Folding of the Pit1 homeodomain near the speed limit

Abstract

Current questions in protein folding mechanisms include how fast can a protein fold and are there energy barriers for the folding and unfolding of ultrafast folding proteins? The small 3-helical engrailed homeodomain protein folds in 1.7 μs to form a well-characterized intermediate, which rearranges in 17 μs to native structure. We found that the homologous pituitary-specific transcription factor homeodomain (Pit1) folded in a similar manner, but in two better separated kinetic phases of 2.3 and 46 μs. The greater separation and better fluorescence changes facilitated a detailed kinetic analysis for the ultrafast phase for formation of the intermediate. Its folding rate constant changed little with denaturant concentration or mutation but unfolding was very sensitive to denaturant and energy changes on mutation. The folding rate constant of 3 × 10⁵ s^-1 in water decreased with increasing viscosity, and was extrapolated to 4.4 × 10⁵ s^-1 at zero viscosity. Thus, the formation of the intermediate was partly rate limited by chain diffusion and partly by an energy barrier to give a very diffuse transition state, which was followed by the formation of structure. Conversely, the unfolding reaction required the near complete disruption of the tertiary structure of the intermediate in a highly cooperative manner, being exquisitely sensitive to individual mutations. The folding is approaching, but has not reached, the downhill-folding scenario of energy landscape theory. Under folding conditions, there is a small energy barrier between the denatured and transition states but a larger barrier between native and transition states.

The homeodomain superfamily is comprised of small three-helical-bundle proteins (1). Their pathways of folding are among the best characterized, with information coming from the effects of point mutations on the folding kinetics of individual members and by comparison of the different members of the family with the same fold but grossly different sequences (2). The mechanism of folding slides from a clear framework mechanism for the engrailed homoeodomain (EnHD) to nucleation-condensation for c-Myb (1, 3). EnHD is an ultrafast folding protein that forms an on-pathway intermediate in 1.7 μs, which rearranges in 17 μs to the native structure. The mechanism is unusually very well-established because the structure of the intermediate has been solved by NMR; it consists of helices 2 and 3 (H2, H3) in native-like conformation with H1 being helical but undocked (4). The H2-turn-H3 motif, is in fact, a stable domain of EnHD (5). Temperature-jump experiments using both IR to monitor secondary structure formation, and Trp fluorescence for tertiary interactions have been used to follow the folding of the wild-type protein, a mutant in which the H1 does not dock at physiological ionic strength, and an H2-turn-H3 construct (5). The H2-turn-H3 domain forms both secondary and tertiary structure in 1.7 μs, and the docking of H1 takes 17 μs. Direct measurements of the motion H1 by photoinduced electron transfer fluorescence-quenching correlation spectroscopy shows that 17 μs is close to the time constant of segmental motion of H1 (6). The folding is sufficiently fast that full atomistic molecular dynamics simulations have been done on the protein on similar time scales to those of experiment to fill in the folding pathway (7, 8). The slide from nucleation-condensation to framework mechanism follows the increase in stability of the helical structure (1).

The folding of the intermediate in ∼2 μs opens up the opportunity of studying the formation of a protein close to the “speed limit” of folding (9). We would also like to analyze the transition states for both steps of the reaction by Φ-values (10). Unfortunately, the amplitudes and temporal resolution of the two phases for the folding of EnHD render detailed kinetic studies difficult. The two observed relaxation times have to be deconvoluted into the four constituent rate constants for folding and unfolding of the intermediate and native states. That process is best performed by measuring the rate constants as a function of concentration of denaturant, and the relaxation rate constants become more difficult to separate with increasing concentrations of denaturant. We examined the kinetics of folding of other members of the superfamily, and found Pit1, the homeodomain from pituitary-specific transcription factor (a member of the POU domain family of transcription factors, with 63 amino acids) (11) to be suitable for such studies. It has an equivalent endogenous Trp probe, Trp48, to that in EnHD. The structure of Pit1, its sequence alignment with EnHD and residues mutated in this study is illustrated in Fig. 1. The amplitudes of the phases were larger and were well separated, being 2.3 and 46 μs, and we were able measure the dependence of the rate constants for folding and unfolding of the kinetics phases on concentration of denaturant for wild-type and mutant proteins to give chevron plots.

Fig. 1.

The structure of Pit1, illustrating the side chains mutated, and the sequence alignment of EnHD and Pit1. The first two residues are from cloning vector.

Results

Folding Kinetics of Pit1.

The kinetic traces for folding of Pit1 were all biphasic, differing by about a factor of ∼20 in time scale with much more easily measured amplitudes, and were fitted to double exponential equations (Fig. 2). The rate constants for wild-type Pit1 in the absence of urea at 25 °C and 50 mM acetate and 100 mM NaCl were 3 × 10⁵ s^-1 for the fast phase and 1.6 × 10⁴ s^-1 for the slow. Typical mutations that are destabilizing, such as R18A and E43A move the onsets of the unfolding limbs of fast and slow phases to lower concentrations of urea, with the converse for stabilizing mutations, such as M32A and V47A (Fig. 3). All of the folding limbs for the fast phase converge at ∼3 × 10⁵ s^-1. The slow phases exhibited rollover, characteristic of multistate kinetics (12–14).

Fig. 2.

Amplitude of the fluorescence change on a 3.5 °C temperature jump of wild-type Pit1 to 25 °C in 2.2 M urea, with excitation at 284 nm and emission at 360 nm. The black line is the fit to a double exponential function, with the residuals plotted below.

Fig. 3.

(Upper) Chevron plots of wild-type and mutants of Pit1 and fits of each phase to the relaxation kinetics. (Lower) Equilibrium denaturation curves, fitted to a standard two-state equation. The inset legends are color-coded for the different mutants (e.g., E43A = Glu43 mutated to Ala43, color coded for the red curves).

The dependence of the logarithms of the relaxation rate constants λ on [Urea] were fitted simultaneously assuming that each phase was two-state and each individual rate constant (k) had the standard linear dependence on [Urea], k = k₀ exp(m[urea]/RT), (Fig. 3). Here, we concentrate on the fast phase.

The rate constant for U → I, k_UI, was (2.96 ± 0.14) × 10⁵ s^-1. The m-value for folding, m_UI( = ∂ log k_UI/∂[Urea]), was -0.19 ± 0.07 kcal mol^-1 M^-1. The rate constant for I → U, k_IU, was (2.2 ± 1.5) × 10⁴ s^-1, m_IU( = ∂ log k_u/∂[Urea]) 0.58 ± 0.13 kcal mol^-1 M^-1. (The equivalent values for the slow steps were -0.10 ± 0.07 and 0.30 ± 0.02 kcal mol^-1 M^-1, respectively.) Substituting these rate constants and their dependence on [Urea] into Scheme 1 in the steady state showed that the concentration of I peaked at 7.8% at 3.1 M urea.

Scheme 1.

The general 3-state mechanism with on-pathway intermediate.

Effects of Viscogens on the Fast Phase.

The rate constants were so high it is likely that they could be limited by diffusion of the chains, which can be detected by the effects of viscogens (15). Ethylene glycol is most suitable for use because it a rare viscogen that does not affect stability (16). We verified that the stability of Pit1 remained approximately constant in up to 20% v/v of ethylene glycol (Fig. 4). As the stability was measured by fitting the 3-state kinetic scheme to a 2-state plot, there will be errors in the absolute values but the data are relatively constant. In any case, the effects of a cosolvent on the relative stability of two states depends on the difference in solvent accessible surface area between those two states, which is very small between the denatured and first transition states of Pit1. A plot of log λ₁ versus viscosity, where λ₁ is > 90% composed of k_f, (Fig. 4) had a small negative slope, indicating that internal friction is partly rate-determining. The intercept at 0 viscosity is 4.4 × 10⁵ s^-1, some 50% higher than in the presence of the viscosity of water. Using the data from ethylene glycol, we corrected the fast chevrons for the effects of the viscosity of urea, using its coefficient of viscosity (17) (Fig. 5). The effect is small, reducing the slope of the folding limb from -0.19 to -0.14 kcal mol^-1 M^-1. The slope of the unfolding limb was similarly increased by 0.05 kcal mol^-1 M^-1.

Fig. 4.

The effect of solution viscosity on the stability (Upper) and rate constant for the fast phase, which is mainly that for k_UI (Lower).

Fig. 5.

The influence of solution viscosity (caused by increasing viscocity with increasing concentration of urea) on fast phase chevron plots of Pit1 (red, corrected for the solution viscosity; blue, uncorrected for the solution viscosity).

An activation energy of 11 kcal mol^-1 for the fast phase was also indicated from its temperature dependence (Fig. 6). Members of the EnHD family, measured over a limited range where the relaxation constant λ₁ ∼ k_UI and λ₂ ∼ k_IN, had similar activation energies.

Fig. 6.

Temperature dependence of the fast and slow relaxation constants for wild-type Pit1 and the mutants EnHD L16A and EnHD L16A/K52A. The data are plotted under conditions where the observed relaxation times are mainly k_UI or k_IN.

Free Energy of Denaturation.

Fluorescence titration of the denaturation of Pit1 in urea gave ΔG_D-N = 3.49 ± 0.14 kcal mol^-1 and m_D-N = 0.92 ± 0.05 kcal kcal mol^-1 M^-1 (Fig. 2). Circular dichroism titration gave ΔG_D-N = 3.17 ± 0.09 kcal mol^-1 and m_D-N = 0.82 ± 0.02 kcal kcal mol^-1 M^-1. Both were calculated by fitting to a standard two-state equation. However, the kinetics was 3-state, and the presence of low populations of intermediate causes free energy and m-values to be underestimated. It was calculated from the measured rate constants above and their urea dependence that I accumulated to 7.8% in 3.1 M urea. ΔG_D-N in water calculated from the rate constants was 4 kcal mol^-1.

Free Energy of Opening of Protein from ²H/¹H-Exchange.

We measured the free energy of opening of the protein by measuring the protection factors for ²H/¹H-exchange as done previously for EnHD (7) (Fig. 7). Under the reaction conditions, the rate constant for the refolding of the protein is much greater than that for exchange, that is EX2 conditions (18). The equilibrium constant between locally folded sites, which are protected against exchange, and locally or globally unfolded states from which exchange takes place (the protection factor) is calculated as described (18). The free energies associated with the degree of equilibrium unfolding of the protein required for exposure of backbone NHs exchange is higher than that measured assuming a two-state mechanism and is in the region on of 4–5 kcal mol^-1 for the best protected positions (Fig. 7). The free energy of the transition from N → I calculated from the ratios of k_IN/k_NI is 2.4 kcal mol^-1. The folding intermediate is thus protected against exchange.

Fig. 7.

Free energy of opening of Pit1 for ²H/¹H-exchange determined under EX2 conditions. The blue horizontal line is the free energy of denaturation calculated from fluorescence and CD titrations with urea, fitting to a 2-state equation, the black is that calculated from the rate constants for the three-state kinetics. The red horizontal line is the energy calculated for N opening to I. The helices span residues 10–22, 28–38, and 42–57 (20).

Discussion

Chevron Plots on the μs Time Scale.

EnHD folds in biphasic kinetics via the fast formation of a highly helical intermediate whose structure has been directly determined to be that of the HTH motif comprising H2 and H3, with H1 helical but not docked (4). Pit1 had biphasic folding kinetics that paralleled that of EnHD with a 2 μs fast phase that had a t_1/2, which was almost exactly the same as for EnHD. The second phase was 2- to 3-fold slower than for EnHD. The better separation of phases and larger amplitudes of Pit1 have allowed an analysis of the fast folding phase, which is not so easily accessible with EnHD.

Remarkably, chevron plots with distinct curvature were observed, unlike those found so far for any other protein folding on this time scale (19). The slope of the folding limb is very shallow, being only 0.19. The sum of m-values for folding and unfolding of the fast phase is 0.77 ± 0.15, compared with 1.17 ± 0.16 for the overall folding, which equates to 66% of the total burial of solvent accessible surface area on folding being manifested in the fast formation of the intermediate. We have not been able to isolate a stable HTH motif from Pit1, as we had done for EnHD. But, the large fraction of buried surface implies that the intermediate is highly structured. An extensive Φ-value analysis implies an average of 61% (± 15% standard deviation) of the native interaction energies are realized in stabilizing the intermediate (20).

Overall Nature of the Transition State for the Fast Phase.

All data conspired to indicate that the transition state for formation of the intermediate was not compact and had little structure. The rate constant in water for the fast phase was about the same for wild type and a series of mutants of differing stability (Fig. 3). Virtually all of the change of stability of the protein on mutation was manifested in changing the rate constant for fast unfolding—It can be seen by inspection that the Φ-values for folding are close to 0 because the folding rate constants converge to the same value at 0 M urea in Fig. 3. In quantitative terms, the average Φ-value for the fast-formed transition state is ∼0.15, based on the total energetics for just the fast phase. The apparent fractional decrease in surface exposure is given by m_UI/(m_ND), which is equal to 0.19/1.17; i.e., 16%. The viscogen ethylene glycol does slow down folding (Fig. 4), showing that chain diffusion is part of the rate-determining step. Allowing for a small correction for the effects of viscosity of urea on m_UI, we calculate there is just a 12% contraction of solvent accessible surface area on formation of the transition state. Thus, the rate-determining step for folding of the intermediate is partly a chain diffusion-limited step followed by rapid formation of the intramolecular noncovalent bonds.

The rate-determining step for the unfolding of the intermediate involves the near complete breaking of all the stabilizing interactions. Every mutation that changes the stability of the folding intermediate affects the rate constant for its unfolding by nearly the full energy of the interaction. This implies that the unfolding reaction is highly cooperative.

Approaching the Speed Limit for Folding.

Eaton and colleagues predict that the speed limit for a protein of N residues is ∼N/100 μs (9), which would be ∼0.6 μs for Pit1. The observed value of 1/k is 3.3 μs. We have measured the segmental motion in the denatured EnHD, the maximum values of 1/k being 0.11 μs (6), similar to that in BBl, which has an activation energy of 7 ± 1 kcal (21), compared with the 11 kcal/mol for the fast phases of Pit1 and EnHD (Fig. 6). It would appear that the apparent activation energies for the fast step of Pit1 and EnHD are 3–5 kcal/mol above that for chain diffusion. Even so, the dependence of the rate constant for folding on viscosity implies that there is an important contribution from internal friction to the energy barrier (16).

Proteins usually do not obey the Arrhenius law for their folding kinetics because there is a large change in specific heat between the denatured and transition states for folding (22). The specific heat change depends on the change in solvent accessible surface between those states, which was found to be small here and so the Arrhenius law is followed over a wider range than normal.

The results on Pit1 may be compared with those on the folding of an engineered mutant of the 35-residue villin headpiece, which folds in 0.7 μs, with an estimated activation energy of ∼1–2 kcal/mol and little or no temperature dependence (23, 24). The rate constants have little dependence on the concentration of denaturant (19). Some of the differences may result from the smaller size of the headpiece, which will make the kinetics and equilibria less sensitive to denaturant as the energetics depends on the change in solvent accessible surface area on denaturation. Also, the transition state may be further along the reaction coordinate, which would give a flatter chevron plot and give a larger change in specific heat of activation, which would flatten an Arrhenius plot. But, in addition, its transition state may move with temperature and denaturant (19).

Downhill Folding.

Energy landscape theory predicts that a denatured state may fold “downhill” when there is a minimal energy barrier (25)—There is minimal formation of noncovalent bonds to give the transition state for downhill folding. The converse of the classical downhill-folding postulate is that the reverse reaction is “uphill”—the noncovalent bonds must be fully broken in the unfolding reaction. We find some of these features for the folding and unfolding of the intermediate of Pit1. But, there is still a significant energy barrier to folding, which may result from several sequential steps, each having a small barrier. Eaton and colleagues conclude that the folding of the villin headpiece, with its even lower activation energy and faster rate, is still not downhill (24). Our data are also consistent with the protein approaching, but not reaching, the downhill-folding limit.

Methods

Protein Expression.

All chemicals were analytical grade or superior. Unless otherwise stated reagents were obtained from Sigma, BDH, or Fisher Scientific whereas ultra pure urea was purchased from Fluka. The plasmid pSEA100, encoding Pit1 gene under T7 promoter was used to express the protein. Cysteine at the position 50 was mutated to serine to prevent the intermolecular disulfide bond formation and therefore C50S mutant is referred to as wild type. The protein was expressed by standard procedures and purified by using Q-Sepharose XL column (Pharmacia) followed by a Source 30S column (Amersham Biosciences) and the reverse-phase C18 monomeric column using Waters HPLC system. Purity and identity were checked by mass spectroscopy. The genes for mutants were prepared by using QuikChange (Stratagene), and the proteins were produced as described above. All proteins were > 95% pure.

Kinetics.

All experiments were performed in 50 mM sodium acetate, 100 mM NaCl at pH 5.5, 25.0 °C unless otherwise stated. All samples for kinetic experiments were degassed and filtered. The concentration of protein ranged from 100–200 μM. The methodology of Temperature-jump kinetic experiments, equilibrium measurements and ²H/¹H-exchange is as described in refs. 5 and 7. For viscosity measurements the viscosity of urea solutions was calculated according to ref. 17.

Exchange of ²H/¹H was measured using a DRX500 Bruker NMR spectrometer fitted with a cryoprobe. Lyophilized protein was dissolved in buffer (50 mM sodium -acetate, 100 mM NaCl, pD 5.5, 25 °C), placed in a NMR tube and ¹⁵N-HSQC spectra measured from 1–2 min onward. Series of additional HSQC spectra were measured at first every 3 min and then at longer time intervals (10 min) for 5 h. The height of each peak over the time series of experiments was fitted to an exponential decay.

Data Analysis.

Kinetic data were fitted simultaneously to 2 equations being the solutions to general 3-state kinetic scheme (Scheme 1):

where p and q are defined as: p = k_UI + k_IU + k_IN + k_NI, and q = k_UIk_IN + k_IUk_NI + k_UIk_NI.

The denaturant-dependence of rate constant was modeled with the standard linear equation: k = k₀ exp(m[urea]/RT), where k₀ is the value of rate constant in the absence of denaturant, m (either positive or negative) is the slope of the dependence of log k on urea concentration, R is gas constant and T is temperature. Data analysis was performed using Kaleidagraph (Synergy Software) and Prism (GraphPad Software).