Mannosylglycerate stabilizes staphylococcal nuclease with restriction of slow β-sheet motions


  • Tiago M. Pais,

    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Av. da República-EAN, Apartado 127, 2780-157 Oeiras, Portugal
    Search for more papers by this author
  • Pedro Lamosa,

    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Av. da República-EAN, Apartado 127, 2780-157 Oeiras, Portugal
    2. Centro de Ressonância Magnética António Xavier, Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Avenida da República EAN, Apartado 127, 2781-901 Oeiras, Portugal
    Search for more papers by this author
  • Manolis Matzapetakis,

    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Av. da República-EAN, Apartado 127, 2780-157 Oeiras, Portugal
    Search for more papers by this author
  • David L. Turner,

    Corresponding author
    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Av. da República-EAN, Apartado 127, 2780-157 Oeiras, Portugal
    • Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Av. da República-EAN, Apartado 127, 2780-157 Oeiras, Portugal
    Search for more papers by this author
  • Helena Santos

    Corresponding author
    1. Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Av. da República-EAN, Apartado 127, 2780-157 Oeiras, Portugal
    • Instituto de Tecnologia Química e Biológica, Universidade Nova de Lisboa, Av. da República-EAN, Apartado 127, 2780-157 Oeiras, Portugal
    Search for more papers by this author


Mannosylglycerate is a compatible solute typical of thermophilic marine microorganisms that has a remarkable ability to protect proteins from thermal denaturation. This ionic solute appears to be a universal stabilizing agent, but the extent of protection depends on the specific protein examined. To understand how mannosylglycerate confers protection, we have been studying its influence on the internal motions of a hyperstable staphylococcal nuclease (SNase). Previously, we found a correlation between the magnitude of protein stabilization and the restriction of fast backbone motions. We now report the effect of mannosylglycerate on the fast motions of side-chains and on the slower unfolding motions of the protein. Side-chain motions were assessed by 13CH3 relaxation measurements and model-free analysis while slower unfolding motions were probed by H/D exchange measurements at increasing concentrations of urea. Side-chain motions were little affected by the presence of different concentrations of mannosylglycerate or even by the presence of urea (0.25M), and show no correlation with changes in the thermodynamic stability of SNase. Native hydrogen exchange experiments showed that, contrary to reports on other stabilizing solutes, mannosylglycerate restricts local motions in addition to the global motions of the protein. The protein unfolding/folding pathway remained undisturbed in the presence of mannosylglycerate but the solute showed a specific effect on the local motions of β-sheet residues. This work reinforces the link between solute-induced stabilization and restriction of protein motions at different timescales, and shows that the solute preferentially affects specific structural elements of SNase.


Cell membranes are permeable to water; hence, the ability to cope with changes in water activity is a prerequisite for cell survival. Most cells are prepared to counteract an increase in the external osmotic pressure by accumulating small molecules designated osmolytes. Organisms adapted to hypersaline environments accumulate high levels of osmolytes such as glycerol or KCl. Even at molar concentrations, these compounds do not disturb the physiological functions of macromolecules, thereby earning the name of “compatible solutes”.1, 2 Interestingly, the role of compatible solutes goes beyond that of balancing the osmotic gradient across the cell membrane, and some of these compounds are involved in the response to other types of stress, such as heat, oxidative, and acid stress.3 Conversely, common osmolytes, such as trehalose or ectoine, are able to confer increased stability on proteins under a range of environmental insults.

Organic osmolytes, and other compatible solutes, fall into few categories of chemical compounds, namely amino acids and derivatives, sugars and derivatives, polyols, and betaines. Trehalose, glycerol, glycine-betaine and ectoine are solutes typical of mesophiles, that is, organisms that grow optimally at moderate temperatures. However, marine organisms that thrive at high temperatures accumulate specific solutes which usually bear a negative charge at physiological pH and are rarely found as part of stress adaptation in mesophilic organisms. Appropriately, these compounds appear to be better than uncharged osmolytes at stabilizing proteins against thermal denaturation.4

For several years, we have studied compatible solutes of organisms adapted to hot environments (hyper/thermophiles). In particular, we are interested in the mechanisms underlying protein stabilization by mannosylglycerate (MG), a solute widespread in hyperthermophiles that is remarkably efficient in the stabilization of protein structures. For example, the melting temperature of SNase has an increment of 8°C in the presence of 0.5M mannosylglycerate.5

Understanding the molecular mechanisms that govern protein stabilization by osmolytes and other compatible solutes became a subject of great interest, not only because of the intrinsic biotechnological significance of increasing the performance of enzymes and other proteins under operational conditions, but also because of the medical relevance of finding effective suppressors of protein misfolding and aggregation. Since the pioneering studies by Nozaki and Tanford,6, 7 many researchers have made important contributions.8–11 There is now a much better understanding of the main thermodynamic features associated with protein folding/unfolding in the presence of uncharged osmolytes, but the effects on the structure and dynamics of the native and the denatured forms remain largely unexplored. In particular, further studies are necessary to clarify the relationship between protein dynamics and stabilization by compatible solutes. Previously, the effect of different solutes on a wide range of backbone motions was monitored by using several NMR methods. Small torsional fluctuations (ps-ns time scale) were studied by spin-relaxation measurements, chemical exchange on the millisecond scale was assessed by magnetization transfer experiments, and events in the second-to-minute time frame were probed with H/D exchange experiments.12 A strong correlation was established between the subnanosecond backbone motions (described by the generalized order parameter) and changes in the melting temperature (Tm) of the protein induced by different solutes. The effect of MG was also evident on the slowest time scale, providing site-specific information about the thermodynamic stability of the protein. However, motions on the millisecond time scale were little affected. Thus, there was strong evidence for a link between restriction of fast protein backbone motions and protein stabilization.

In the light of these results, we consider it important to address the following questions. Is the dynamic behavior of protein side-chains on the ns-ps timescale also important for the mechanism of protein stabilization by compatible solutes? Is the effect of stabilizing solutes on the slow timescale motions specific for the global motions of the protein, as previously suggested?13 And do stabilizing solutes affect the protein folding/unfolding pathway?

To study the effect of solutes on the dynamics of protein side-chains, we measured 13C relaxation rates of the methyl groups of a staphylococcal nuclease (SNase) variant (P117G/H124L/S128A), providing 46 probes of motion on the subnanosecond time scale. Although they are typically located in the close packed interior of proteins, methyl bearing side-chains usually display significant motions in addition to the rotation of the methyl group. Analysis of 13C relaxation data using the model-free formalism14, 15 is a common procedure to extract information about the motion of the carbon–carbon bond that connects the methyl group to the side-chain, and the rotation of the C[BOND]H vectors about the symmetry axis. The methyl groups are expected to report on the effects of different solutes on the dynamics of the protein hydrophobic core in the subnanosecond time scale.

The thermodynamic equilibrium of proteins implies unfolding and refolding of the molecules, even under conditions that favor the folded form. This creates transient non-native states that represent a minute fraction of the protein population under normal conditions and allows the protein folding/unfolding pathway to be studied using NMR H/D exchange experiments. Most backbone NHs exchange with the solvent only when the protein molecule is in one of the non-native states. In the so-called EX2 exchange regime, proton exchange rates depend on the equilibrium constant between the native and non-native states.16 The motions involved in visiting these states may be local motions that expose little new surface area, or nearly global motions that expose large patches of normally buried peptide segments.16 Stabilizing solutes reduce the exchange rates of NHs,12, 13, 17 which implies a shift in the equilibrium of the protein population towards the native state. The observation of a greater effect on the NHs that exchange more slowly has been explained as a consequence of solutes opposing large increases in protein surface area, as the slower exchanging protons are expected to exchange only through large scale unfolding motions. However, not all slower-exchanging NHs require large protein motions; hence, it is not easy to establish a relationship between the effect of stabilizing solutes and exchange rates. Investigating the scale of the motions associated with NH exchange may clarify the matter and deal with our second query: is the effect of stabilizing solutes really specific for the global motions of the protein?

The dependence of individual H/D exchange rates on denaturant (urea) concentration can be used to differentiate between local and global backbone motions. The slope of this dependence, called the m-value, relates to the increase in surface exposure that occurs when the protein visits the higher energy exchange-competent state while remaining under largely native conditions (folded population around 99%).18, 19 Values close to zero are characteristic of local motions while large m-values indicate that NH exchange requires nearly global motions.20, 21 Englander and coworkers found that protein assembly proceeds via cooperative folding of protein segments which they called foldons,20 with H/D exchange measurements used to show which amides are involved in each foldon. Experiments of this type in the absence and presence of MG should address our last question regarding the possible effect of stabilizing solutes on the cooperative unfolding motions of the protein.

By addressing these questions, this work shows that stabilization of a variant of SNase by MG has little effect on side-chain dynamics but restricts local fluctuations of NHs as well as global opening motions. Moreover, there is an intriguing insight into the mechanism of stabilization in the observation of specific effects of MG on the local motions of the β-sheets in SNase.


H/D, hydrogen/deuterium; MG, mannosylglycerate; SNase, staphylococcal nuclease; Tm, melting temperature


Assignment and solution structure of the PHS variant of SNase

We have identified 143 out of 149 possible amide resonances in the 1H-15N HSQC spectrum of SNase (Fig. S1 in the Supporting Information). Residues A1, T2, K6, K53, Y54, and E57 could not be observed in the 1H-15N HSQC; however, with the exception of A1 and K53, the side-chains of these residues have been identified in all the 13C HSQC-based spectra. All the missing amide resonances originate from amino-acids located in regions of high flexibility, which is to be expected since proton exchange is otherwise slow at the low pH (5.2) at which the NMR data were collected.

Overall, 84% of all possible resonances were assigned, including 93% of backbone resonances, 90% of the aliphatic 1H and 13C and 45% of all aromatic protons. The chemical shifts of a small number of groups are far removed from the average values seen in the BMRB database. These are K9 (HE2/HE3), D77, and M98. The uncommon values of HE of K9 are likely caused by its proximity to the aromatic ring of Y92.

The solution structure of this hyperstable variant of SNase is very similar to previously published crystal structures. The RMSD between the average NMR structure and the crystal structure (PDB entry 1EY8) is 0.96 Å. The main differences are in the loop spanning residues 42–52 (Supporting Information Fig. S2). This region exhibits the lowest order parameters,12 indicating that the loop is flexible. As can be seen in Figure 1, the region is not well defined in the NMR structure, since very few meaningful NOEs can be extracted for it. Further differences are observed in the 83–86 range, encompassing the loop connecting sheets 2 and 3. This part of the structure is well defined and the reason for this structural discrepancy is not immediately evident.

Figure 1.

Ribbon representation of the 20 best calculated structures of a SNase variant (P117G, H124L, S128A) calculated at 37°C. Figure prepared with MolMol.56 The N-terminal is at the bottom of the picture and the C-terminal is at the top. The less-structured loop spanning residues 42–52 is indicated in the figure as the “large” loop. [Color figure can be viewed in the online issue, which is available at]

The assigned chemical shifts have been deposited in the BioMagResBank with the accession number 18013, and the structures were deposited at RCSB with the PDB ID 2lkv (Supporting Information Table S1 shows the statistics on the structure determination).

Relaxation data and diffusion tensor

The relaxation rates of 46 13CH3 resonances from 30 amino acid side-chains were measured. Signal to noise ratios were generally above 30 for the peaks in the first time point spectra; slightly lower sensitivities were obtained for the NOE experiment without 1H saturation. Longitudinal relaxation decay curves were well fit with single exponentials. Most of the transverse relaxation decay curves were best represented by a biexponential function22 as assessed by F-test analysis. Uncertainties in the optimized parameters of R1 and R2 are generally below 10% while the standard deviation of duplicate NOE experiments was typically no greater than 5%.

The solution structure of the SNase variant determined by NMR spectroscopy was used as the structural model for the model-free calculations, with coordinates adjusted to the principal axes of the diffusion tensor using the program Tensor2.23, 24 Additionally, the coordinates of the methyl protons were averaged along the symmetry axis (the axis described by the C[BOND]C bond connecting the methyl group to the side-chain).

Sequential fitting of the 13C relaxation data to the five model-free models was performed using the RELAX software.25, 26. The complete RELAX script is provided as Supporting Information (Fig. S3).

Internal mobility

In the absence of solutes, the majority of the methyl groups (42 out of the 46) required the so-called extended model of the Lipari-Szabo formalism (S2s, S2f, and τs) to explain the relaxation data. For the remaining groups, the less complex models two (S2 and Rex) and four (S2, τe, and Rex) were sufficient to describe the data. In some cases, S2 values were found to depend on the model-free model used to fit the relaxation data. Therefore, we consider only the methyl group datasets that were fitted with the same model in all experimental conditions: 36 methyl groups fitted with model 5 (S2s, S2f, and τs). The average τs for these methyls is below 0.15 ns and the slowest value is just under 0.25 ns, whereas τc values determined for this protein are larger than 6 ns.

The dynamic behavior of side-chains, as represented by differences in the generalized order parameter of methyl groups (Supporting Information Fig. S4), seems largely unaffected by the presence of stabilizing or destabilizing solutes. Most of the changes observed in the S2 values are within the uncertainty limits estimated by Monte Carlo simulations. The significant variation observed for Ile72δ and Leu124δ coincides with methyl groups displaying S2 values above the theoretical limit of 0.111 for perfect tetrahedral geometry of the methyl group. More modest S2 variations occur for Ile18γ, Val66γ2, Ile92δ, and Val104γ2 but with no apparent correlation with the concentration of MG (not shown). In the case of the two valines, similar S2 variations are observed in the presence of glycerol (0.60M) and KCl (0.25M), two nonstabilizing compounds used as controls of viscosity and ionic strength, respectively. The location of these side-chains in the protein structure is not associated with a particular region of the protein or with the degree of solvent exposure.

The average S2 value of the side-chains clearly does not respond to changes in protein stability in the same way as the backbone dynamics determined in earlier work12 (Fig. 2).

Figure 2.

Comparison of the average S2 values of the backbone NHs and methyl groups as a function of the SNase melting temperature (Tm). The left vertical axis shows the S2 values of backbone (full circles) and the right axis shows the S2 values of the side-chains bearing methyl groups (empty circles). The Tm of SNase is shown for the protein in the presence of a given solute: no solute (NS), 0.25M KCl (KCl), 0.60M glycerol (Gly), 0.25M urea (Ur), and MG at 0.15M, 0.25M, and 0.35M. The error bars represent the error of the average calculated from the individual S2 uncertainties.

Analysis of hydrogen exchange

The exchange reaction of an NH occurs only in open conformations (NHopen) in which 1H is accessible to the solvent, as in Eq. (1).16, 27

equation image(1)

In the steady state, the experimentally observed exchange rate (kex) for any given NH is described by Eq. (2).

equation image(2)

In the bimolecular regime (EX2) the closed state (NHclosed) is highly favored. The closing rate (kcl) is much larger than the opening rate (kop) and also much larger than the exchange rate of a freely accessible amide proton (kch). The EX2 regime is typical of stable folded proteins and has been confirmed to hold for all measured exchange rates of a P117G, H124L SNase mutant.28 With these approximations, an equilibrium constant of opening can be defined and the free-energy of the structural opening reaction (ΔGHX) can be derived from the observed exchange rate, (kex) [Eqs. (3) and (4)].

equation image(3)
equation image(4)

The values for kch can be calculated as described in Bai et al.29 so that the free energies may be obtained from the measured exchange rates.

Amide exchange may result from local fluctuations, involving one or few H-bond breaks, or from so called global opening motions of the protein that involve the cooperative breaking of a large number of hydrogen bonds. Englander and colleagues20 also considered the possibility of “subglobal” opening motions associated with the cooperative unfolding of protein subunits—the foldons. However, throughout this report the term “global” will be used to describe all opening motions that do not result from local fluctuations. These exchange modes make different contributions to the free energy of exchange20 which may, in principle, be distinguished by studying the dependence of ΔGHX on the concentration of denaturant. Local motions expose little new surface area of the protein to the solvent, making them relatively insensitive to the denaturant concentration, whereas global opening motions increase the protein surface area significantly and expose additional denaturant binding sites. As a result, global motions become dominant at higher concentrations of denaturant.

Two possible models were considered to describe the dependence of ΔGHX with added denaturant (urea). In these, exchange occurs either through global (g) opening motions alone as in Eq. (5), or both local (lc) and global motions contribute to exchange, as in Eq. (6).20, 21

equation image(5)
equation image(6)

Here, ΔGHX(g) and ΔGHX(lc) are the free energies at zero denaturant concentration of the global and local opening reactions, respectively, and mi is the dependence of the ΔGHX on the denaturant concentration (i = 1 for local motions and i = 2 for global motions). These m-values correlate with the additional denaturant binding sites of the protein in the transient open state.18, 19, 30, 31 An F-test (α = 0.10) was used to establish whether the use of the more complex model was statistically significant, that is, if more than one opening mode contributes to the exchange reaction.

Dependence of H/D exchange rates on urea concentration

The signal to noise ratio of the amide resonances in the first HSQC spectra was 20–30, allowing accurate determination of peak volumes through most of the decay time and good precision in the fitted exchange rate values. The decay of the resonances observed at 35°C was generally well described by a mono-exponential function and the profile was not changed by the presence of urea or MG.

The exchange rates of 68 backbone NHs were determined, accounting for about 50% of the total. The presence of MG made the decay of some amide resonances too slow to be accurately determined within the experimental time (ca. 60 h) but these became measurable as the concentration of urea increased.

The urea unfolding curve of SNase determined by circular dichroism (Supporting Information Fig. S5) shows that 99% of the protein population is still folded at 2M urea, the highest concentration used in the exchange rate measurements without MG, hence the EX2 exchange regime should still hold under these conditions. Therefore, Eq. (4) was used to extract ΔGHX values for the amide protons. The variation of ΔGHX values with the concentration of urea shows four types of profile, as illustrated in Figure 3. Profile 1 is characterized by a two parameter model [Eq. (5)], m2 and ΔGHX(g), while Profiles 2 and 3 are characterized by a four parameter model [Eq. (6)]: m1 and m2, and ΔGHX(lc) and ΔGHX(g). Profile 3 differs from Profile 2 because its m1 value is essentially zero. The less common Profile 4, illustrated by the nearly flat line of Arg87 is a limiting case of Profile 2 (Supporting Information Table S2 lists the values of the model parameters determined for each NH).

Figure 3.

Typical profiles of the dependence of ΔGHX on urea concentration. The points are derived from amide H/D exchange experiments with SNase and the lines are fitted using Eqs. (5) and (6).

The values of ΔGHX increase significantly in the presence of MG. Nearly all NHs are stabilized in the presence of 0.33M mannosylglycerate (Supporting Information Fig. S6), with a maximum increase of 2.3 kcal/mol for Y93. There is a weak correlation between the ΔGHX of a given amide proton at zero denaturant concentration and ΔΔGHX due to the presence of MG (R2 = 0.43). There is no significant correlation between the extent of the global motion (m2-value) and the value ΔΔGHX(global) (Fig. 4).

Figure 4.

Free-energy variation of the global opening motions as function of the m2-value. The change in ΔGHX(g) of the global opening motions induced by 0.33M mannosylglycerate is plotted as a function of the respective m2-value. ΔΔGHX(g) is obtained by subtracting the free-energy values obtained by fitting the dependence of amide proton exchange on urea concentration in the presence of mannosylglycerate from those obtained in its absence.

Despite the very different ΔGHX values exhibited by the NHs of SNase, many eventually merge into common isotherms as the concentration of urea increases. This effect is illustrated by residues 129, 135, and 137 in Figure 3. Common ΔGHX isotherms suggest cooperative unfolding and are characterized by similar m2-values, ΔGHX(g) values, and structural relatedness. The experimental data available to fit the parameters m2 and ΔGHX(g) are sometimes sparse, in which case small errors in the H/D exchange measurements produce significant variations in the values of the fitted parameters. Therefore, observing the whole trend of the ΔGHX curve, instead of relying solely on the numerical values, seems more informative. In this way, we identify 45 amide protons as likely to be involved in ten different ΔGHX isotherms (Fig. 5). The presence of MG does not appear to cause any significant change in the compositions of the groups, though groups four and six appear to merge (Supporting Information Fig. S7), which seems to result from a greater increase of free-energy values in group six relative to group four.

Figure 5.

Partially unfolded units of SNase in the absence and presence of mannosylglycerate (0.33M), characterized by merging ΔGHX(g) isotherms. The ΔGHX(g) isotherms reveal cooperative unfolding reactions of groups of amide protons promoted by urea in the absence and presence of mannosylglycerate (MG). The lines in the plot represent the best fit of the measured exchange rates using either Eq. (5) or (6). Each color identifies amide protons involved in a particular partially unfolded unit: (1) purple for 36 and 37; (2) red for 97, 100, 102, 106, 107; (3) blue for 105, 128, 132, 134; (4) black for 24, 26, 30, 35, and 90–94; (5) light blue for 99, 101, 103, 104, 109, 110, 126, 130; (6) dashed line for 22, 23, 25, 27; (7) gray for 62 and 66, (8) green for 39, 108, 136, 139; (9) orange for 73, 74, 75; 10) 111, 129, 135, 137. The vertical axis has the same units in both plots. [Color figure can be viewed in the online issue, which is available at]

Large m-values tend to be associated with the more slowly exchanging amide protons, that is, those with higher ΔGHX. However, 30% of the amides with ΔGHX values greater than 5 kcal mol−1 display Profiles 2 or 3 with low initial m-values, precluding a general conclusion. The presence of MG has only marginal effects on the m-values and, in general, does not change the type of profile exhibited by a given amide proton. However, 6 amide protons (25, 67, 92, 99, 106, and 108) clearly shift from a two parameter model [Eq. (5)] to a four parameter model [Eq. (6)]. Two of these amide protons are involved in an H-bond pair (Ile92 and Val99), two share the same H-bond acceptor (Gln106 and Leu108), and they are all located close to the hinge regions of the protein.


This study complements our previous work on the effect of mannosylglycerate (MG), a stabilizing solute accumulated by many (hyper)thermophilic organisms, on the backbone motions of a hyperstable SNase variant. In that study, we found a correlation between restriction of backbone dynamics and protein stabilization by MG. The analysis is now expanded to include fast side-chain motions, as determined by 13C spin relaxation, and slower opening motions, as determined by H/D exchange measurements.

Fast dynamics of methyl bearing side-chains from 13C relaxation measurements

Model-free analysis of the 13C relaxation data obtained in the presence of the different solutes showed that model 5 (S2s, S2f, and τs) provided the best fit for the majority of the experimental data. This suggests a high degree of mobility of the protein side-chains and reorientation of the symmetry axis. The variation in the S2 values between different fitted models has been noted previously and attributed to deviations from the extreme narrowing limit ((ωτe)2 ≪ 1), in particular when τe approaches 1/10 of τc.32

In contrast to the significant restriction of the backbone dynamics in SNase, there is no generalized restriction of side-chain fast dynamics as a function of MG stabilization (Fig. 2). In other words, the side-chains remain essentially unperturbed in the presence of MG. Since there is a theoretical correlation between changes in S2 values and changes in conformational entropy33–35 these results suggest that MG reduces the entropy of the folded backbone and not of the side-chains. A decrease in unfolding entropy was proposed to be the origin of SNase A stabilization by MG below 65°C in a study based on differential scanning calorimetry and picosecond time-resolved fluorescence spectroscopy.5 These observations imply that the reduced entropy of the backbone is more than offset by other motions.

Local and global opening motions in amide proton exchange

We followed the model of protein H/D exchange proposed by Linderstrøm-Lang and developed by others16, 27, 36 in which exchange occurs by opening motions of different extent that expose the hydrogen to the solvent. Because the amount of denaturant used here does not significantly change the population of folded protein (Supporting Information Fig. S5), we are monitoring thinly populated unfolded states under normal conditions.

The majority of the NHs monitored show evidence of exchange via local motions (i.e., m-value near zero, as in Profile 3), which contrasts with the less stable wild type SNase A in which nearly all NHs (61 out of 64) exchange via global opening motions37 (i.e., large m-values, as in Profile 1). Amide protons with ΔGHX values that do not change significantly (m-value < 0.1) throughout the range of urea concentrations tested here (Profile 4) may indicate residual protection in the unfolded state of the protein.20 Gly88 may be an example of this as it was found in a native-like loop conformation in the SNase mutant Δ131Δ, which is considered to be a good model of the unfolded structure of the wild type SNase.38 Alternatively, the greatest urea concentration used here may not be sufficient for global motions to become dominant.

It is important to note that the interpretation of m-values is not completely consensual.39 Wooll et al. have shown that it is possible, at least in theory, for denaturant independent exchange of many amide protons to be consistent with significant surface area exposure.39 Nevertheless, in practice, the m-value of a number of proteins has been shown to correlate strongly with changes in solvent accessible surface area.31

Effect of MG on the ΔGHX urea dependence profiles

Based on m2-values, ΔGHX(global), and structural relatedness we defined 10 groups of amide protons with merging unfolding isotherms and, therefore, possible cooperative unfolding. For some residues, the spread of the different global ΔGHX isotherms is not sufficient to allow a definite assignment to one group or another. Nevertheless, it is clear that the relative position of the several ΔGHX isotherms remains unchanged in the presence of MG (Fig. 5). The m-values are also little affected by the presence of solute, which was also noted for trimethylamine N-oxide, another stabilizing solute.40 These observations suggest that, despite the substantial increase in stability conferred by MG, the protein unfolding/folding pathway remains the same. Previous work from our group reached a similar conclusion but using different biophysical methods.5 A recent study on a similar SNase mutant (P117G/H124L) points to the existence of at least four cooperative unfolding structural units (foldons) for this protein.28 Considering that the solute has little effect on the m-values and on the merging ΔGHX(g) isotherms in the hyperstable SNase variant studied here, it is likely that the protein foldons are very similar in the absence and presence of the stabilizing solute.

Specific effects of MG on local motions

Many of the ΔGHX curves obtained in the presence of MG (0.33M) are virtually superimposable on those obtained in the absence of solute, with a shift of the ΔGHX curve along the axis of urea concentration by about 1.2M (see Fig. 6 for examples). This overlap suggests that 0.33M mannosylglycerate is canceling the effect of 1.2M urea. The superposition of data points from lower concentrations of MG (star symbols in Fig. 6) indicates that the cancellation occurs for all solute concentrations in the ratio of about 3.5 parts of urea to 1 of MG.

Figure 6.

Comparison of the dependence of free energies of amide H/D exchange on urea concentration, determined at 35°C in the absence and presence of MG (0.33M). The data obtained without MG is shifted along the concentration axis to show the similarity of the profiles. Empty symbols represent data obtained in the presence of 0.33M mannosylglycerate and are plotted in relation to the axes above the panels; filled symbols represent data obtained without MG and are plotted in relation to the axes below the panels. The lines represent the best fit using Eqs. (5) and (6). Star symbols represent selected ΔGHX values obtained in the presence of 0.15 and 0.24M mannosylglycerate and without urea; their positions on the concentration axis are shifted in proportion to the MG concentration.

The presence of mannosylglycerate (0.33M) causes six NHs to shift their ΔGHX urea dependence profile from Type 1 to Type 2 or 3 (e.g., residue 108 Fig. 6). This is a simple consequence of the general increase in ΔGHX(g) values, such that local motions can be seen to contribute to the exchange. However, the quality of cancellation of effects in 3.5:1 urea/mannosylglycerate mixtures is variable. Some amide protons show marked deviations, with increased ΔGHX(local) values in the presence of MG, such as residue 39 in Figure 6. This suggests that MG may have other effects besides counteracting the urea. Furthermore, this observation does not fit the notion that stabilizers specifically restrict global motions.

To analyze the distribution of changes in ΔGHX, we subtracted, for each NH, the ΔGHX value obtained with 0.33M mannosylglycerate and 1.2M urea from the ΔGHX value obtained without solutes, and plotted them as a function of position in the structure (Fig. 7). The width of the Cα trace of SNase in Figure 7 has been set proportional to this difference and clearly shows that the largest effects are associated with β-sheets. Thus, it appears that MG has a specific effect in restricting the local motions of NHs in β-sheets in SNase. Alternatively, this effect may be associated with the β-barrel subdomain of the protein. This region was shown to hold residual structure in non-native states of SNase.41, 42 Mutations in this region were found to have a greater effect on the structure of the unfolded state than on the native states.43–45 The similarity of m-values determined in the absence and presence of MG suggest no such differential structural impact. However, the existence of a late-unfolding intermediate rich in the β-barrel subdomain could explain the greater effect of the solute on the local motions of this region. An argument against this hypothesis comes from the work of Englander's group on foldons of another SNase mutant,28 where the outer half of the β-barrel is expected to unfold early in the denaturation process.

Figure 7.

Distribution of deviations from cancellation of the effects of urea and MG on amide H/D exchange. The Cα trace of SNase is shown with radii proportional to the differences between ΔGHX values obtained in the presence of 1.2M urea and 0.33M mannosylglycerate and in the absence of solutes. Residues with NHs that exchange partly through local motions are colored in red, those exchanging only via global motions are in blue, and residues without data are in grey. The figure was created with MolMol.56 [Color figure can be viewed in the online issue, which is available at]

Further examples would be necessary to confirm that this is a stabilizing effect associated with a specific type of secondary structure or subdomain rather than the response of one part of the SNase fold to an increase in effective surface tension across the whole protein. However, this effect goes some way to explain the observation that the extent of stabilization depends on the specific solute/protein pair examined.

In summary, our studies on the slower time scale of H/D exchange measurements in the presence of urea revealed that mannosylglycerate does not alter the folding/unfolding pathway of SNase, and restricts the global unfolding motions of the protein to a similar extent throughout the protein. The local motions, however, are restricted to varying extents, with some NHs experiencing no change while others, particularly those in β-sheets, have changes in ΔGHX that are comparable with those for global motions. This work underlines the importance of studying the widest possible range of motions in exploring the link between protein dynamics and protein stabilization by compatible solutes. Furthermore, it suggests that MG restricts local motions in specific structural elements, which may be an important clue to explain the variability in the degree of stabilization conferred on different proteins.

Materials and Methods

Preparation of protein samples

The protein used for these studies is a hyperstable variant of SNase which differs from the wild type amino acid sequence in three positions: P117G, H124L, S128A.46 The expression and purification protocols for SNase samples were as described elsewhere,12, 47 except for the medium composition, which differs in the nitrogen and carbon sources. For the uniformly 13C/15N labeled SNase sample, [13C6]-glucose (4 g/L) and 15NH4Cl (2 g/L) were used as sole carbon and nitrogen sources. Fractionally 13C labeled SNase was prepared using a mixture of 15% [2-13C]acetate, 15% [1-13C]acetate, and 70% [1,2-12C]acetate at a final concentration of 4 g/L as sole carbon source, and 14NH4Cl (2 g/L) as nitrogen source. This labeling strategy results in a protein molecule where carbon sites have a 15% probability of being labeled with 13C. Finally, uniformly 15N labeled SNase samples were obtained using 15NH4Cl (2 g/L) as sole nitrogen source, while unlabeled SNase was obtained using LB rich medium. The concentration of purified protein stocks was estimated by the solution absorbance at 280 nm and the extinction coefficient of 0.93 (cm−1 mg−1 ml).5 The protein purity was assessed by SDS-PAGE electrophoresis with Coomassie staining. Purified protein stocks were flash-frozen with liquid nitrogen in small drops, as this reduces protein aggregation, and stored at −80°C.

Circular dichroism spectroscopy

Circular dichroism (CD) experiments were performed at 35°C on a Jasco J-815 spectropolarimeter equipped with a Peltier thermostated cell. Spectra of 15 μM SNase samples (acetate buffer, 60 mM, pH 5.5) were recorded in the range of 200–250 nm using a 0.1 mm path length cell and the average of 7 scans. Chemical unfolding of SNase was studied by monitoring the CD signal intensity at 222 nm in the presence of several urea concentrations (0–6 M). Nonlinear least-squares fitting of the data was carried out using the Santoro-Bolen equation.48

NMR spectroscopy

All NMR experiments were performed at the ITQB magnetic resonance center, CERMAX. Spectra for the assignment of SNase resonances and for 13C relaxation measurements were collected on a Bruker AvanceIII 800 spectrometer (Bruker, Rheinstetten, Germany) operating at 800.33 MHz, equipped with a TXI-Z H C/N/-D (5 mm) probe. Experiments for H/D exchange measurements were performed on a Bruker AVANCEIII 500 spectrometer (Bruker, Rheinstetten, Germany), operating at 500.13 MHz, using a 5 mm QXI inverse detection probe-head with pulsed field gradients along the Z axis.

Data were processed with Topspin 2.1 (Bruker Biospin) and analyzed with CARA version 1.8.4. Backbone sequential assignments were obtained from the analysis of 1H-15N HSQC and triple resonance HNCA, HNCO, HN(CO)CA, CBCA(CO)NH, and HNCACB experiments. Side-chain resonances were assigned with the combined use of 15N HSQC-TOCSY, 15N HSQC-NOESY, HBHA(CO)NH, (H)CCH-TOCSY, and 1H-13C HSQC-NOESY spectra. The 1H-13C HSQC-NOESY and (HB)CB(CGCD)HD spectra were used to assign aromatic side-chains.

Structure determination

Initial models of the structure were calculated with the program UNIO′10,49, 50 using the 1H-1H NOESY, 3D 1H-15N HSQC-NOESY, 3D 1H-13C HSQC-NOESY spectra and chemical shift lists tailored to each spectrum. For each calculation cycle, 100 structures were produced using CYANA2.1 protocols,51 with 10,000 simulated annealing steps. In this procedure, the residue ranges of 6–42 and 56−145 were used for the calculation of the RMSD. After the first cycle, H-bond constraints were introduced in accordance with H/D exchange data and the observed secondary structure. After 7 iterative calculation cycles, bundles of 20 structures were generated, together with upper limit constraint lists. These constraint lists were then used for de novo calculations of structures in explicit water using the RECOORD protocols.52 First, 80 structures were calculated using the included CNS protocols.52 The 20 best of these structures were then used for explicit water calculations based on the smallest number of NOE violations and the ten structures with the lowest calculated energy were used to form the final NMR bundle. The quality of the calculated structures was evaluated using the suit of programs iCING (

13C relaxation measurements

13C relaxation rates were measured for the methyl groups of 1 mM SNase samples fractionally labeled with 13C (60 mM acetate-d4 buffer, pH 5.1), at 37°C in the absence and presence of the following compounds: mannosylglycerate (0.15, 0.25, and 0.35M), KCl (0.25M), glycerol (0.60M), and urea (0.25M). The concentrations of KCl and glycerol are chosen to match the ionic strength and viscosity of 0.25M mannosylglycerate, respectively. The pulse sequences used to measure 13C longitudinal (R1), and transverse (R2) relaxation rates and steady-state heteronuclear {1H-13C}NOE (h-NOE) were based on published schemes22 with the following modifications: gradient pulses (sine shape) were included and the last carbon pulse was a 180° adiabatic shaped pulse (Crp60,0.5,20.1) of 500 μs. Spectra for R1 and R2 relaxation measurements were acquired as pseudo 3D data while spectra for h-NOE determination were acquired in an interleaved fashion. The R2 pulse sequence includes a CPMG heat compensation block so that all time points are acquired in similar conditions. These pulse sequences were designed to minimize cross relaxation effects and to minimize the effects of differential 1H relaxation.22 Perdeuteration of the protein has been proposed as a means to suppress remote proton relaxation,53 but this chemical modification was not considered here since solute/protein interactions are likely to be affected.

Longitudinal relaxation rates (R1) were measured using relaxation periods of 10, 50, 100, 250, 350, 600, 800, and 1,200 ms while transverse relaxation rates (R2) were obtained using relaxation periods of 14, 28, 42, 56, 70, 83, 97, and 111 ms. Peak volumes were determined using the integration mode of TopSpin 2.1 software (Bruker) and their time dependence fitted to a mono exponential (for R1) or biexponential function (R2) by nonlinear least squares. The transverse 13C relaxation of an isolated methyl group rapidly rotating about its 3-fold axis and attached to a macromolecule is better characterized by a biexponential decay of the form y = 0.5A[exp(-tR2f) + exp(−tR2s)] due to the inner and outer 13C multiplet components.22 In this equation, R2f and R2s represent the fast and slow components of the transverse relaxation rate; the value used for the subsequent model-free analysis is the average of the relaxation rates, R2 = (R2s + R2f)/2. Error estimates were obtained from the standard deviation of the fitted coefficients. Experiments for h-NOE measurements were performed in duplicate and determined from the ratio of the peak volume of each resonance in the presence and absence of 1H saturation (Vsat/Vnonsat). The NOE was allowed to build for 5 s. For the control experiment (no NOE) a minimum relaxation delay of 7 s was used.

Model-free analysis of 13C relaxation data

Model-free analysis of 13C relaxation data made use of the diffusion tensor parameters obtained from 15N relaxation measurements performed previously under the same experimental conditions;12 since 15N relaxation measurements provided a larger data set, it is expected that the diffusion tensor values are more precise than could be obtained from the 13C relaxation data. Specific, fully anisotropic, diffusion tensor values were used for each experimental condition: no solute, KCl, glycerol, urea, and MG at 0.15, 0.25, and 0.35M. Relaxation data were analyzed with the RELAX software package25, 26 and the Lipari-Szabo formalism assuming an internuclear 1H-13C distance of 1.07 Å and negligible effect of chemical shift anisotropy.14, 15 The dynamics of 13C labeled methyl groups were characterized by one of five models: (1) S2; (2) S2, and τe; (3) S2 and Rex; (4) S2, τe, and Rex; (5) S2s, S2f, and τs. For each methyl group, the best model was selected based on the lowest Akaike's information criteria. The fifth and most complex model considers two contributions, a fast motion (S2f) corresponding to methyl group rotation about its symmetry axis (the carbon–carbon bond connecting the methyl group to the side-chain), and a slower motion (S2s) with a correlation time, τs, corresponding to the reorientation of the symmetry axis.54, 55 The simpler models consider a fast motion described by the generalized order parameter, S2, and assess the presence of chemical exchange (Rex) and effective correlation time (τe).

H/D exchange experiments

Samples used in H/D exchange experiments contained 1 mM of 15N labeled SNase with 0.15, 0.24, 0.33M MG or no MG, and urea at several concentrations. The reaction of H/D exchange was initiated by dissolving lyophilized protein samples in deuterated solutions (acetate-d4 buffer, 60 mM, pD 5.2) with increasing concentrations of urea. The sample was thermally equilibrated at 35°C for 10 min in the spectrometer before collecting sequential 1H-15N HSQC spectra. The first spectrum started less than 15 min after the exchange reaction was initiated. Spectra were acquired with 1,024 complex data points, 136 time increments, and 4 scans per increment and each took less than 15 min.

Cross-peak volumes of the NH resonances present in the HSQC spectra were measured with TopSpin 2.1 software (Bruker) and fitted to a mono-exponential decay function using the Solver package of Microsoft Excel, providing the exchange rate for each detected NH resonance.


The authors thank Prof. Bertrand Garcia-Moreno for providing the plasmid of the hyperstable SNase and Dr. Luis Fonseca for valuable help with the statistical analysis.