are more strongly affected by the presence of crowding agents, resulting in a stabilization of the native state by destabilizing the unfolded states.2 In the presence of lysine crowding agents, the test protein shows thermodynamic behavior almost identical to that in the presence of excluded volume only crowding agents. The only difference between the two systems is the ability for hydrogen bonds to form between the test protein and the lysine crowding agents, which causes a slight destabilization of the test protein (Tf = 359.6 K in comparison with Tf = 361.3 K for the excluded volume system). Because the crowding agents are fixed in an αhelical configuration, there are minimal sites available for hydrogen bonding. Thus, attractive interactions have a small effect and the excluded volume interactions play a dominant role in the stability of the test protein in the presence of lysine crowding agents. While excluded volume interactions and hydrogen bonding are also present in the serine crowding agent system, there is a

(4)

where d⟨χ⟩/dT is the temperature derivative if χ, max[d⟨χ⟩/ dT] is its maximum value, Tmax is the temperature at which the maximum occurs, and ΔT is the full width at half-maximum in the peak of d⟨χ⟩/dT. Values of Ω for each system are given in Table 1. Crowding agents that are less hydrophobic than the test protein decrease the cooperativity, while those that are more hydrophobic increase the cooperativity. Histidine, which has a hydrophobicity nearly equal to that of the test protein (0.26 for histidine vs 0.234 for (AAQAA)3), has almost no effect on the cooperativity of the test protein. This is consistent with that found by Tsao and Dokholyan,47 who saw a decrease in the cooperativity using excluded volume (i.e., hydrophilic) crowders but not so with the results of Lewis et al.,48 who saw an increase in cooperativity due to crowding of identical molecules. However, both of these simulations use Go-like models, so the comparison with our results is limited. In addition to the folding cooperativity, the “foldability” of the test protein is affected by the crowding agent hydrophobicity. The foldability is defined by46 σf =

Tθ − Tf Tθ

(5)

where Tθ is the collapse temperature and is the temperature at which the heat capacity as well as the fluctuations in the potential energy are at a maximum. The folding temperature, Tf, was defined previously. Proteins with small values of σf are considered to be good folders and exhibit two-state behavior while those with σf ≈ 1 encounter long-lived misfolded structures. Tθ was determined using the fluctuations in the G

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

values of the crowding agent hydrophobicity, the native state is thermodynamically stable while the unfolded state is stable at intermediate values. As the hydrophobicity of the crowding agents increases, first the free energy of the unfolded states decreases faster than that of the folded states, and then the trend reverses with the free energy of the folded state decreasing faster. Unlike the serine, histidine, and tyrosine systems, where the enthalpic hydrophobic interactions were seen to destabilize the folded state of the test protein, for the methionine and leucine systems (the two most hydrophobic crowding agents), the enthalpic interactions are stabilizing. Strong hydrophobic attractions cause four crowding agents to pack around the test protein at temperatures below the folding temperature. This favorably energetic, packed configuration, highlighted in Figure 9 by the dashed circle, results in an increase in side chain−side chain contacts. As a result, the folded state has more contacts than the unfolded state: an average of 69 test protein− methionine side chain−side chain contacts for the folded state versus 57 for the unfolded state at T = 322.4 K. These interactions result in a lower energy configuration and greatly stabilize the test protein but at a significant entropic cost. Table 1 gives values for the change in entropy upon unfolding, ΔS. For tyrosine, methionine, and leucine crowders, all of which form this packed structure, ΔS is between 120 and 155 cal/mol· K, compared with just 15−18 cal/mol·K for the excluded volume, lysine, and serine crowders, which do not for this structure. The large ΔS is due to loss of translational and rotational motion of the four packed crowders. Histidine, which shows limited packing, has an intermediate value of ΔS = 73.7 cal/mol·K and is comparable to that in dilute solution, 89.1 cal/ mol·K. Because of the large ΔS of unfolding, this energetically favorable but entropically unfavorable packing conformation is only lower in free energy when the strength of the side chain− side chain interactions is very high, that is, when the crowding agents are very hydrophobic. For hydrophilic crowders such as lysine and serine, the entropic cost outweighs the energetic gain, and this conformation is not favored. This stabilization effect differs from the entropic stabilization characteristic of macromolecular crowding. The packed conformation stabilizes the native state through association with crowding agents as opposed to destabilizing the enfolded states by decreasing their conformational entropy.2 That ΔS decreasing slightly as the crowding agents become more hydrophobic from tyrosine to methionine to leucine is due to the strong interaction between the unfolded test protein and a single crowder previously discussed with the radius of gyration. The magnitude of these ΔS values is consistent with those reported in the literature.50 While this “packed” arrangement is most prominent for the more hydrophobic simulations, similar side chain−side chain interactions are also observed to a limited degree in the less hydrophobic systems. Figure 10 shows the temperature dependence of the average number of packed crowding agents, Nc, which ranges from 0 to 4, where 0 represents no crowding agents packed around the test protein and 4 represents the maximum packing of four crowding agents. For all temperatures, as crowding agent hydrophobicity increases, the average number of crowding agents packed around the test protein also increases. For the excluded volume and lysine crowding agents, packing is not observed at any temperature due to the lack of side chain−side chain hydrophobic attractions. For the serine and histidine crowding agents, limited packing is observed that quickly decreases and eventually disappears as temperature

notable difference between the stability of the test protein in the lysine and serine simulations. This difference is a destabilization of the native state through a stabilization of the unfolded states by favorable hydrophobic attractions between the test protein and the crowding agents. The extended, unfolded states have more contact with the crowding agents than the compact, native state: at T = 322.4 K, an average of 15 side chain−side chain contacts for the extended states compared with an average of 13 for the folded state. Consequently, the hydrophobic interactions stabilize the unfolded states more than the folded states, resulting in a net decrease in the stability of the native state of the test protein and a decrease in the folding temperature compared to lysine. However, the stabilization due to the excluded volume effect dominates the destabilization due to hydrophobic attractions, resulting in a folding temperature for the serine system that is still above that of the test protein in dilute solution. As crowding agent hydrophobicity is increased from serine (ε = 0.11) to histidine (ε = 0.26) and tyrosine (ε = 0.49), the unfolded states become increasingly stabilized relative to the folded states. For histidine and tyrosine, the destabilizing attractive interactions dominate over the entropic, excluded volume stabilization. This is shown in Figure 3 where the folding temperature of the test protein for both histidine and tyrosine crowding agents is below that of the test protein in dilute solution. As the hydrophobicity of the crowding agents increases, the free energy of the unfolded state (Gu) and of the folded state (Gf) decrease due to favorable enthalpic interactions. However, the free energy of the unfolded states decreases more rapidly than that of the folded states. At T = 322.4 K, Gu is lower than Gf for the histidine and tyrosine crowding agent systems, and the unfolded state is thermodynamically stable. As the hydrophobicity is further increased to methionine (ε = 0.67) and leucine (ε = 1.00), this trend is reversed: Gf decreases faster than Gu, and Gf again becomes lower than Gu. The native, folded state is restabilized. The overall behavior of the free energies is shown schematically in Figure 8, where Gu (dashed line) and Gf (solid line) at the folding temperature of the isolated monomer are plotted against increasing hydrophobicity and cross at two points. The actual data show identical behavior, although it is difficult to see graphically because the difference between the two values is small compared with the entire range of G. For both low and high

Figure 8. Schematic representation of the free energy of the folded (solid line) and unfolded (dashed line) as a function of crowding agent hydrophobicity. For low values of ε, the unfolded states are stabilized more than the folded state with increasing crowder hydrophobicity, while the reverse is true for high values of ε. H

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 9. Representative configurations showing the test protein (AAQAA)3 (red, center of each image) surrounded by hydrophobic crowders (gray, left image) and hydrophilic crowders (cyan, right image). The hydrophobic crowders pack close to the test protein to maximize contacts (inside dashed circle). No such packing is observed with hydrophilic crowders. Images generated using VMD.43

Figure 10. Number of packed crowding agents Nc as a function of the temperature. Data for the excluded volume and lysine systems are coincident. Colors are the same as in Figure 2

increases. For the tyrosine, methionine, and leucine crowding agents, maximum packing is observed at low temperatures and is mostly maintained as temperature increases until the folding temperature is reached, after which the packed configuration disintegrates. The degree of packing, as measured by Nc, correlates with the increase in folding temperature of the test protein as the crowding agent hydrophobicity is increased from histidine to leucine. The “compactness” of the packed configuration is also dependent on the hydrophobicity of the crowding agents. The radial distribution function gc(r) in cylindrical coordinates at T = 240 K is shown in Figure 11. gc(r) is calculated with the z axis defined by the average position of the α carbons on residues 2, 3, 4, and 5 and on residues 11, 12, 13, and 14. When the test protein is in the helical configuration, the z axis is coincident with the long axis of the helix. The behavior of gc(r) is consistent with that of Nc shown in Figure 10. The three most hydrophobic crowding agents (leucine, methionine, and tyrosine) show a distinct double peak between

7 and 13 Å, corresponding to the sides of the crowder helices nearer and farther from the test protein. (See the left image in Figure 9.) The minimum at 10.5 Å is due to the hollow center of the crowder α-helix. The two peaks are lower and less welldefined for histidine crowders and are all but vanished for serine. Lysine and excluded volume crowders show no peaks in the crowder density in this range, again consistent with the behavior of Nc shown in Figure 5. As hydrophobicity increases, the average distance between the test protein and the packed crowding agents decreases: 9.97 Å for serine, 8.84 Å for histidine, 8.64 Å for tyrosine, 8.53 Å for methionine, and 8.44 Å for leucine (as measured by the largest peak in gc(r)). Note that for the most hydrophobic crowding agents, there is a noticeable depletion layer between 14 and 22 Å. It is interesting to note that the presence of histidine crowding agents has little effect on the folding behavior of the test protein. While there is a very small destabilization of the native state at 322.4 K, the entropy of unfolding, the I

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

Figure 11. Radial distribution function of crowding agents at T = 240 K. At this temperature, the test protein is an α-helix for all systems. Data for the excluded volume and lysine systems are coincident. Colors are the same as in Figure 2

helices, in particular, and are consistent with their destabilizing behavior.6 While (AAQAA)3 is a relatively small protein and our system only consists of 10 proteins, we believe that our results are significant with respect to the greater study of protein folding in cellular environments. α-Helices are a dominant structural motif in cellular proteins, and the effect of crowding on their stability has implications for the stability of the protein as a whole. The main effects of crowding are due to the four crowding agents nearest the test protein, as shown by the behavior of both Nc and gc(r) and by the average distance between the centers of mass of the test protein and the crowding agents. (See Figure S3 in the Supporting Information.) We do not anticipate that systems with more crowding agents, but a single test protein would show markedly different behavior. This is subject to verification in future work. This work shows that attractive interactions play a significant role in protein stability and therefore should be accounted for in future computational work in addition to the excluded volume effect. Taking into account these different interactions may help elucidate the mechanisms involved in many misfolded protein diseases such as Alzheimer’s, prion diseases, and Parkinson’s disease. As the hydrophobicity of the crowding agents increases, we see a changeover from a destabilizing effect to a stabilizing effect. A changeover in the stability was observed by Sirur and coworkers for proteins in a spherical cavity.28 However, in that study, the changeover was from stabilizing to destabilizing. For weak protein−cavity interactions the native state was stabilized due to favorable contacts between the protein and the wall. For a weak attraction, the native state experiences a smaller entropic cost for forming contacts with the cavity and is thus stabilized. As the strength of the attraction increased, the unfolded states, which are able to form more contacts with the cavity, are favored and the effect changes from stabilization to destabilization. While we do not see the stabilization due to favorable protein−crowder interactions for low values of the hydrophobicity, the preferential stabilization of the unfolded

cooperativity, the foldability, the average helicity, and the free energy surface are all remarkably similar to those in dilute solution. In addition, the number of side chain−side chain contacts is nearly identical, 27 for the folded state and 28 for the unfolded state at T = 322.4 K. We ascribe this to the similarity in hydrophobicity between the test protein (ε = 0.234) and the histidine crowders (ε = 0.260). The side chain− side chain interactions are all very similar, and the test protein shows little preference toward interacting with itself versus with the crowding agents.



CONCLUSIONS We have identified three mechanisms that play a role in the stabilization or destabilization of an α helix in a crowded environment. These are (1) the stabilization due to the excluded volume effect of crowding agents, (2) a destabilization of the native state due to favorable stabilizing interactions between the crowding agents and the unfolded states of the test proteins, and (3) the stabilization of the native state due to the packing of strongly hydrophobic crowding agents around the test protein. At ε = 0.00, the native state is stabilized by excluded volume interactions. As the hydrophobicity of the crowding agents is increased, the stability of the native state decreases due to increasingly stabilized unfolded states. As the hydrophobicity continues to increase, the native state is increasingly stabilized due to packing. The net effect is that for intermediate values of crowding agent hydrophobicity (0.2 < ε < 0.6), the destabilizing attractive interactions with the crowding agents can overcome the stabilizing effect of their excluded volume. This range is physiologically relevant: the average value of the hydrophobicity in mammalian cells is ε = 0.33,51 suggesting that destabilization due to attractive test protein−crowder interactions may play a role in cellular processes. In addition, many experiments are performed using protein crowding agents such as bovine serum albumin (average hydrophobicity: ε = 0.353) and lysozyme (average hydrophobicity: ε = 0.346). Our results are important in understanding their effect on test proteins and the stability of J

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B

(2) Minton, A. P. Models for excluded volume interaction between an unfolded protein and rigid macromolecular cosolutes: Macromolecular crowding and protein stability revisited. Biophys. J. 2005, 88, 971−985. (3) Sasahara, K.; McPhie, P.; Minton, A. P. Effect of dextran on protein stability and conformation attributed to macromolecular crowding. J. Mol. Biol. 2003, 326, 1227−1237. (4) Cheung, M. S.; Klimov, D.; Thirumalai, D. Molecular crowding enhances native state stability and refolding rates of globular proteins. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 4753−4758. (5) Dhar, A.; Girdhar, K.; Singh, D.; Gelman, D. H.; Ebbinghaus, E.; Gruebele, M. Protein stability and folding kinetics in the nucleus and endoplasmic reticulum of eukaryotic cells. Biophys. J. 2011, 101, 421− 430. (6) Miklos, A. C.; Sarkar, M.; Wang, Y.; Pielak, G. J. Protein crowding tunes protein stability. J. Am. Chem. Soc. 2011, 133, 7116−7120. (7) Rosen, J.; Kim, Y. C.; Mittal, J. Modest protein-crowder attractive interactions can counteract enhancement of protein association by intermolecular excluded volume interactions. J. Phys. Chem. B 2011, 115, 2683−2689. (8) Benton, L. A.; Smith, A. E.; Young, G. B.; Pielak, G. J. Unexpected effects of macromolecular crowding on protein stability. Biochemistry 2012, 51, 9773−9775. (9) Feig, M.; Sugita, Y. Variable interactions between protein crowders and biomolecular solutes are important in understanding cellular crowding. J. Phys. Chem. B 2012, 116, 599−605. (10) Wang, Y.; Sarkar, M.; Smith, A. E.; Krois, A. S.; Pielak, G. J. Macromolecular crowding and protein stability. J. Am. Chem. Soc. 2012, 134, 16614−16618. (11) Harada, R.; Tochio, N.; Kigawa, T.; Sugita, Y.; Fieg, M. Reduced native state stability in crowded cellular environment due to proteinprotein interactions. J. Am. Chem. Soc. 2013, 135, 3696−3701. (12) Kim, Y. C.; Mittal, J. Crowding induced entropy-enthalpy compensation in protein association equilibria. Phys. Rev. Lett. 2013, 110, 208102. (13) Kurniawan, N. A.; Enemark, S.; Rajagopalan, R. Crowding alters the folding kinetics of a β-hairpin by modulating the stability of intermediates. J. Am. Chem. Soc. 2012, 134, 10200−10208. (14) McGuffee, S. R.; Elcock, A. H. Diffusion, crowding & protein stability in a dynamic molecular model of the bacterial cytoplasm. PLoS Comput. Biol. 2010, 6, e10000694. (15) Cossins, B. P.; Jacobson, M. P.; Guallar, V. A new view of the bacterial cystol environment. PLoS Comput. Biol. 2011, 7, e1002066. (16) Predeus, A. V.; Gul, S.; Gopal, S. M.; Feig, M. Conformational sampling of peptides in the presence of protein crowders from AA/ CG-multiscale simulations. J. Phys. Chem. B 2012, 116, 8610−8620. (17) Qin, S.; Zhou, H.-X. FFT-based method for modeling protein folding and binding under crowding: Benchmarking on ellipsoidal and all-atom crowders. J. Chem. Theory Comput. 2013, 9, 4633−4643. (18) Yap, E.-H; Head-Gordon, T. Calculating the bimolecular rate of protein-protein association with interacting crowders. J. Chem. Theory Comput. 2013, 9, 2481−2489. (19) Qin, S.; Zhou, H.-X. Further development of the FFT-based method for atomistic modeling of protein folding and binding under crowding: Optimization of accuracy and speed. J. Chem. Theory Comput. 2014, 10, 2824−2835. (20) Ignatova, Z.; Krishnan, B.; Bombardier, J. P.; Marcelino, A. M. C.; Hong, J.; Gierasch, L. M. From the test tube to the cell: Exploring the folding and aggregation of a β-clam protein. Peptide Sci. 2007, 88, 157−163. (21) Inomata, K.; Ohno, A.; Tochio, H.; Isogai, S.; Tenno, T.; Nakase, I.; Takeuchi, T.; Futaki, S.; Ito, Y.; Hiroaki, H.; Shirakawa, M. High-resolution multi-dimensinal NMR spectroscopy of proteins in human cells. Nature 2009, 458, 106−109. (22) Ebbinghaus, S.; Dhar, A.; McDonald, J. D.; Gruebele, M. Protein folding stability and dynamics imaged in a living cell. Nat. Methods 2010, 7, 319−323.

states over the folded state is the same in both systems. However, for high crowder hydrophobicities, two differences result in the stabilization of the native state seen here. First, when folded, the helical test protein (AAQAA)3 has all of its residues exposed. The stabilizing packed arrangement provides these already exposed residues with favorable contacts with the hydrophobic crowders. Second, our crowding agents can move to maximize contacts, while in the case of the cavity, the wall is unable to deform. Together, these allow for the stabilized, packed conformation seen here. The stabilizing effect due to packing of hydrophobic crowders around the folded test protein is perhaps the most surprising result of this study. It makes it clear that the behavior of the helical crowding agents is a critical factor in affecting the stability of the test protein and that association can increase the stability of the proteins. Because this effect clearly depends on the geometry of both the crowding agent and the folded test protein, its extension to larger test proteins and to crowders of different shapes is unclear. For larger proteins with buried residues, such as globular proteins and those studied by Sirur et al.,28 one would expect that the unfolded state would be stabilized because it would expose more residues to the crowding agents than the folded state and the packed structure would not form. In addition, as previously noted, the helical conformation of the crowding agents limits the role that intermolecular hydrogen bonding plays. We anticipate that crowding agents with more exposed hydrogen bond donors, such as those with β sheets, will have a larger destabilizing effect on the test protein. This is a direction of future work in our group.



ASSOCIATED CONTENT

* Supporting Information S

Figure S1 shows the potential energy surface for all eight test protein−crowder systems. Figure S2 shows representative partially folded configurations of the test protein surrounded by histidine and tyrosine crowding agents. Figure S3 shows the distance between the centers of mass of the test protein and each of the nine crowding agent molecules for each crowding agent system. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: (413) 538-2449. Fax: (413) 538-2327. Author Contributions †

B. M., S. M., and S. N. contributed equally.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS A.E.v.G. gratefully acknowledges support from a Cottrell College Science Award from the Research Corporation. We would also like to thank an anonymous reviewer for helpful comments and suggestions.



REFERENCES

(1) Cheung, M. S. Where soft matter meets living matter − protein structure, stability, and folding in the cell. Curr. Opin. Struct. Biol. 2013, 23, 212−217. K

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry B (23) Crowley, P. B.; Chow, E.; Papkovskaia, T. Protein interactions in the Escherichia coli cystol: An impediment to in-cell NMR spectroscopy. ChemBioChem. 2011, 12, 1043−1048. (24) Ghaemmaghami, S.; Oas, T. G. Quantitative protein stability measurement in vivo. Nat. Struct. Biol. 2000, 4, 879−822. (25) Schlesinger, A. P.; Wang, Y.; Tadeo, X.; Millet, O.; Pielak, G. J. Macromolecular crowding fails to fold a globular protein in cells. J. Am. Chem. Soc. 2011, 133, 8082−8085. (26) Betancourt, M. R.; Thirumalai, D. Exploring the kinetic requirements for enhancement of protein folding rates in the GroEL cavity. J. Mol. Biol. 1999, 287, 627−644. (27) Jewett, A. I.; Baumketmer, A.; Shea, J.-E. Accelerated folding in the weak hydrophobic environment of a chaperonin cavity: Creation of an alternate fast folding pathway. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 13192−13197. (28) Sirur, A.; Knott, M.; Best, R. B. Effect of interactions with the chaperonin cavity on protein folding and misfolding. Phys. Chem. Chem. Phys. 2014, 16, 6358−6366. (29) Spitzer, J. J.; Poolman, B. Electrochemical structure of the crowded cytoplasm. Trends Biochem. Sci. 2005, 30, 536−541. (30) Scholtz, J. M.; York, E. J.; Stewart, J. M.; Baldwin, R. L. A neutral, water-soluble, α-helical peptide: The effect of ionic strength on the helix coil equilibrium. J. Am. Chem. Soc. 1991, 113, 5102−5104. (31) Shalongo, W.; Dugad, L.; Stellwagen, E. Distribution of helicity within the model peptide Acetyl(AAQAA)3amide. J. Am. Chem. Soc. 1994, 116, 8288−8293. (32) Ferrara, P.; Apostolakis, J.; Calflisch, A. Thermodynamics and kinetics of folding of two model peptides investigated by molecular dynamics simulations. J. Phys. Chem. B 2000, 104, 5000−5010. (33) Chebaro, Y.; Dong, X.; Laghaei, R.; Derreumaux, P.; Mousseau, N. Replica exchange molecular dynamics simulations of coarse-grained proteins in implicit solvent. J. Phys. Chem. B 2009, 113, 267−272. (34) Bereau, T.; Deserno, M.; Bachmann, M. Structural basis of folding cooperativity in model proteins: Insights from a microcanonical perspective. Biophys. J. 2011, 100, 2764−2772. (35) Bereau, T.; Deserno, M. Generic coarse-grained model for protein folding and aggregation. J. Chem. Phys. 2009, 130, 235106. (36) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.; Weissig, H.; Shindylov, N.; Bourne, P. E. The protein data bank. Nucleic Acid Res. 2000, 28, 235−242. (37) Huang, D. M.; Chandler, D. Temperature and length scale dependence of hydrophobic effects and their possibly implications for protein folding. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 8324−8327. (38) Homouz, D.; Stagg, L.; Wittung-Stafshede, P.; Cheung, M. S. Molecular crowding modulates folding mechanism of α/β protein apoflavodoxin. Biophys. J. 2009, 96, 671−680. (39) Kim, J.; Straub, J. E.; Keyes, T. Replica exchange statistical temperature molecular dynamics algorithm. J. Phys. Chem. B 2012, 116, 8646−8653. (40) Church, M. S.; Ferry, C. S.; van Giessen, A. E. Thermodynamics of peptide dimer formation. J. Chem. Phys. 2012, 136, 245102. (41) Wang, F.; Landau, D. P. Determining the density of states for classical statistical models: A random walk algorithm to produce a flat histogram. Phys. Rev. E 2001, 64, 056101. (42) Kim, J.; Keyes, T.; Straub, J. E. Iteration-free, weighted histogram analysis method in terms of intensive variables. J. Chem. Phys. 2011, 135, 061103. (43) Humphrey, W.; Dalke, A.; Schulten, K. VMD - Visual Molecular Dynamics. J. Mol. Graphics. 1996, 4, 33−38. (44) Perham, M.; Stagg, L.; Wittung-Shafshede, P. Macromolecular crowding increases structural content in folded proteins. FEBS Lett. 2007, 581, 5065−5069. (45) Stagg, L.; Zhang, S.-Q.; Cheung, M. S.; Wittung-Stafshede, P. Molecular crowding enhances native structure and stability of α/β protein flavodoxin. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 18976− 18981. (46) Klimov, D. K.; Thirumalai, D. Cooperativity in protein folding: From lattice models with sidechains to real proteins. Fold. Des. 1998, 1, 127−139.

(47) Tsao, D.; Dolkholyan, N. V. Macromolecular crowding induces polypeptide compaction and decreases folding cooperativity. Phys. Chem. Chem. Phys. 2010, 12, 3491−3500. (48) Lewis, J. I.; Moss, D. J.; Knotts, T. A., IV Multiple molecule effects on the cooperativity of protein folding transitions in simulations. J. Chem. Phys. 2012, 136, 245101. (49) Kim, J.; Keyes, T.; Straub, J. E. Relationship between protein folding thermodynamics and the energy landscape. Phys. Rev. E 2009, 79, 030902. (50) Robertson, A. D.; Murphy, K. P. Protein structure and the energetics of protein stability. Chem. Rev. 1997, 97, 1251−1267. (51) Okayasu, T.; Ikeda, M.; Akimoto, K.; Sorimachi, K. The amino acid composition of mammalian and bacterial cells. Amino Acids 1997, 12, 379−391.

L

DOI: 10.1021/jp512630s J. Phys. Chem. B XXXX, XXX, XXX−XXX