Article Information

• PDF (514 kb)
• Full Text with Large Figures

PubMed

• Articles by N. Cremades
• Articles by J. Sancho

Related Articles

• Macromolecular Crowding Induces a Molten Globule State in th...

  Macromolecular Crowding Induces a Molten Globule State in the C-Terminal Domain of Histone H1 Biophysical Journal, Volume 93, Issue 6, 15 September 2007, Pages 2170-2177 Alicia Roque, Inma Ponte and Pedro Suau Abstract We studied the secondary structure of the C-terminal domains of the histone H1 subtypes H1° (C-H1°) and H1t (C-H1t) in the presence of macromolecular crowding agents (Ficoll 70 and PEG 6000) by IR spectroscopy. The carboxyl-terminal domain has little structure in aqueous solution but became extensively folded in the presence of crowding agents. In 30% PEG, C-H1° contained 19% -helix, 28% -sheet, 16% turns, and 31% open loops. Similar proportions were observed in 30% Ficoll 70 and for C-H1t in both crowding agents. The proportions of secondary structure motifs were comparable to those of the DNA-bound domain. Kratky plots of the small-angle x-ray scattering showed that in crowding agents the C-terminus had the compaction of a globular state. Progressive dissipation of the secondary structure and a linear increase in partial heat capacity with temperature together with increased binding of ANS indicated that the C-terminus is not cooperatively folded in crowded conditions. Native-like secondary structure and compactness in absence of folding cooperativity indicate that the C-terminus in crowding agents is in a molten globule state. Folding of the C-terminus in crowded conditions may increase the rate of the transition toward the DNA-bound state and facilitate H1 diffusion inside cell nuclei. Abstract | Full Text | PDF (153 kb)

• Molecular chaperones: Avoiding the crowd

  Molecular chaperones: Avoiding the crowd Current Biology, Volume 7, Issue 9, 1 September 1997, Pages R531-R533 R.John Ellis Summary The involvement of two types of molecular chaperone in folding newly synthesized proteins can be rationalized in terms of the crowded nature of the intracellular environment. Recent work sheds light on how these chaperones recognise their substrates and protect them from the problems of macromolecular crowding. Summary | Full Text | PDF (302 kb)

• Molten globules

  Molten globules Current Biology, Volume 9, Issue 9, 6 May 1999, Pages R313 Christina Redfield Full Text | PDF (52 kb)

• …more

Copyright © 2008 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 95, Issue 4, 1913-1927, 15 August 2008

doi:10.1529/biophysj.108.130153

Proteins

Molten Globule and Native State Ensemble of Helicobacter pylori Flavodoxin: Can Crowding, Osmolytes or Cofactors Stabilize the Native Conformation Relative to the Molten Globule?

N. Cremades and J. Sancho^,  

Biocomputation and Complex Systems Physics Institute and Departamento de Bioquímica y Biología Molecular y Celular, Facultad de Ciencias, Universidad de Zaragoza, Zaragoza, Spain

Address reprint requests to J. Sancho, Departamento de Bioquímica y Biología Molecular y Celular, Facultad de Ciencias, Universidad de Zaragoza, Zaragoza, Spain.

It has been suggested that partly unfolded conformations of proteins that are close in energy to the native state under native conditions are relevant for protein function and in the onset of protein conformational defects 1,2. At this time, NMR techniques, such as hydrogen exchange dynamics, have been used to detect and characterize partly unfolded forms (PUFs) in the native ensemble 3,4,5. Classical thermodynamic analysis of partly unfolded forms arising under a variety of solution stress conditions can also shed light on the structure and energetics of these conformations and help to define native state ensembles. In this respect, global analysis of thermal unfolding equilibria 6,7 can provide good estimations of free energy differences between native and partly unfolded conformations, and can even delineate low-resolution structures of intermediates 6,8,9. One partly unfolded form of proteins, the molten globule, has received much attention due to its fascinating spectroscopic properties and to proposals that have implicated it in a variety of functional roles, such as protein folding 10,11 and protein membrane translocation 12,13. Yet, it is still difficult to know whether molten globules belong to native state ensembles at neutral pH or whether molten globules or other partly unfolded forms can respond to native stabilizing conditions, such as crowding, osmolytes, or cofactor binding, in the same fashion as native states. The fact is that although the stability of molten globules against full unfolding has been investigated over the years, less is known about the relative stability of native and molten globule conformations of proteins.

We use here the flavodoxin from the pathogenic bacteria Helicobacter pylori to investigate these issues. H. pylori is the only living entity, to our knowledge, able to survive in the human stomach, which is because of the presence of an H⁺-stimulated cytoplasmic urease 14,15. The cytoplasmic pH of H. pylori has been proposed to vary between 8 and 5 depending on external pH and urea concentration in the medium 16,17,18. Since fluctuations in pH and urea concentration might exert a strong influence on the relative populations of the native state ensemble 2, it is possible that the bacteria use in vivo stabilizing mechanisms, such as macromolecular crowding or osmolytes, to counterbalance potential transient destabilizations of the native state. Flavodoxins are electron transfer proteins involved in many different reactions 19. H. pylori flavodoxin shuttles electrons from pyruvate to NADP⁺ in an electron transport pathway specific for epsilon-proteobacteria 20. It has high sequence and structure homology with other members of the flavodoxin family 21,22 but a more exposed flavin mononucleotide (FMN) binding pocket due to the presence of an alanine residue where a conserved tryptophan is typical for most flavodoxins. This feature, which is being exploited to target new specific inhibitors for H. pylori eradication 23, makes the affinity for the FMN cofactor the lowest we have found for a long-chain flavodoxin (K_d=4.4nM at pH 7.0). Stability studies have demonstrated that H. pylori apoflavodoxin has both the highest stability toward full thermal unfolding and the lowest cooperativity (with two partially unfolded intermediates) we have yet observed for flavodoxins 24. Recently, we reported that at pH 2.0, it also populates a monomeric, partially unfolded conformation with molten globule characteristics 25 that seems structurally different from the two thermal intermediates identified at neutral pH. A similar conformation was described at acidic pH for the apoflavodoxin from Anabaena, which could not be characterized because it was not monomeric 26. However, a truncated version exhibiting molten globule characteristics at neutral pH was investigated by equilibrium phi-analysis 8,27 and was shown to be homogenously expanded and to display a nativelike secondary structure. Very little is known of the energetics of the equilibrium between the molten globule of H. pylori apoflavodoxin and the native state of known crystal structure 21,22 and dominant at neutral pH 25.

Since the apoflavodoxin molten globule is monomeric and unfolds in a reversible manner 25, it constitutes an interesting model to investigate molten globule energetics, and to determine whether molten globules belong to the cohort of partly unfolded conformations, close in energy to the native one, that form the native state ensemble 2,28. In addition, it provides a means to investigate whether common natural mechanisms that are known to stabilize native proteins against full unfolding are similarly effective to stabilize and recover functional native states in conditions where partially unfolded conformations become populated. According to our analysis, H. pylori flavodoxin remains functional between pH 2.0 and 10.0 thanks to the strong, native-specific ability of the FMN cofactor to bind to the native state, which compensates for the fact that at low pH the molten globule becomes more stable than the native conformation of the apoprotein. However, neither crowding agents nor osmolytes can significantly stabilize the native state relative to the molten globule, which indicates that these physiological mechanisms are not effective to minimize partial unfolding events leading to intermediates in the cell. A global analysis of all available thermodynamic information on the folding and binding equilibria, using data presented here combined with those from previous studies 24,25, has allowed us to delineate the dynamics of the free energy landscape of H. pylori flavodoxin along the pH axes. At neutral pH, the ensemble contains, in addition to the native state, the molten globule (this work) and two previously characterized equilibrium intermediates 24. Incidentally, the intermediate of the ensemble closer in free energy to the native state is the only one that has been observed to become dominant under mutational stress in the structurally homologous Anabaena flavodoxin 29,30.

Recombinant H. pylori flavodoxin was purified from a culture of E. coli cells (BL21) harboring the pET28a plasmid, which contains the flavodoxin gene 23. Apoflavodoxin was separated from holoflavodoxin in a MonoQ10column (FPLC, Amersham, Piscataway, NJ), with a 0- to 1-M linear gradient of NaCl in Tris-HCl, pH 8. Apoflavodoxin was also obtained from holoflavodoxin by precipitation with trichloroacetic acid 31.

UV fluorescence emission and far-UV and near-UV circular dichroism (CD) spectra were acquired at 25±0.1°C at several pH values in both native and denaturing conditions (6 M urea or 90°C). Fluorescence emission spectra (300–400nm, with excitation at 280nm) were recorded in an Aminco-Bowman Series 2 spectrometer, and circular dichroism spectra in a Jasco 710 (Tokyo, Japan) or an Applied-Photophysics Chirascan (BioTools, Jupiter, FL) spectropolarimeter. Protein samples at different pH values were prepared by mixing concentrated buffer and protein solutions. Apoflavodoxin concentrations were 2–20μM for fluorescence, 20μM for far-UV CD, and 20–40μM for near-UV CD experiments. All final buffer solutions were of 10mM ionic strength. For fluorescence and near-UV CD, a cuvette of 1cm pathlength was used, and for far-UV CD, a cuvette of 0.1cm pathlength.

The chemical stability of apoflavodoxin has been determined in the 1.5–5.0pH interval, using urea as a denaturant. Protein samples (2μM for fluorescence and 20μM for far-UV CD) were prepared at the desired pH and urea concentration by mixing 900-μL urea solutions with 100-μL aliquots of buffered apoflavodoxin. The ionic strength of the different buffers in the final protein solutions was 10mM. Protein samples were equilibrated at 25°C for at least 30min before recording their spectroscopic signals. Unfolding curves were recorded throughout the 1.5–5.0pH interval by emission fluorescence (ratio of 320/360nm emission, with excitation at 280nm). Additional unfolding curves were recorded at pH 2.0 and 5.0 by far-UV CD (at 222nm) and at pH 5.0 by near-UV CD (at 291nm). The influence of crowding on protein stability was determined by recording urea unfolding curves using protein samples that also contained dextran (35,000–45,000 average mol wt) (Sigma, St. Louis, MO) at a concentration of 100mg/mL.

Unfolding data were analyzed as described 32,33, assuming a two-state equilibrium where the free energy of unfolding, ΔG, is considered to be a linear function of denaturant concentration, D:

(1)

(2)

(3)

Individual fitting of the different curves obtained at a single pH value to the three-state model was unsatisfactory, because each of these curves shows an apparent two-state behavior. We thus performed a combined analysis of the curves obtained at pH 2.2, 2.6, and 3.17. In the analysis, the value of each curve was previously calculated from a spectral deconvolution (see below) and fixed in Eq. (3). On the other hand, the m₁ and m₂ parameters, representing the urea dependencies of the free energies, were kept the same for the three pH values, since no significant changes are expected in the narrow 2.2–3.2pH interval. To further reduce fitting dependencies, the unfolded state signal (S_U, m_U) was treated as a shared parameter for the three curves, whereas the slope of the signal of the intermediate state (i.e., the molten globule) was fixed at zero 34.

Thermal denaturation at pH 2.0 (in 23mM sodium phosphate buffer) was assessed by fluorescence emission (emissions ratio 320/360nm, which is proportional to the advance of the unfolding, because no change in intensity takes place at 360nm 8) and far-UV circular dichroism (222nm) for apoflavodoxin. Apoflavodoxin concentration was 2–20μM for fluorescence measurements and 20μM for far-UV CD. Global fit of the spectroscopic unfolding curves of the apo form was performed assuming a two-state unfolding model,

(4)

(5)

For holoflavodoxin, the unfolding was also assessed by near-UV CD (291nm) and absorbance at 291nm. Holoflavodoxin was always used at 20μM concentration. The clear nonsuperimposition of the unfolding curves monitored by different spectroscopic techniques suggested a three-state model for the equilibrium unfolding of the holoprotein. The equation used to globally fit the curves was

(6)

To obtain an accurate value for the unfolding heat capacity change of the molten globule, thermal unfolding at pH 2.0 was carried out in the presence of different urea concentrations (0, 0.4, 0.8, 1.0, 1.2, 1.4, and 3.0M). A global fit of these curves was performed using a modified version of Eq. (4), assuming that S_N, m_N, S_U, and m_U show a linear dependence of urea concentration and that the free energy of unfolding follows:

(7)

(8)

(9)

(10)

The populations of native state and molten globule at 25°C, in the absence of denaturant, and at different pH values (1.8, 2.0, 2.2, 2.4, 2.5, 2.75, 3.1, 3.7, 4.1, and 5.0), were calculated by deconvolution of far-UV CD spectra, using

(11)

(12)

(13)

Once the populations of the molten globule and native states in the pH 1.5–5 interval are known, their stabilities relative to the unfolded conformation have been determined from urea unfolding curves at different pHs that have been globally fitted to a three-state model in the pH transition region (pH values 2.2, 2.6, and 3.2), and to a two-state model at pH 1.6 and 2.0 (molten globule/unfolded) and pH 3.9, 4.4, and 5.0 (native/unfolded). The stabilities of the molten globule (ΔG_MG−U) make it possible to determine the number of residues, n₂, in the molten globule whose pK_a in the unfolded state is sufficiently different to modify the proton content in the states and the mean pK_a values in those states, and using Eq. (14).

(14)

According to the scaled particle theory 40,41, the difference in free energy between two conformations of a given protein (herein the native and molten globule states of apoflavodoxin: ΔG_MG−N) is influenced by excluded volume effects exerted by crowding agents (herein dextran) according to

(15)

(16)

(17)

The thermodynamic parameters of the interaction between molten globule apoflavodoxin and its FMN cofactor were determined in 23mM phosphate buffer pH 2.0 at 25°C by fluorescence spectroscopy and isothermal titration calorimetry (ITC) (VP-ITC calorimeter, Microcal LLC, Northampton, MA). The FMN used in both ITC and spectroscopic determinations was supplied by Sigma-Aldrich (>95% pure). In the ITC experiments, a 10-μM flavodoxin solution was loaded in the calorimetric cell and titrated with 90μM FMN dissolved in the same buffer. Both solutions had been previously degassed. Between injections, it took 400s to recover the baseline. On the other hand, the quenching of flavin fluorescence at 525nm upon addition of small aliquots of apoflavodoxin was recorded and the following equation was used to fit the fluorescence data:

(18)

At low pH, H. pylori flavodoxin populates a molten globule conformation that is compact, rich in secondary structure, and devoid of, or perhaps exhibiting very weakened, tertiary native interactions, and which displays exposed hydrophobic patches where the fluorescent probe ANS can bind 25. It is slightly expanded relative to the native state (30% volume increase) and remains monomeric at pH 2.0 25, unlike the molten globule previously found at acidic pH for the homologous Anabaena apoflavodoxin. H. pylori apoflavodoxin molten globule thus constitutes a fine model to investigate the energetics of this kind of conformation, which have been implicated in the mechanism of protein folding 10,11 and a variety of physiological processes, such as protein membrane translocation 12,13 or enzyme inactivation 13,45, or as constituents of substrates of chaperonins 46. It has also been proposed that they participate in the mechanism of amyloid fibril formation 47. On the other hand, it has become increasingly clear that, in native conditions, the native state of many proteins is accompanied by a cohort of partly unfolded conformations of low population but close in energy and ready to become dominant under a variety of stress conditions 2. Whether molten globules belong to the native ensemble remains to be determined. Here, we have investigated the conformational stability of H. pylori apoflavodoxin molten globule and its relationship to the native state ensemble, which for flavodoxins includes up to two partly unfolded conformations that become dominant upon heating 6,24 and, for one of them, upon mutation 29,30. Stability has been first studied at pH 2.0 using urea-induced and thermally induced unfolding. Both chemical and thermal unfolding are reversible, as is shown by the fact that the unfolding curves obtained from refolded protein samples are very similar to those obtained from fresh samples (same thermodynamic parameters with a signal lost of ∼10%, not shown). In addition, no protein concentration dependence was noticed (from 2μM to 20μM), which indicates that the unfolding is free from complicating oligomerization or aggregation processes. For urea-induced unfolding, fluorescence and far-UV CD curves are superimposable (Figure 1B) and have been globally fitted to a two-state model (Eq. (2)). The stability of the molten globule is very low (1.13±0.05kcal/mol; mean of three experiments ± SE); and the m value of 1.33±0.043kcal/(mol×M), well below the m value of 2.32±0.02kcal/(mol×M) described for the native conformation at pH 5 25, is consistent with the molten globule being more hydrated than the native state.

Display large version of this figure

Figure 1

Apoflavodoxin unfolding at pH 2.0. (A) Urea-induced unfolding and (B) thermally induced unfolding of the molten globule state of apoflavodoxin followed with fluorescence (squares) and far-UV CD (circles). For a better visual comparison, data were normalized so that the signals are from roughly 0 to 1. The lines represent the global fit to a two-state unfolding model.

Fluorescence and far-UV CD thermal unfolding curves (Figure 1B) can also be superimposed and have also been globally fitted to a two-state model (Eq. (4)). The molten globule thus exhibits a much simpler thermal unfolding behavior than the native conformation, which unfolds via a four-state equilibrium 24. The temperature of mid-denaturation (49.6±0.5°C; mean of three global fits ± SE) is not particularly low, and it lies between the values of T_m1=31.9 and T_m2=57.9°C described for the native conformation at pH 9.0 24. The enthalpy change (25±1kcal/mol) is nevertheless much lower than that of the native state at pH 9.0 at the same temperature (∼85kcal/mol at 49.9°C), which is expected for a conformation with debilitated tertiary interactions. The thermal unfolding has not been studied by differential scanning calorimetry because, due to the low enthalpy change, the protein concentration required is high and leads to protein precipitation at high temperatures (unlike in the low protein concentration range used for spectroscopic thermal unfolding). This phenomenon, observed at high protein concentrations and high temperatures, seems to be preceded by conversion into a conformation of high beta content (not shown).

To extrapolate the stability data to lower temperatures and compare thermal and chemical unfolding, the ΔC_P value is required. Often, ΔC_P values derived from spectroscopic thermal unfolding curves are completely unreliable 48. However, when cold denaturation is observed within the experimental temperature window, the ΔC_P values so obtained are reasonable, because cold denaturation is strongly linked to ΔC_P. As shown in Figure 1B, cold denaturation of the molten globule is clear and the ΔC_P value derived from the global fit of the thermal unfolding curves (1.1±0.2kcal/(mol×K); mean of three determinations ± SE) should be a good approximation. Nevertheless, to obtain a more precise ΔC_P value, thermal unfolding of the molten globule has been carried out in the presence of different urea concentrations, which induces a more pronounced cold denaturation in the experimental temperature window (Fig. 2). At 1.4M urea, the unfolding curve is almost symmetrical, showing two clear transitions, one corresponding to heat denaturation and the other, at low temperatures, to cold denaturation. At 3M urea, the protein is completely unfolded, which is consistent with the urea unfolding curve in Figure 1A, and no thermal transition is observed. The smooth linear temperature dependence of the fluorescence of the denatured state at 3M urea from 0 to 85°C (at 320nm—not shown, but see Figure 2A for the ratio of emissions at 320 and 360nm) suggests that the cold- and heat-denatured states may exhibit a similar degree of compaction. Global fits of the curves to either the linear model 35,36,37 or the urea binding model 38,39 are good (Figure 2A), the global fitting error being slightly lower for the linear model. The two models yield similar values for the thermodynamic parameters of molten globule thermal unfolding in the absence of denaturant (T_m=50.4±0.3 or 47.9±0.7°C, ΔH_m=26.8±0.8 or 23±1kcal/mol, ΔC_P=1.09±0.02 or 0.98±0.03kcal/(mol×K) for the linear and urea-binding models, respectively) (Figure 2A). At 25°C, the linear model provides an m value of 1.36kcal/(mol×M), in excellent agreement with the value obtained from simple urea unfolding curves: 1.33±0.04kcal/(mol×M) (Figure 1A). The urea-binding model estimates that ∼143±40 urea molecules bind preferentially to the unfolded state relative to the molten globule, with a binding enthalpy of −1.76±0.09kcal/mol and a binding entropy of −0.014±0.005kcal/(mol×K) per urea molecule, in good agreement with the values proposed by Makhatadze and Privalov (−2.1±0.5kcal/mol and −0.013±0.002kcal/(mol×K), respectively) for urea unfolding of native proteins 39. Therefore, preferential binding of urea to the unfolded state leads to significant decreases of the apparent enthalpy and entropy of unfolding, which makes cold denaturation take place at higher temperatures. On the other hand, the heat capacity change of urea binding to the protein is calculated at 0.04±0.001kcal/(mol×K), which means that urea increases the curvature of the stability curve 49, thus further raising the temperature of cold denaturation of the molten globule. From the thermodynamic values calculated using the linear model (see Table 4), the conformational stability of the molten globule at 25°C in the absence of denaturant is of 0.99±0.07kcal/mol, in good agreement with the value derived from chemical denaturation analysis (1.13±0.05kcal/mol). Overall, the thermal and chemical stability data consistently indicate that at 25°C and pH 2.0, 85% of the apoflavodoxin molecules are in the molten globule conformation and 15% are cold-denatured.

Display large version of this figure

Figure 2

Cold denaturation in apoflavodoxin. (A) Thermal unfolding curves of apoflavodoxin at pH 2.0 in the absence of urea (squares) and in the presence of 0.4 (circles), 0.8 (triangles), 1 (rhombus), 1.2 (stars), 1.4 (crosses), and 3 M (open arrowheads) urea. All curves were globally fitted to the two-state model for thermal unfolding coupled to either the linear extrapolation method (solid lines) or the urea-binding method (dotted lines) for chemical unfolding. (B) The molar fraction of the molten globule state obtained for each experimental curve and its theoretical value calculated using the parameters obtained for the two-state unfolding model coupled with the linear extrapolation model for the interpretation of urea influence (solid lines).

It has been reported that ΔC_P and m values of protein unfolding are correlated to protein size 32,50. Apoflavodoxin from H. pylori contains 164 residues. The ΔC_P of the native state at pH 9 (calculated, as described by Cremades et al. 24, to be 2.7kcal/(mol×M)) is consistent with a protein of ∼147 residues, whereas the ΔC_P of the molten globule at pH 2.0 corresponds to that of a native protein of 70 residues. It thus seems that the molten globule is hydrated so that about half of the residues behave as if they where exposed to solvent as much as in the denatured state. Exactly the same picture is provided by the m values, which predict a length of 182 residues for the native state but only 96 for the molten globule 32,50.

At pH 5.0, the native conformation of H. pylori apoflavodoxin, of known x-ray structure 21 and well characterized thermodynamic behavior 24,25, is largely dominant. However, as the pH drops, the stability balance between native state and molten globule switches, so that at pH 2.0 most folded protein molecules are molten globules. At intermediate pH, different populations of both states are expected. To determine the free energy difference of the molten globule/native equilibrium and the influence of pH therein, a careful spectroscopic and thermodynamic global analysis has been performed combining data from the pH range 1.5–5.0. First, native and molten globule molar fractions at the different pHs in the transition region have been calculated from deconvolution of apoflavodoxin spectra recorded at different pH values from 1.5 to 5. Deconvolution was based on far-UV CD spectra (Figure 3A, inset). Fluorescence and near-UV CD spectra were not used due to strong influence of protonation in fluorescence emission intensities and weak near-UV CD spectra, respectively. The native fraction as a function of pH is shown in Figure 3A and the free energy difference between native and molten globule in Figure 3B. The transition from native state to molten globule is driven by preferential proton binding to the molten globule of around five acidic residues of the protein, the pK_as of which are higher in the molten globule (2.9) than in the native state (2.0). The reason for increased pK_as in the molten globule can be either that the affected residues are more buried in the molten globule than in the native state, which seems unlikely, or that they establish specific interactions that stabilize the ionized forms in the native state, but are broken in the molten globule conformation. As the pH drops, the increase in proton concentration promotes a preferential stabilization of the molten globule, which is more easily protonated than the native state. A tentative identification of the acidic residues involved in the native/molten globule transition has been made by calculating the pK_as of all apoflavodoxin ionizable residues in the native structure 21 using the program PROPKA 51. Six candidate residues are Asp-33, Asp-63, Asp-75, Asp-88, Glu-98, and Asp-140, the pK_as of which are calculated to be <3.1. These residues appear to be spread on the protein surface (not shown). Therefore, if the prediction is correct, the H. pylori apoflavodoxin molten globule could be a homogenously destabilized conformation, as has been described for a molten-globule-like shortened version of Anabaena apoflavodoxin, whose low-resolution structure was solved by equilibrium Φ-analysis 8. On the other hand some of these residues are located at or near the FMN binding site, which suggests that this site could be severely distorted in the molten globule. Two contrasting structural models have been proposed for molten globules: expanded conformations with locally stable native secondary-structure subdomains that are partially folded but permit water penetration 52, or conformations with a dry, native-like hydrophobic core plus an unfolded surface 53. Either model can explain both the increased volume and the low heat capacity and m values of the apoflavodoxin molten globule. However, its close-to-native helical content argues in favor of the first model, because the helices of its α/β fold are located at the surface 21.

Display large version of this figure

Figure 3

The native/molten globule quilibrium as a function of pH. (A) Native state molar fraction obtained at 10 different pH values from 1.8 to 5 from deconvolution of far-UV CD spectra (inset). Solid lines represent the fit of the data to Eq. (12). (B) pH dependence of the free energy difference between native and molten globule states obtained from fitting native fraction data to Eqs. (12).

The stabilities of the native and molten globule states relative to the unfolded one have also been determined, as explained in the Methods section. The global fit to the three-state model at pH values with significant molar fraction of the three species (pH 2.2, 2.6, and 3.2) is good (Figure 4A). The populations of each state at these intermediate pH values are shown in Figure 4BD. The overall stability of the protein (ΔG_N−U=ΔG_N−MG+ΔG_MG−U) at different pHs was fitted to Eq. (13) plus Eq. (14), and from this fit, the number of molten globule residues involved in its pH-induced unfolding transition has been determined (Table 1). Molten globule unfolding is driven by preferential binding of protons to the denatured state. Three side chains whose pK_as are lower in the molten globule (2.9) than in the unfolded state (3.9) are involved. These groups are most likely carboxylates, but the lack of a structural model for the molten globule precludes any tentative assignment. That the mechanism driving molten globule unfolding at low pH is the same one that promotes conversion of the native state into the molten globule highlights the fact that the differences between native protein conformations and molten globules are in many respects of a quantitative rather than qualitative nature 27.

Display large version of this figure

Figure 4

Three-state apoflavodoxin urea unfolding at pH values where native, molten globule, and unfolded states are significantly populated. (A) Global fit of urea-unfolding curves at pH 2.2, 2.6, and 3.2 to a three-state unfolding model (Eq. (3)). (B–D) Calculated populations as a function of urea concentration of native (thick solid lines), molten globule (dashed lines), and unfolded states (thin solid lines) at pH 2.2, 2.6, and 3.2, respectively.

Table 1 Parameters defining the conformational stability of native and molten globule H. pylori apoflavodoxin as a function of pH ()

	n1				n2
	4.9±0.6	2.0±0.3	2.9±0.1	2.7±0.32	2.8±0.2	3.9*	6.4±0.2

Free energies are given in kcal/mol. Fitting errors are reported. Temperature and ionic strength are 25°C and 10mM, respectively.

* The value was fixed during the fitting to reduce dependency between parameters. It represents a weighted average of the expected pKas of aspartic and glutamic residues in unfolded H. pylori flavodoxin.

Macromolecular crowding has been shown to stabilize compact forms of proteins 41,54,55. According to excluded volume theory 55, crowding agents provide a nonspecific force that promotes a reduction of total excluded volume, leading to stabilization of the most compact conformations of proteins. Therefore, the higher the volume difference between conformations, the greater the stabilizing effect. According to the scaled-particle theory 56,57,58, the free energy difference between two protein conformations depends linearly on crowding agent concentration. For the molten globule/native state equilibrium in H. pylori apoflavodoxin, the m parameter in Eqs. (16) is expected to be 0.00093 (kcal×L)/(mol×g), much lower than the value of 0.0156 (kcal×L)/(mol×g) calculated for the unfolded/native equilibrium. With those m values, a 1kcal/mol stabilization of the native state relative to the molten globule would require 1000g/L dextran in the media, whereas for global unfolding of the native protein, such stabilization would be achieved with only 50g/L.

To test the performance of the theory and to investigate whether crowding can induce a significant stabilization of native proteins relative to equilibrium intermediates close in energy, the effect of dextran on the stability of H. pylori apoflavodoxin was tested in several pH conditions that favor either the molten globule or the native state, and also at a pH value where the two conformations display the same molar fraction (pH 2.55). At this pH value, different dextran concentrations from 0 to 250g/L were used, and spectroscopic properties (fluorescence emission maximum wavelength and mean residue ellipticity at 222 and 208nm) were carefully monitored to try to sense slight variations in native and molten globule populations (Figure 5A). Within the 0–250g/L dextran concentration range, the wavelength of maximal emission remains constant at 328.5nm, compared to 323.5nm for the native conformation at pH 5.0, and the ellipticity similarly remained constant. Therefore, 250g/L dextran seems unable to significantly shift the conformational equilibrium toward the native state.

Display large version of this figure

Figure 5

Stabilization of native and molten globule apoflavodoxin by dextran. (A) Spectroscopic signals of apoflavodoxin at pH 2.55 (50% of the protein in the native state and the other 50% in molten globule conformation) at different concentrations of dextran (solid symbols): the wavelength of the maximum in fluorescence emission intensity (squares) and the CD signal at 222nm (circles) and 208nm (triangles). For comparison, the signals of the protein at pH 5 (100% of the protein in the native state) are also shown as open symbols. (B) Urea-unfolding curves of apoflavodoxin in the absence (open symbols) and presence (solid symbols) of 100mg/mL of dextran, assessed by fluorescence emission at pH 2 (squares), pH 5 (circles), and pH 7 (triangles). Black lines are individual fits of the data to Eq. (2).

On the other hand, the influence of 100g/L dextran on the conformational stability of the molten globule at pH 2.0 and of the native state at pH 5.0 and 7.0 was determined by urea denaturation (Figure 5B) and compared with the predictions from the scaled-particle theory (Table 2) together with the excluded volume theory by Ogston 42. Since both molten globule and native state experience large volume increases when they unfold 25, theory predicts significant and similar stabilizations at ∼1.5–1.6kcal/mol. Indeed, similar stabilization of 0.8–0.9kcal/mol (see Table 2) was observed for the native state at pH 5 and 7, and the molten globule at pH 2 is also significantly stabilized by ∼0.5kcal/mol. The lack of perfect agreement between the lower observed stabilizations and theory predictions may be attributed to various things, including inaccuracy of the value of the radius of the unfolded conformation introduced in Eq. (17) and the possibility that the unfolded state at low pH is more compact than at higher pH values. Whatever the reason, it seems clear that the two compact conformations analyzed here (the native state and the molten globule) are notably stabilized by crowding effects against full unfolding. In a similar way, strong stabilization of a homologous flavodoxin by crowding agents has been described recently 59 using analysis of circular dichroism thermal unfolding curves, which in flavodoxins typically monitor the conversion of a compact intermediate into the fully unfolded state 59 and therefore report on a conformational transition of large volume change. In contrast, we have shown here that, according to our spectroscopic analysis and also according to theory predictions, the relative stability of the native and molten globule states in solution conditions where they coexist (pH 2.55) does not appear to be modified by crowding agents. This is an important issue, because many proteins are transformed, under a variety of stress conditions, into partially unfolded conformations (such as molten globules 27, thermal equilibrium intermediates 6,24, or partly unfolded mutants 29,30) rather than the unfolded state. It should be borne in mind that for proteins experiencing non-two-state equilibrium unfolding, it is useful to distinguish between the relevant stability associated with the N↔I equilibrium and the residual stability of the intermediate conformation (I↔U), the two stabilities that add up to the global stability of the protein (N↔U) 60,61. According to scaled-particle theory and to the observations reported here, such proteins, which may well be a majority within the proteome, will not benefit from crowding as a stabilizing force, because the volume changes involved in the N↔I transitions will not be large in many cases.

A variety of organisms use protein-stabilizing osmolytes, such as trimethylamine N-oxide and the disaccharides α, α-trehalose and sucrose 62, to survive at extreme temperatures, in dehydration, and under other stress conditions 63,64,65,66. Proteins are more stable in the presence of osmolytes, because osmolytes preferentially increase the Gibbs energy of the protein denatured state 67,68,69,70,71, likely due to unfavorable interactions between polypeptide backbone and osmolyte (the so called “osmophobic effect” 69). Examples of mechanisms proposed to account for this preferential exclusion include solvophobicity 69,72, surface tension 73,74,75, excluded volume 76,77, water structure changes, and electrostatic repulsions 68. Recently, the effect of sugars on the conformation of several unfolded proteins has been studied and it has been proposed that volume exclusion by high concentrations of sugars can induce a compact molten globule conformation from the acidic unfolded state of some proteins 78,79. In this study, we tested whether osmolytes can similarly induce the transformation of molten globules into native structures. We first tried trimethylamine N-oxide 63,80,81, but it makes the molten globule aggregate, as do several other salts we have tested (not shown), which prompted us to use a neutral osmolyte such as sucrose. Shifts in the native/molten globule equilibrium at pH 2.55 (where they are present in equimolecular concentrations) were investigated in 20μM apoflavodoxin solutions by recording the CD signal at 222nm as a function of sucrose concentration (from 0 to 1.5M). Although the helical content increases at high sucrose concentrations approaching that of the native state, the expected concomitant decrease of the minimum at 208nm 25 (Figure 3A, inset) does not take place (not shown). On the other hand, the wavelength of maximal tryptophan fluorescence emission moves to lower values at high sucrose concentrations, suggesting a lower exposure to solvent of the tryptophan residues. Nevertheless, the same effect is observed at pH 5.0, where the native conformation is prevalent (not shown). It seems thus that sucrose increases both molten globule and native state compactness without driving the molten molecules into the native conformation. The reason this osmolyte can drive conversion of unfolded proteins into molten globules 78,79 but fail to drive conversion of molten globules into native proteins (this work) may also be related to the volume differences between native proteins and molten globules being smaller than those between molten globules and unfolded states.

Osmotic stress analysis 82 is sometimes used to determine hydration changes from osmolyte-induced changes in equilibrium constants. Our data could be interpreted, within the framework of osmotic stress analysis, as indicating that molten globule and native apoflavodoxin are similarly hydrated, which seems unlikely considering their very different m-values of urea unfolding.

Preferential ligand binding to a specific protein conformation stabilizes that conformation relative to alternative conformations 83,84,85,86. To determine whether FMN, the natural flavodoxin cofactor, can bind to the apoprotein in acidic conditions and rescue the native state, FMN was added at different concentrations to a 20μM apoflavodoxin solution at pH 2.55. An equimolar 20-μM concentration of FMN was enough to recover the native wavelength of maximal fluorescence emission and the native CD signals at both 222 and 208nm, which became almost identical to those exhibited by holoflavodoxin at pH 5 (not shown). FMN can therefore bind to the native state at acidic pH, thus shifting the molten globule/native state equilibrium toward the native conformation.

The affinity of the apoflavodoxin:FMN complex was then determined at pH 2.0, where all the protein is in the molten globule conformation, by both fluorescence quenching of FMN upon apoprotein binding and by ITC (Figure 6AB, respectively), and the corresponding binding parameters, essentially identical for the two techniques, are reported in Table 3. Compared to the functional complex at pH 7.0 24, the apparent affinity is 100 times lower, whereas the enthalpy and entropy of binding are much higher (Table 3). It should be noted, however, that these are not the true binding parameters to native apoflavodoxin because part of the binding energy of the complex is utilized to shift the molten globule/native equilibrium toward the binding-competent native state. The binding energy was therefore corrected by adding up the previously determined free energy difference between native state and molten globule at pH 2.0. As regards the enthalpy and entropy of binding, they are not expected to change significantly with pH. First, FMN binding at pH 7.0 and 9.0 has been reported to take place without proton exchange 24, and second, very similar binding free energies are obtained at pH 2.0 (Table 3, corrected value), 7.0, and 9.0 24. Therefore, we propose that the intrinsic entropy and enthalpy of binding at pH 2.0 are similar to those at pH 7, and that the differences between these values and the apparent values experimentally determined by titration of the molten globule with FMN provide estimations of the enthalpy and entropy contributions (−23.6 and 25.7kcal/mol, respectively) associated with the conversion of the molten globule to the native state at pH 2.0 and 25°C. According to this interpretation, adding up the enthalpic and entropic contributions, the conformational stability of molten globule apoflavodoxin at pH 2.0 relative to the native state is calculated at 2.1kcal/mol, which agrees well with the value directly determined from deconvolution analysis of far-UV CD spectra (ΔG_MG−N=1.64kcal/mol, Figure 3B). The relatively high value of the enthalpy change from the molten globule to the native state (yet only one-half of the unfolded/native enthalpy change at pH 9.0 at 25°C 24) suggests that many internal interactions are formed or strengthened during the conformational change, whereas the unfavorable entropic contribution indicates that ordering of the polypeptide chain offsets the expected increase in solvent entropy.

Display large version of this figure

Figure 6

Rescue of native apoflavodoxin at pH 2.0 by FMN binding. (A) FMN fluorescence quenching upon apoflavodoxin binding and (B) isothermal titration calorimetry were used to measure the affinity constant of the flavodoxin/FMN complex at pH 2.0 and to estimate the thermodynamic parameters of binding.

The equimolecular mixture of apoflavodoxin and FMN at pH 2.0 and 25°C presents spectroscopic properties similar to those at pH 5.0, 7.0, and 9.0 24 (Fig. 7). This spectral coincidence indicates that there is a single holoflavodoxin conformation in the 2.0–9.0 pH interval. The stability of holoflavodoxin at pH 2.0 has been studied by monitoring thermal unfolding curves with different spectroscopic techniques (Figure 8A). The unfolding is reversible provided the temperature is kept below 62°C; otherwise, the released FMN group hydrolyzes and the apoprotein refolds but cannot bind FMN (not shown). The different unfolding curves are not superimposable (Figure 8A), which demonstrates the appearance of an equilibrium intermediate. The intermediate has lost the near-UV CD signal characteristic of the associated FMN cofactor and accumulates at temperatures where the molten globule conformation of the apoprotein is still stable (Figure 8B). That the intermediate corresponds to the molten globule is also shown by the thermodynamic parameters obtained from global three-state analysis of the four holoflavodoxin unfolding curves at pH 2.0 (Table 4). Indeed, the second transition takes place at a temperature and with an enthalpy change nearly identical to those determined for molten globule thermal unfolding (49.6±0.5°C and 25±1kcal/mol). The thermal unfolding of holoflavodoxin at pH 2.0 can thus be described by the scheme

Display large version of this figure

Figure 7

Spectroscopic characterization of holoflavodoxin at different pH values. (A–C) Fluorescence (A), far-UV CD (B), and near-UV CD spectra (C) of holoflavodoxin at 25°C and pH 2 (thick black solid line), pH 5 (gray solid line), and pH 7 (gray dashed line). The spectra of holoflavodoxin at pH 2 at 90°C (dotted line) and apoflavodoxin at pH 2 at 25°C (thin black solid line) are also shown for reference.

Display large version of this figure

Figure 8

Holoflavodoxin thermal unfolding at pH 2.0. (A) Thermal unfolding of holoflavodoxin (equimolecular mixture of FMN and apoflavodoxin, 20μM each) assessed by protein fluorescence (open squares), far-UV CD (open circles), near-UV CD (solid squares), and absorbance at 291nm (solid circles). For a better comparison, the data were normalized so that the different curves run from roughly 0 to 1. The solid lines represent the global fit to a three-state unfolding model. (B) Populations of native (circles), molten globule (triangles), and thermally induced unfolded (squares) states. The populations of the different conformations in the absence of FMN are represented by thick gray lines for the molten globule and thin gray lines for the unfolded state.

Table 4 Thermal unfolding of flavodoxin at pH 2.0

		Tm1 (°C)	ΔHm1 (kcal/mol)	ΔCP1 (kcal/(mol×K))	Tm2 (°C)	ΔHm2 (kcal/mol)	ΔCP2 (kcal/(mol×K))	ΔG* (kcal/mol)
	Apoflavodoxin†	−	−	−	50.4±0.3	26.8±0.8	1.09±0.2	0.99±0.07
	Holoflavodoxin	38.8±0.5	44±1	1.5 ± 0.1	49.8±2	23±3	1.09‡	2.2±0.4

* Global unfolding free energy (Holo↔U or MG↔U) at 25°C. † Parameters obtained by global analysis of apoflavodoxin cold denaturation (data in Fig. 2). ‡ ΔCP2 for holoflavodoxin was fixed to the ΔCP,MG-U value of apoflavodoxin.

The native ensemble of proteins, i.e., the alternative conformations that are not far in stability from the native state under native solution conditions, is attracting much interest because it is increasingly clear that the ensemble is involved in the dynamics associated with protein function and that, on the other hand, it might be related to the occurrence of conformational diseases 2. Valuable information on native state ensembles has been obtained so far by NMR equilibrium studies 3,4,5. Here, we have used conventional protein stability techniques to characterize the native state ensemble of a model protein at acidic pH. Temperature, urea concentration, cofactor concentration, and pH markedly modify the relative stabilities of holoprotein, native apoprotein, and molten globule. On the other hand, the native ensemble of this protein at neutral and mildly basic pH has been characterized previously using similar techniques and was shown to include two partly unfolded intermediates 24. The picture that emerges from these analyses (Fig. 9) provides insight into the ensembles present as a function of pH. It has been pointed out that only conformations with free energies within 3–4kcal/mol of the native-state free energy are likely to become significantly populated under native conditions and thus should be considered part of the native ensemble 2. At neutral and mildly acidic pH values, the native state ensemble of H. pylori apoflavodoxin is constituted by two partly unfolded intermediates plus the molten globule (Fig. 9). The more stable intermediate 24, whose structure is similar to that described for the single Anabaena apoflavodoxin thermal intermediate 6, is so close in energy to the native state that it becomes dominant upon small temperature increases. There is also strong evidence that it becomes the dominant conformation of a variety of partly unfolded point mutants 29,30, and that it appears to be related to the mechanism of FMN binding 21. The second intermediate, of lower stability, appears at higher temperatures, and little is known of its structure except that its helix content is low. The molten globule, with high helix content, is ∼3kcal/mol above the native state and represents the least stable intermediate in the native ensemble. Not surprisingly, it has not yet been observed at these pH values under either heat or mutational stress (point mutations). However, more drastic protein destabilization of Anabaena apoflavodoxin by removal of the terminal helix gives rise to the molten globule conformation 8. Only native apoflavodoxin can directly bind the cofactor, but the alternative conformations in the native ensemble are in equilibrium with native apoflavodoxin, and thus FMN drives them into the native state. Four partially unfolded forms (PUFs) of the homologous apoflavodoxin from Azotobacter vinelandii have been detected and studied by NMR 87. PUF 1 is very similar to the native state, PUF 2 corresponds to the more stable intermediate in H. pylori thermal unfolding, and PUF 4 is essentially an unfolded conformation. It is possible, but uncertain, that PUF 3 is related to the second thermal intermediate in H. pylori apoflavodoxin thermal unfolding, but none of these PUFs appear related to the molten globule. As the pH approaches 4.0, preferential proton binding to the molten globule reduces its energy gap with the native state, and below pH 2.5, the molten globule is dominant. Even at pH 2.0, cofactor binding to the tiny fraction of native apoflavodoxin effectively shifts the equilibrium toward the native apoflavodoxin:FMN complex. The two intermediates that at neutral pH are closer to the native state than the molten globule appear to be destabilized at lower pH, and no longer populate upon heating.

Display large version of this figure

Figure 9

Energy landscape of H. pylori flavodoxin native state ensemble as a function of pH, showing the free energy along the pH axes (with G=0 for native apoflavodoxin) of the different flavodoxin conformations that have been observed to become significantly populated. Native state apoflavodoxin is shown in black; molten globule in red; the more stable thermal intermediate observed at pH 7 and 9 in blue; the less stable thermal intermediate in dark green; the thermal unfolding state in orange (the different degree of unfolding proposed for these conformations in Cremades et al. 24 is schematically depicted); the urea-unfolded state in light green; and the flavodoxin/FMN complex in gray. The molten globule is in pale red in the pH 6–9 range to indicate that its energy is assumed to be that exhibited at pH 5. This assumption is based on the fact that the two histidine residues of the flavodoxin, responsible for the variations in protein stability in this pH region 25, are on the surface 21, and no large differences in histidine protonation are expected between the native and molten globule states. This is borne out by the fact that the molten globule is not observed in the thermal unfolding at neutral-basic pH, which indicates that it is less stable than the two thermal partially unfolded states.

It is clear that much information concerning protein function and dysfunction is confined in the details of the native state ensemble, and that stability differences between the native conformation and less stable, minor conformations of the ensemble may be more important than the overall stability of the native protein relative to the unfolded state 2,60,61. Indeed, partly folded intermediates, such as those shown here to constitute the native state ensemble of H. pylori apoflavodoxin, are thought to be critical intermediates involved in the onset of conformational diseases, including aggregation and subsequent fibril formation 47,88,89,90,91. Although NMR methods are very useful to detect and characterize the conformation of dominant species within the native state ensemble 3,4,5, conventional stability analysis can greatly help to understand the energetics of the ensembles and explain their fluctuations under changing solution conditions.