Hydration of the folding transition-state ensemble of a protein

Abstract

The exponential nature of protein folding kinetics suggests the existence of rate-limiting energy barriers between the folded and the unfolded states. The lack of detailed understanding of this barrier and of the attendant rate-limiting ensemble of states still limits our ability to describe mechanisms of folding in detail. Although use of ϕ-value analysis (1–4) to describe the character of the folding transition state ensembles (TSE) remains controversial (5;6), a self-consistent view is emerging (7–12). Considerable differences in the TSE of different proteins have been observed (13;14), but in general, native-like structures are thought to dominate the TSE, possibly even with many side chains already fully dehydrated but with few of them satisfying their native contacts (5;6;15).

Computational analyses of ϕ values have influenced significantly the prevailing views on the structural character of the TSE (16–28), which is thought to be determined primarily by the topology of the protein, not by detailed energetic factors (17;24;29;30). Some computational studies suggest that TSEs consist mostly of expanded and distorted native structure. In MD-based studies, subsets of nucleating interactions are assigned key roles (19;21;22;27), but this is not fully consistent with experimental data (5;6;31).

Despite this progress in characterizing the TSE, the central question of the role of hydration in determining folding rates and mechanisms remains unanswered. Computational studies suggest that the conformations populated in the TSE consist mostly of native structure with a highly solvated hydrophobic core. The final step in the folding reaction is thought to involve packing of the hydrophobic core and concomitant, cooperative squeezing out of water from the core (16;20;25;32). Water in the core of the TSE could prevent nonnative contacts from being formed (28). Experimentally it has been shown that replacement of a hydrophobic core sidechain with an isosteric polar one can alter the hydration of the TSE when residues crucial to the folding mechanism are targeted (32). These experimental findings can be modeled using a desolvation potential (16). They support the view of the TSE as a highly hydrated state.

The emerging view of TSEs as a highly hydrated, native-like state is not consistent with previous results from pressure-jump relaxation studies of wild type (WT) staphylococcal nuclease (SNase) (33). These studies revealed that the rate-limiting step in folding was accompanied by significant dehydration, and that the TSE for this protein was relatively dry. Indeed, pressure perturbation offers a unique experimental means for probing the role of hydration changes in protein conformational changes (34) because pressure effects arise from differences in the specific volumes of the various (folded, intermediates, unfolded, transition) states of the protein. These differences in specific volume are linked directly to the degree and type (polar or non-polar) of hydration, and to the existence or loss of cavity volume (35;36), which also involve hydration or dehydration, respectively. Most proteins studied to date exhibit a larger specific volume in the folded state than in the unfolded state, hence the application of pressure leads to their unfolding. Our earlier p-jump relaxation studies demonstrated that the effect of pressure on the stability of WT SNase arises primarily from a pressure-induced decrease in the folding rate constant due to a positive activation volume for this reaction, suggesting significant dehydration of the transition state ensemble (33). Similar p-jump relaxation studies on the folding of six other proteins indicated that significant (although less) dehydration accompanied the rate-limiting step in folding of these proteins as well (37–42).

To probe the state of hydration of the TSE of SNase in depth, we performed pressure unfolding experiments with variants in which a deeply buried hydrophobic residue, Val-66, was replaced with Arg, Lys, His, Asp, Glu, or Asn. Charged and polar groups are well hydrated in water, therefore removal of these groups from water for placement into the core of the protein is energetically unfavorable. The introduction of ionizable and polar groups in the protein interior were expected to perturb the TSE of the protein, and to allow clarification of the role of water in the final stages of folding with unprecedented level of detail.

METHODS

Replacement of core hydrophobic positions with polar or ionizable residues is energetically disfavored (43–50). For this reason, the variants with the internal position Val-66 substituted with Arg, Lys, His, Asp, Glu, or Asn were engineered using a hyperstable form of SNase known as Δ+PHS after the set of substitutions (P117G, H124L, S128A, G50F, V51N) and deletions (Δ44–49). WT SNase and all of the Δ+PHS proteins were produced and purified as described previously (51). The pressure-jump experiments were carried out in either Tris or bis-Tris buffer (Aldrich) 50 mM at the pH indicated in Table 1 to avoid pressure-induced changes in the pH (52). GuHCl (Aldrich) was dissolved into the appropriate buffer, the pH adjusted and used at the concentration indicated in Table 1. Protein concentration was near 70 μM.

High pressure experiments were carried out using a stainless steel high pressure cell designed in our laboratory and fitted with sapphire windows. The high pressure generating equipment has been previously described (53). High purity (18 MOhm) water was used as the pressure generating liquid. Fluorescence measurements were made using the detection from an ISS KOALA instrument (ISS Champaign IL) and ISS excitation source and monochromator coupled to the high pressure cell via an optical fiber. The excitation and emission monochromator slits were at 8 nm and the excitation wavelength was 295 nm, while the emission was set at 335 nm. Emission spectra (310–450 nm) were also acquired once the sample had equilibrated. The acquisition time base was 1 second. Pressure jumps were performed by pumping the pressure with the valve to the high pressure cell closed, followed by rapid opening of the valve. Solution conditions were adjusted with pH or guanidine hydrochloride to decrease the stability of the proteins so that pressures below 4 kbar (400 MPa) were sufficient to unfold the variants tested. We have shown previously that changes in pH or in the concentration of GuHCl change the stability of SNase without affecting the volume change measured by pressure unfolding (53;54).

Equilibrium unfolding profiles were analyzed for the volume change (ΔV_f) and free energy of folding (ΔG_f) using the analysis program Bioeqs and an analytical rather than numerical model of 2-state folding.

Uncertainties on the recovered free energy and volume change values were determined by rigorous confidence limit testing, and take into account correlations between the fitted parameter.

The pressure-jump relaxation profiles at each pressure were analyzed according to classical transition state theory for a single exponential decay:

For a simple two state reaction, the relaxation time τ at each pressure is the inverse sum of the folding and unfolding rate constants k_f(p) and k_u(p) at that pressure.

These rate constants are exponentially dependent upon pressure and the activation volumes for the reactions.

Thus the plots of ln τ vs p are fit using non-linear least squares analysis for the values of the activation volumes (ΔV^≠_f and ΔV^≠_u) and rate constants (k_fo and k_uo) for folding and unfolding at atmospheric pressure. We have constrained the analysis to only two parameters, either those for folding (WT) or unfolding (variants), using the equilibrium volume change and free energy for folding according to

such that

and

such that

Uncertainties reported are calculated from the diagonal of the correlation matrix and correspond to 95% confidence limits.

RESULTS AND DISCUSSION

Proteins with internal polar or ionizable residues

Fluorescence-detected pressure-jump relaxation experiments were performed with V66K, V66R, V66H, V66D, V66E, and V66H variants of the Δ+PHS hyperstable nuclease. Crystal structures of all but the His-66 variant have been determined previously under room temperature and cryogenic conditions (Figure 1) (45;47;48;50;55). With the exception of the V66R variant, in which the β-barrel releases a strand to expose Arg-66 to water (Karp, Gittis, & García-Moreno, in preparation), under conditions near neutral pH, where the internal ionizable groups are neutral, the residues at position-66 are indeed buried in the core of the protein in a hydrophobic pocket. Under these conditions there are also no detectable differences between the native states of the variant proteins and of the background protein. One or two internal water molecules have been observed hydrating the internal polar moieties in cryogenic structures with internal Asp, Glu, Asn, Gln, Tyr, and Trp at position 66. Molecular dynamics simulations also suggest that 2 or 3 water molecules can penetrate into the hydrophobic core of WT SNase at room temperature and pressure, even in the absence of internal polar or ionizable groups (56). Internal water molecules have never been observed crystallographically in the WT protein.

Open in a separate window

Figure 1

Schematic of the structure of the SNase variant V66K with the residue K66 in space-filled representation (50).

The pKa values of the internal ionizable groups have been measured previously: 5.7 for Lys-66 (46–48), 8.9 for Asp-66 (48), 9.2 for Glu-66 (45), < 4 for His-66, >10 for Arg-66 (Karp, Gittis, Garcia-Moreno, in preparation). These pKa values are shifted substantially relative to the pKa values of ionizable groups in water. This effect is driven largely by the dehydration of the ionizable groups when they are in internal positions in the protein (46).

Equilibrium Unfolding by Pressure

The fluorescence-detected pressure-induced unfolding profiles of wild type and variant Snase (Figure 2) were fit to a 2-state unfolding model. The values for the volume change and free energy for folding at atmospheric pressure (ΔV^o_f and ΔG^o_f) recovered from this analysis are listed in Table 1. The ΔV^o_f for wild-type SNase is within error of our previously reported measurement of 70 ±10 ml/mol (at 21 rather than 23 °C) (33;57). The volume change for folding the Δ+PHS was 41 ml/mol, significantly smaller than that of the wild-type. The substitutions and deletions used to create the Δ+PHS hyperstable form of SNase are at external sites and produce only local changes in the structure (43;44). We have also shown previously that the P117G+H124L double substitution has no effect on the volume change of unfolding (54). The other mutations and deletions in Δ+PHS are unlikely to result in significant structural changes since they are all exposed. Hence, the difference in overall volume change between the Δ+PHS and WT SNase must arise from stabilization of residual structure in the unfolded state in the Δ+PHS variant. Clearly introduction of ionizable residues in the hydrophobic core disrupts this residual structure in the unfolded state, rendering the overall volume change upon unfolding comparable to that of the WT Snase.

Open in a separate window

Figure 2

Equilibrium fluorescence detected high pressure unfolding profiles for Snase and its variants. a.) WT Snase, b.) Δ+PHS+V66K, c.) Δ+PHS+V66D, d.) Δ+PHS+V66H, e.) Δ+PHS, f.) Δ+PHS+V66R, g.) Δ+PHS+V66E, h.) Δ+PHS+V66N. Conditions are given in Table 1. The temperature was 23 degrees C. Lines through the points represent fits to the data to equation 1 as described in the text.

With the exception of Arg-66, the other internal ionizable groups were largely neutral under the conditions tested (45;46;48). We expected a contribution to the pressure effects from the electrostriction related to the ionization of these internal ionizable residues upon unfolding. This should have led to an increase the magnitude of the ΔV^o_f above that of the background Δ+PHS protein. While the variants with V66K, V66E, and V66D all exhibited ΔV^o_f values greater than for the Δ+PHS background protein, consistent with electrostriction, those with V66R, V66H, and V66N unfold with ΔV^o_f values that are as large or even larger than for the proteins in which electrostriction was anticipated, and yet in these variants electrostriction cannot be a factor because these groups are either neutral (His, Asn) or charged (Arg) in both folded and unfolded states at the pH where the experiments were performed. Although we have shown previously that an increase in the size of the small internal cavity around position 66 will result in larger ΔV^o_f values (35), this also cannot be the reason behind the larger ΔV^o_f values measured for the variants with internal polar and ionizable side chains because these side chains would actually decrease the volume of the small internal cavity. Instead, the large ΔV^o_f must reflect the destabilization of residual structure in the unfolded state. We note that the free energy values cannot be compared among variants since the conditions of the experiments were chosen in order to tune the pressure-induced unfolding into the appropriate range for our apparatus. However, the free energy measured by pressure and chemical denaturation for all variants are in agreement (data for chemical denaturation not shown), corroborating the validity of the thermodynamic parameters obtained by pressure unfolding.

Pressure-jump relaxation kinetics

Pressure-jump relaxation profiles at multiple pressures were obtained for all the variants. The Δ+PHS protein was not studied because the relaxation was faster than the 2 s dead time of our instrument. This very fast relaxation time is consistent with a large increase in the folding rate for this hyperstable variant. All of the other variants exhibited single exponential relaxation curves, indicative of simple 2-state kinetic behavior, as has been shown previously for WT SNase at high pressure under a variety of conditions (33;57). This loss of kinetic complexity at high pressure appears to arise from the pressure inhibition of proline isomerization (54;58).

For WT SNase (Figure 3a) the relaxation times increase significantly with increasing pressure, as previously reported (33;57). Strikingly, the V66K, V66R, V66D and V66E variants of Δ+PHS exhibited the opposite behavior (shown in Figure 3b for the Δ+PHS/V66K variant). As pressure was increased, the relaxation times became significantly shorter. The relaxation profiles at each pressure for each variant were analyzed for a single exponential relaxation time. Plots of the natural logarithm of these relaxation times (ln τ vs. pressure) are shown in Figure 4 for WT SNase and for the V66N, V66K, V66R, V66D, and V66E variants of Δ+PHS. The ln τ vs. pressure plots for all of the variants were analyzed according to equations (5) and (6) for the activation volumes and rate constants for either folding or unfolding.

Open in a separate window

Figure 3

Pressure jump fluorescence detected relaxation profiles a) WT Snase; (O) correspond to intensity relaxation after a jump from 1350 to 1550 bar, wheras (Δ) correspond to the relaxation after a jump from 2920 to 3120 bar. b) Δ+PHS+V66K variant, (O) correspond to intensity relaxation after a jump from 1100 to 1320 bar, wheras (Δ) correspond to the relaxation after a jump from 2710 to 2920 bar. The data were fit to a single exponential decay model (equation 4), and the lines drawn through the points correspond to the fits. Conditions are those in Figure 2. The data illustrate that for the wild type, relaxation slows significantly as pressure is increased, whereas for the Δ+phs+V66K variant, increasing pressure leads to much faster relaxation.

Open in a separate window

Figure 4

Dependence of the natural logarithm of the relaxation time on the final pressure. The values for the relaxation times are those obtained from the fits of the data such as those in Figure 3. The lines through the points represent fits of the data according to equation 5, as described in the text. a.) WT Snase, b.) Δ+PHS+V66N, c.) Δ+PHS+V66K, d.) Δ+PHS+V66R, e.) Δ+PHS+V66D, f.) Δ+PHS+V66E.

The analysis for the WT protein yielded, as previously reported, a large activation volume for folding and a very small negative activation volume for unfolding (Table 1). In contrast, for the V66K, V66R, V66D, and V66E variants, the activation volume for folding was much smaller than for the wild type protein, while that for unfolding was very large and negative (Figure 4, Table 2).

The differences in folding mechanisms of wild type and variant proteins are described schematically by the volume diagrams in Figure 5. The activation volumes for folding and unfolding combine to yield the overall equilibrium volume changes. The equilibrium unfolding volume changes are somewhat different among the variants tested here. Because the structures of the native state are comparable for the proteins studied, the differences in equilibrium volume change must be linked to differences in the degree of hydration of the unfolded state. In the case of the wild type protein, the volume of the TSE lies near that of the folded state. This is consistent with our prior interpretation that the TSE of WT SNase is highly dehydrated (33;57). In contrast, the V66K variant exhibits a TSE whose volume is much closer to that of the unfolded state, and thus much more hydrated that the WT TSE. For the V66N andV66E the equilibrium unfolding volume change is larger than for the other variants. Since there are no significant structural differences among the native states of these proteins, this is also more likely to reflect increased hydration of the unfolded state. Also since the folded states of the Δ+PHS background variant and the WT Snase are structurally equivalent, and the Δ+PHS mutant is significantly more stable than the WT, it is highly unlikely that these background stabilizing mutations are responsible for the more open, hydrated transition state observed in the variants of Δ+PHS bearing ionizable residues in their hydrophobic core.

Open in a separate window

Figure 5

Volume diagrams for the folding/unfolding reactions of WT Snase, and two of the variants Δ+PHS+V66K and Δ+PHS+V66N. The values for the activation volumes were obtained from the fits of the data such as those in Figure 4 and the equilibrium unfolding volume. All values were normalized to the same volume for the folded state, based on the fact that the structures of these proteins are nearly identical.

The analysis of the pressure-jump relaxation data presented above is grounded on classical transition state theory. The kinetic and energetic basis of pressure-induced protein unfolding has also been analyzed previously (59) with energy landscape theory of protein folding (60), using a square well potential that includes a desolvation barrier (61;62). In this theoretical treatment of a model β-barrel protein, increasing pressure was found to actually lower the transition barrier. Nevertheless, the overall result of pressure was to decrease the folding rate because of a large effect of pressure on the reconfigurational diffusion coefficient, D_Q. This pressure-induced decrease in D_Q suggests that in these simulations, the increase in the rate of folding by increasing pressure was surpassed by the increase in the roughness of the landscape with increasing pressure. We have evidence from iso-energetic pressure studies of WT Snase at varying concentrations of stabilizers and denaturants ((57); Frye, Niemeyer, Royer, Garcia and Onuchic, unpublished data) that pressure may indeed diminish the pre-exponential factor by a factor of 1.7/kbar. However, the overall decrease in the folding rate, over 1 kbar, represents a factor of 7 (Figure 4), demonstrating that pressure indeed increases the free energy at the folding barrier for this protein. The contribution of the pre-exponential factor is, at most, about 25% of the total effect of pressure in this system. In the case of the variants with internal ionizable residues, pressure leads to an increase of the relaxation rate, rather than slowing it down. A general effect of pressure on the roughness of the folding landscape cannot be invoked to explain our observations. Indeed, pressure may also decrease the D_Q in the case of the variants studied here, but this effect, if it exists, is compensated by a pressure-induced increase in the unfolding rate constant linked to a large, negative activation volume for unfolding.

CONCLUSIONS

The pressure-jump relaxation experiments have revealed a striking and unique difference in the folding mechanisms of wild type SN and of variants with internal polar or ionizable groups. Consider as a representative example the case of the V66K variant. The structure of the V66K variant of SNase is essentially identical to structure of the wild type protein (47;50). Under the conditions in which pressure unfolding was performed, the stability of the V66K protein and of its parent protein are quite similar, and the rate constants for folding and unfolding differ only by factors of 10 and 3 respectively. Even the equilibrium pressure unfolding profiles of the two proteins are almost indistinguishable. Yet the pressure dependence of their folding kinetics suggests that these two proteins fold via very different mechanisms.

The differences in the state of hydration of the TSE in the variants and in the wild type protein can be rationalized in terms of the high unfavorable free energy that would be required to neutralize and dehydrate a charged side chain if it were to be buried in a native-like TSE. Internal polar and ionizable groups would promote a hydrated TSE in which the polar and ionizable group destined to occupy internal positions in the native state can remain charged and fully solvated. The pressure unfolding data of the variants with internal polar or ionizable groups is very significant because it reinforces the view that the TSE of wild type SNase is highly dehydrated.

In many computational studies, based primarily on Go models and desolvation potentials, the native structure is achieved before water is expelled from the hydrophobic core (16;20;25;28;32). This would appear to be inconsistent with our observations of a highly dehydrated TSE for WT SNase. It is important to note that not all proteins behave the same way. Wild type SNase (33) and CspB (38), for example, appear to have highly dehydrated TSEs. The other five proteins that have been examined by pressure unfolding (trp repressor (37), P13^MTCP1 (39), CI2 (40), tendamistat (41), and P23 (42)) exhibit specific volumes for their TSEs that lie one half to three/fourths of the way between those of the unfolded and folded states. This is consistent with the TSEs of these proteins having lost water relative to the unfolded state, but nonetheless remaining significantly hydrated. Indeed in replica exchange MD simulations of the folding of protein A (61), about 25% of the hydrating waters remain at the TSE, and are expelled in the final step in folding. This amount of water remaining in the core at the TSE is actually in quite good agreement with the results of the pressure relaxation studies, in which 50–75% of the water is expelled at the TSE, leaving about 25%–50% behind. Wild type SNase appears to be somewhat of an exception (perhaps due to the important size of its hydrophobic core compared to most of the other proteins studied by p-jump techniques) - its TSE appears to be much closer to that of the native state than is observed in other proteins. We note as well that the activation and equilibrium volume changes calculated from the pressure effects on folding are indicative of the overall (thermodynamic) degree of hydration, and are not a quantitative measure of the number of water molecules, such that a direct comparison between the simulations and the experiments is not yet possible, particularly since they were not performed on the same protein systems.

The present results demonstrate that substitution of a single amino acid can drastically alter the state of hydration of the TSE. The role of hydration in determining folding mechanism is likely to be highly sequence dependent, and influenced by the presence of uncompensated polarity in the protein interior. These data also illustrate the high plasticity of the TSE, which apparently adapts structurally to an energetically costly change in sequence. In the case of wild type SNase, the folding TSE is already quite dehydrated, thus the last step in folding does not involve major dehydration of the hydrophobic core. Instead, it might involve the establishment of the native contacts necessary to stabilize side chains in their native-like conformations in an already dehydrated core (5;6;15;31).

Acknowledgments

Abbreviations

References