INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

One of the challenges facing molecular biology is determining the rules that govern the acquisition by a nascent polypeptide chain of its three-dimensional and functional structure. Rapid progress in genome sequencing has made it all the more urgent to solve this problem. However, although protein folding is an extremely active field of research combining aspects of biology, chemistry, biochemistry, computer science, and physics, the detailed mechanisms of folding are not entirely clear (for a review see Brockwell et al., 2000, and references therein). A complete understanding of protein folding requires the physical characterization of both native and denatured states and evaluation of the thermodynamic parameters of the system. This involves obtaining information concerning the structure and dynamics of proteins denatured under various conditions. Over the last ten years or so, a large amount of structural information has been obtained experimentally on unfolded proteins, using various techniques including circular dichroism, fluorescence, nuclear magnetic resonance (NMR), small-angle x-ray, or neutron scattering. In contrast, much less attention has been paid to describing the dynamic properties of the denatured state. The main results obtained come mainly from NMR experiments with ¹⁵N relaxation (Tollinger et al., 2001; Barbar et al., 2001; Bai et al., 2001; Yao et al., 2001) and molecular dynamic simulations combined with NMR (Wong et al., 2000).

In recent years, incoherent quasielastic neutron scattering (IQENS) has been used to describe the internal dynamics of proteins (for a review, see Smith, 2000). IQENS directly probes the internal dynamics of biomolecules on the picosecond time scale, providing information on diffusive motions and the geometry of the motions observed (Bée, 1988). These two components change significantly during denaturation. IQENS is a dynamic technique complementary to NMR and molecular dynamic simulation (Dellerue et al., 2001). This technique has already been applied to studies of the internal dynamics of native-state proteins (Réat et al., 1997, 1998; Fitter et al., 1998; Lehnert et al., 1998; Zaccai, 2000) focusing on the characterization of the dynamical transition around 180-200 K (Doster et al., 1989; Ferrand et al., 1993; Andreani et al., 1995; Fitter et al., 1997; Fitter, 1999; Bicout and Zaccai, 2001), role of hydration water (Fitter et al., 1996; Zanotti et al., 1997, 1999; Pérez et al., 1999), relationship between protein dynamics and thermal stability (Fitter and Heberle, 2000; Fitter et al., 2001; Tsai et al., 2000, 2001), solvent dependence of internal dynamics (Demmel et al., 1997; Cordone et al., 1999; Réat et al., 2000), effects of dynamics on enzyme activity (Daniel et al., 1998, 1999; Dunn et al., 2000), and analysis of vibrational modes (Cusack and Doster, 1990; Diehl et al., 1997; Andreani et al., 1997; Paciaroni et al., 1999; Bizzarri et al., 2001). Few studies have investigated the internal dynamics of the partially or completely denatured states of proteins (Receveur et al., 1997; Kataoka et al., 1999a,b; Russo et al., 2000b; Bu et al., 2000, 2001; Fitter et al., 2001; Tehei et al., 2001).

We have studied the picosecond dynamics of a globular protein in solution during heat-induced unfolding. We used a model protein, neocarzinostatin (NCS), which displays a folding pattern typical of a large protein family, the immunoglobulin fold. NCS belongs to a family of antitumor proteins of bacterial origin that contain a labile chromophore. The protein component, apo-neocarzinostatin, the object of this study, is a 113-amino-acid protein with two short disulfide bridges, located within different loops. Its structure (Adjadj et al., 1992; Kim et al., 1993) essentially consists of a seven-stranded antiparallel -sandwich, which forms a cavity. Equilibrium unfolding experiments, with unfolding induced both thermally and chemically, were performed to characterize the unfolding pathways of NCS in terms of structure. Chemically-induced unfolding was monitored by small-angle neutron scattering, fluorescence, and differential scanning calorimetry (Russo et al., 2001). The results suggest that NCS was highly stable, and that a distribution of states existed, with various degrees of residual structure, during the unfolding process. The highly unfolded state is well described by an excluded volume chain model, as described by Russo (2000), and Russo et al. (2000a). The heat-induced unfolding of NCS was followed by small-angle x-ray scattering, in the 20-80°C temperature range (Russo, 2000; Pérez et al., 2001). It was shown that the native globular protein, in buffered H₂O at pH 7 (radius of gyration R_g = 14 Å), begins to unfold at ~63°C and is completely unfolded at 77°C, displaying Kratky Porod chain behavior (R_g = 26.3 Å). The midrange temperature of the transition is T_m  68°C. Performing the experiment in buffered D₂O shifts the transition by ~3°C toward higher temperatures (Russo, 2000), giving a midrange temperature T_m  71°C. Like chemically induced unfolding, heat-induced unfolding is complex, and may involve a collection of intermediate states leading to the fully unfolded conformation.

We report here an IQENS study of changes in the internal dynamics of NCS during thermal denaturation. We investigated four temperatures covering the range from 20°C to the midrange temperature (71°C), working with a solution in heavy water. Data were analyzed with a model that dissociates local diffusive motions and the Brownian motion of the whole protein in solution (Pérez et al., 1999). The inferred dynamic parameters are a mean of those for the various species present in the solution at each temperature.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Sample preparation

Recombinant apo-NCS (mass = 11079 g/mol) was produced in Escherichia coli and purified as previously described (Heyd et al., 2000). The purified protein was lyophilized, dissolved in D₂O, and extensively dialyzed against D₂O to replace the labile hydrogen atoms by deuterium. The protein was lyophilized again and dissolved in 100 mM phosphate buffer at pD 7. A first series of neutron measurements was performed at temperatures 20.8°C, 61.8°C, and 71°C, with a 58 mg/ml solution of NCS. A fresh solution (concentration, 42 mg/ml) was used for measurement at the beginning of the transition (T = 66.3°C), to minimize the aggregation effects due to the sample being maintained for a long period in the transition zone. We used 100 mM phosphate buffer at pD 7 for background subtraction.

Experimental procedure

The experiment was carried out at the Orphée reactor of the Leon Brillouin Laboratory (Gif-sur-Yvette, France), using the MIBEMOL time-of-flight spectrometer. The spectrometer was operated with an incident energy of 2.27 meV, a wavevector range of 0.3 < Q < 2.0 Ź and an energy resolution half width at half maximum (HWHM) of 0.048 meV. All samples were placed in a slab 1.3 mm thick, oriented at 45° with respect to the incident beam. All samples were measured for 24 h at each temperature, to facilitate statistical analysis. All experimental spectra were corrected for the contribution made by the sample holder. They were also normalized using the vanadium standard and corrected for transmission and geometry effects. The resulting data were analyzed with the LLB programs in (, ) space, where  is the scattering angle and  is the neutron time of flight. Conversely, the equations used to describe the data are expressed in the (Q, ) space, where

₀ is the incident neutron energy and ₀ the wavelength, is the energy transfer between the neutron and the sample, m_n is the neutron mass and D_SD, the sample-to-detector distance.

Data analysis

Solvent subtraction

(1)

1

View larger version (22K): [in this window] [in a new window]   FIGURE 1   Time-of-flight experimental scattering spectra of the protein solution and the corresponding buffer at 20.8°C. The momentum transfer at the elastic peak is Q0 = 4 sin(/2)/ = 1.48 Å1. The resulting curve obtained after subtraction of the buffer contribution according to Eq. 1 gives the contribution of the protein alone.

Incoherent dynamic structure factor

(2)

(3)

(4)

(5)

(6)

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Figure 2 shows two examples of the fitting procedure using Eq. 5 at two different temperatures, 20.8°C and 61.8°C, at which the protein is still in its native state. The narrow Lorentzian function corresponds to the global motion of the protein in solution, and the large Lorentzian function corresponds to internal diffusive motion.

View larger version (19K): [in this window] [in a new window]   FIGURE 2   Experimental scattering spectra corrected for the solvent contribution at two temperatures (A) 20.8°C and (B) 61.8°C. The momentum transfer at the elastic peak is Q0 = 1.48 Å1. The solid line is the fit based on Eq. 5. The two Lorentzian components are shown as dotted lines.

Brownian diffusion

Before analyzing internal dynamics, we checked that global motion was correctly deconvoluted by the fitting procedure. The apparent diffusion coefficient of the protein in solution, D_app, can be inferred from changes in the width of the narrow Lorentzian function L₁(Q, ) with Q. For each temperature, changes in the HWHM of the Lorentzian function L₁(Q, ) appear to follow a quadratic function of Q, as expected for a free diffusion motion, ₁(Q) = D_app · Q² (Fig. 3). The slope of ₁(Q) against Q² gives the value of the apparent diffusion constant, D_app. These values are plotted in Fig. 4 as a function of temperature. For comparison, we also performed quasielastic light scattering experiments, to measure expected changes of the translation diffusion coefficient D_tr with T. Light scattering was measured with a 4-mg/ml solution of NCS in H₂O phosphate buffer at pH 7. For comparison with the time-of-flight results in D₂O, the translation diffusion coefficient values D_tr were normalized, at each temperature, according to the ratio between the viscosities of solutions in D₂O and H₂O, then corrected for the constant 1.27 (Pérez et al., 1999) to account for the effect of rotational motion. We also took into account the shift (3°C) of the transition toward higher temperatures for the D₂O buffer. The inferred values are plotted in Fig. 4. We also calculated expected changes in the diffusion coefficient of the compact protein with T, based on the relationship
where
and R_g is the radius of gyration of the native state, R_g = 14 Å. The calculated values are plotted as a straight line in Fig. 4.

View larger version (21K): [in this window] [in a new window]   FIGURE 3   Changes in 1(Q), the HWHM of the Lorentzian component arising from the protein global motions, with Q2 at each experimental temperature. The slope of 1 versus Q2 is indicated by a line and provides the value of Dapp.

View larger version (14K): [in this window] [in a new window]   FIGURE 4   Changes in the protein apparent diffusion constant Dapp with temperature, as derived from neutron () and light scattering data (). The calculated values, evaluated as described in the text, are plotted as a straight line.

The orders of magnitude of the diffusion coefficients derived from the neutron scattering data appear to be consistent with calculated values and with values deduced from measurements of light scattering. The effect of global motions may therefore be considered to be properly deconvoluted.

Internal dynamics

Quasielastic amplitude

(7)

View larger version (18K): [in this window] [in a new window]   FIGURE 5   (A) Changes in pseudo-EISF A0(Q) at all investigated temperatures, as derived from the data analyzed with Eq. 5. The lines result from the fit of A0(Q), based on a model of free diffusion within a sphere of variable radius a. (B) Distribution of the sphere radius f(a), determined by the fitting procedure, for each temperature.

(8)

1

0.5

View larger version (12K): [in this window] [in a new window]   FIGURE 6   Variation of mean sphere radius, , resulting from the model of free diffusion within a sphere and of the fraction of immobile scatterers, p, as a function of temperature.

Quasielastic width

Mean square vibrational amplitude

View larger version (12K): [in this window] [in a new window]   FIGURE 7   Mean square vibrational amplitude u2 as a function of temperature. Dashed line, expected change in vibrational motion amplitude u2 of the native NCS with temperature.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

We present here the dynamic changes that occur before and during the first steps in the heat denaturation of NCS. We observed slight but clear changes in the internal dynamics of the native NCS protein on the picosecond time scale, as a result of the increase in temperature. Moreover, all internal dynamics parameters, including the amplitude of diffusive motions, the fraction of immobile scatterers, and the mean squared vibrational amplitude, underwent stronger changes during the heat-induced unfolding.

Brownian diffusion

Because the sample was a protein solution, we took into account in data analysis the considerable parasitic contribution of bulk solvent and Brownian motion of the protein. We estimated Brownian motion using a model developed for a native protein, which we assumed, to a first approximation, to be valid in our conditions. This enabled us to determine the protein self-diffusion constant (global translation and rotation).

One technical point concerns the width of the narrow Lorentzian function L₁(Q, ), which takes into account the global diffusion motions of the protein in solution. The values of the HWHM, ₁(Q), were generally below the experimental resolution used (0.048 meV). Thus the characteristic time of Brownian diffusion varies, as a function of Q, between 20 and 60 ps at a temperature of 20°C, and between 7 and 26 ps at T = 66.3°C, whereas the correlation time corresponding to the spectrometer resolution is 14 ps. Although any motion with a correlation time higher than the inverse of the resolution is usually considered to produce elastic scattering and is therefore ignored, it has been shown that this is not the case when dealing with the global motions of a molecule in solution (Pérez et al., 1999). In such cases, the HWHM ₁(Q) can be considered as a broadening of the resolution, which is measurable even if much smaller than the resolution itself.

Internal diffusion

To ensure that we correctly interpreted the data for p (fraction of immobile protons), we analyzed the distribution of nonexchanged hydrogen atoms on the native protein, using the Protein Data Bank file (PDB file: 1noa). The NCS protein consists of 113 amino acids, comprising a total of 1510 atoms: 778 atoms of C, N, O, and 732 H atoms. Only 172 of these hydrogen atoms are exchangeable. The remaining protons were considered to be uniformly distributed over the protein. In the case of native NCS, incoherent scattering arises predominantly from these 560 nonexchanged hydrogen atoms: 24% on the backbone, 39% on the side chains of residues involved in the -sheet structure, and 37% on the side chains of residues in random coil structures.

We inferred from our neutron scattering analysis that the proportion (1  p) of mobile hydrogen atoms was 33% at T = 20.8°C. This value is very similar to the value (37%) calculated for the number of protons involved in the side chains of random coil structures. It therefore seems likely that these protons, which are very exposed to the solvent, display mainly diffusive dynamics on the picosecond time scale. According to this hypothesis, the hydrogen atoms of the backbone and the side chains involved in the beta structure are less mobile, and their dynamics have little effect.

If the temperature is increased to 61.8°C, just below the heat denaturation transition, some hydrogen atoms of the backbone and -sheet side chains progressively acquire enough kinetic energy to diffuse locally, but with a weak amplitude because the protein is still compact. Because these protons explore a restrained spatial domain, it follows that the motions are, on average, more confined than those at ambient temperature.

This interpretation is supported by the results of ¹³C NMR experiments performed at 35°C (Mispelter et al., 1995) and 50°C (E. Adjadj, personal communication) showing that, at 35°C, motions within the -sandwich are more constrained than those within the loops, and that, at 50°C, the backbone becomes more flexible and the residues of the beta structure are less tightly constrained.

During the unfolding transition, we observed a pronounced decrease in the fraction p of immobile hydrogen atoms and in the mean motion amplitude . The decrease in p is easy to interpret, given that, upon partial (or total) unfolding, many more hydrogen atoms are able to diffuse than in the native, structured, state. The decrease in at 71°C is more surprising, because we would expect relatively large amplitudes of diffusion for hydrogen atoms in an unstructured state. However, it should be noted that the value of is strongly dependent on the A₀ values in the small-Q range, where only one experimental point is available. In contrast, p being the asymptotic value of A₀ at high Q, its value can be accurately determined. Moreover, the data analysis based on Eq. 5 may not be perfectly adequate when the population of partly unstructured intermediate state becomes important. This point has yet to be clarified.

Debye-Waller

The Debye-Waller factor was evaluated in our approach as the ratio between the intensity scattered with an energy transfer greater than 1 meV and total intensity at a given value of Q. The expected change as a function of Q², exp(Q²u²/3), was then checked experimentally. The resulting mean square displacement, u², includes only the motions corresponding to energy transfers greater than 1 meV, in particular the vibrational motions that give rise to the inelastic part of the scattered intensity. For the structured native protein, most of the diffusive motions on the picosecond time scale involve an energy transfer of less than 1 meV. Therefore, u² can be considered to account only for vibrational motions. This is important because the u² values usually published account both for diffusive and vibrational motions (Doster et al., 1989; Ferrand et al., 1993; Andreani et al., 1995; Réat et al., 1997, 1998) and appear to be much higher than those found here. However, it is possible to reconcile our measurements of u²_vibr with the measurements of u²_total reported elsewhere by extrapolating to room temperature the linear change in u²_total with temperature before the so-called dynamic transition at ~200 K. Indeed, at temperatures lower than the transition temperature, only vibrational motions can occur, and u²_total is equal to u²_vibr. This linear extrapolation to room temperature, as expected for motions depending on a harmonic potential, gave a value of 0.12 Å² for myoglobin (Doster et al., 1989), 0.10 Å² for superoxide dismutase (Andreani et al., 1995), and 0.15 Å² for bacteriorhodopsin (Ferrand et al., 1993). These values are similar to the value of 0.12 Å² obtained for NCS at 20.8°C.

As shown by the dashed line in Fig. 7, the values of u² before the unfolding transition, at 20 and 61.8°C, are perfectly compatible with the expected linear change in vibrational motion amplitude with temperature. However, a clear departure from this linear law is observed when the protein begins to unfold, with values of u² being higher than expected. One possible explanation for this behavior is the expected change in harmonic potential that occurs as soon as the protein structure begins to change dramatically. The harmonic potential of a partially or totally unfolded protein should be softer than that of the native protein, giving larger motion amplitudes. Alternatively, diffusive motions associated with energy transfers greater than 1 meV may occur in the (partially) unfolded protein. These motions would contribute to the Debye-Waller factor as an additional term to the value of u².

	CONCLUSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

IQNS was used to follow changes in the internal dynamics of a small all- protein, NCS, during the first steps of thermal denaturation. The internal dynamics of the native fold at 21°C is consistent with diffusive motions arising from the side chains of the polypeptide loops external to the protein core. Based on structural data, we believe that the protein backbone is slightly flexible at room temperature, and that the side chains, which are involved in the -sandwich, are globally constrained. It is therefore not possible to detect the movement of these side chains on the time scale used. This is consistent with recent results (Dellerue et al., 2001) obtained with a globular protein, C-phycocyanin, which showed monotonic variation of dynamic parameters with distance from the protein core. If temperature increases to 61.8°C, just below the heat denaturation transition, the backbone of NCS becomes more flexible and the -sandwich residues less constrained. Evidence for this change is provided in particular by the increasing number of protons with detectable diffusive motions. If the temperature is then increased to the half-transition temperature, almost all the protons in the protein acquire the ability to diffuse locally. To complete our analysis of dynamics during thermal unfolding, we plan to study the picosecond dynamics of the fully denatured state. In addition, experiments with different resolutions would provide additional information about the dynamics of the unfolding process.