Salt-specific stability and denaturation of a short salt-bridge forming \alpha-helix

Salt-specific stability and denaturation of a short salt-bridge forming -helix

Joachim Dzubiella Physics Department T37, Technical University Munich, 85748 Garching, Germany. E-mail: jdzubiel@ph.tum.de
Abstract

The structure of a single alanine-based Ace-AEAAAKEAAAKA-Nme peptide in explicit aqueous electrolyte solutions (NaCl, KCl, NaI, and KF) at large salt concentrations (3-4 M) is investigated using s molecular dynamics (MD) computer simulations. The peptide displays -helical structure without salt and destabilizes with the addition of NaCl in agreement with experiments of a somewhat longer version. It is mainly stabilized by direct and indirect (“”)EK salt bridges between the Lys and Glu side chains and a concomitant backbone shielding mechanism. NaI is found to be a stronger denaturant than NaCl, while the potassium salts hardly show influence. Investigation of the molecular structures reveals that consistent with recent experiments Na has a much stronger affinity to side chain carboxylates and backbone carbonyls than K, thereby weakening salt bridges and secondary structure hydrogen bonds. At the same time the large I has a considerable affinity to the nonpolar alanine in line with recent observations of a large propensity of I to adsorb to simple hydrophobes, and thereby ’assists’ Na in its destabilizing action. In the denatured states of the peptide novel long-lived ( ns) ’loop’-configurations are observed in which single Na-ions and water molecules are hydrogen-bonded to multiple backbone carbonyls. In an attempt to analyze the denaturation behavior within the preferential interaction formalism, we find indeed that for the strongest denaturant, NaI, the protein is least hydrated. Additionally, a possible indication for protein denaturation might be a preferential solvation of the first solvation shell of the peptide backbone by the destabilizing cosolute (sodium). The mechanisms found in this work may be of general importance to understand salt effects on protein secondary structure stability and further systematic experiments in this matter are strongly encouraged.

thanks: To whom correspondence should be addressed. E-mail: jdzubiel@ph.tum.de

I Introduction

Identifying and understanding the mechanisms by which salt stabilizes or destabilizes protein structures is a highly complex issue as, in general, they result from a subtle balance of specific ion-solute and ion-water interactions Baldwin (1996); Tobias and Hemminger (2008). Even the seemingly simpler question how ions interact with planar nonpolar surfaces and thereby influence their macroscopic interfacial properties is still a matter of ongoing research Jungwirth and Tobias (2006); Kunz (2006); Horinek and Netz (2007). It was found for instance that large anions have a surprisingly strong propensity to adsorb to nonpolar surfaces in contrast to the classical continuum picture, where ions are always repelled from low dielectric spaces. While for relatively homogeneous solutes (nonpolar atoms, simple colloids, etc.) salt effects can often be at least qualitatively interpreted by rather nonspecific mechanisms such as electrostatic screening, changes in solvent surface tension Baldwin (1996), or dielectric properties Barthel et al. (1998); Vila et al. (2000), the analysis for mixed polar and nonpolar entities on a molecular level, as given by proteins, is far more complicated. It has been shown recently that even the simple cations sodium and potassium, which are vertical neighbors in the periodic table, exhibit strikingly different behavior in the interaction with protein surfaces, where sodium is strongly favored over potassium Vrbka et al. (2006); Uejio et al. (2008). One apparent reason is the strong attraction of sodium with protein carboxylates or backbone carbonyls in accord with the ’law of matching water affinities’, relating ion hydration energies to their respective tendency to form ionic pairs Collins (1997, 2004). Similar specifc binding behavior can also be observed for the strong ionic denaturant Guanidinium Mason et al. (2007). It is, in fact, the distribution of water and ions close to the protein surface, i.e., their preferential interaction to (or exclusion from) the latter that intimately relates the change in their chemical potential and that of the macromolecule. This useful view has more and more sharpened in the recent years Timasheff (1993); Parsegian et al. (2000); Timasheff (2002); Shimizu and Smith (2004); Shulgin and Ruckenstein (2005); Smith (2006) and has been successfully used to analyze salt-specific hydrophobic association Ghosh et al. (2005); Thomas and Elcock (2007).

In particular for large concentrations ( M) salt effects are pronounced, highly sequence and salt-type specific, and typically lead to changes in protein solubility, stability, and/or denaturation that result in the so-called Hofmeister series for the precipitation of proteins Baldwin (1996). Although an order of magnitude higher than at typical physiological conditions ( M), large salt concentrations play an important biochemical role in the broad field of protein crystallization Dumetz et al. (2007), in food industry as fermentation additives Dyer (1951), and for the function and stability of biotechnologically interesting halophilic (salt-loving) enzymes Lanyi (1974). Additionally the study of protein structures in high salinity solvents is instructive as salt-specific effects are amplified and, important from a computational perspective, can be sampled more efficiently in molecular dynamics (MD) simulations, a typical tool nowadays for the theoretical study of protein folding, function, and stability Karplus and McCammon (2002).

An ubiquitous and fundamental secondary structure element of proteins is the -helix which is stabilized by backbone hydrogen bonds typically involving four amino acids (aa) per turn. The majority of short ( 10-20 aa) isolated helices derived from proteins are unstable in solution, unless specific side-chain interactions stabilize them. Among these it has been demonstrated that alanine based peptides have the strongest intrinsic helix propensity Marqusee and Baldwin (1987); Marqusee et al. (1989); Spek et al. (1999); Chakrabartty et al. (1994); Scholtz et al. (1991). Theoretical investigations of the properties of short helices are vast, either with focus on stability and structure Sorin and Pande (2005); Garcia and Sanbonmatsu (2002); Vila et al. (2000); Levy et al. (2001); Gnanakaran and Garcia (2005); Fedorov et al. (2007) or folding kinetics Hummer et al. (2000); Monticelli et al. (2005); Wang et al. (2007). The action of salt however, has been studied only very recently by MD simulations of the trialanine peptides in sodium halide solutions Fedorov et al. (2007). Strongly specific effects on peptide compactness (helicity is ill defined for such a short peptide) and different interface affinities of the ions have been reported in line with the arguments before.

A very instructive model peptide was experimentally introduced by Marqusee and Baldwin Marqusee and Baldwin (1987) who showed with circular dichroism (CD) and other methods that short (16-17 aa) alanine-based peptides display high helix propensity when implanted with Glu and Lys residues that possibly form stable salt bridges at pH = 7 (loosely defined, a salt bridge is a weak ionic bond between positively and negatively charged amino acid side chains). A sensitive dependence of helicity on the spacing of the charged groups was detected, i.e., largest helix probability ( helicity) resulted if glutamic acid (E) and lysine (K) were spaced 4 residues apart as in “EK()” = Ac-AEAAAKEAAAKEAAAKA-NH, while the reversed order with three spacings, “KE()” = Ac-AKAAEAKAAEAKAAEA-NH, led to the lowest helicity (). It was argued that the strong sensitivity may have its origin in the interaction of the charged groups with the electrostatic helix dipole and the larger steric and conformational freedom of the side chains in the “EK()” case allowing for a higher salt bridge forming potential. A quantification and molecular details of these hypotheses however, remained elusive. Importantly, it was shown that addition of a simple monovalent salt like NaCl can monotonically destabilize the “EK()” helix with increasing concentration ( at 3 M) at pH = 7, which could not be interpreted by screening only Marqusee and Baldwin (1987).

Further work on other peptides revealed that the stabilizing potential of salt bridges strongly depends on sequence context and suggested that they can contribute significantly to thermal stability, obviously an appealing issue for protein engineers Makhatadze et al. (2003); Olson et al. (2001). In particular, in a recent MD study of the thermal stability of a 20 aa long “EK()” peptide it was indeed found that salt bridges contribute to thermal stability but not significantly Ghosh et al. (2003). Another, possibly more important stabilizing mechanism was identified, where the relatively long Lys side chain shielded specific backbone hydrogen bonds from the direct interaction with water. This screening mechanism by long side chains has been observed also for other peptides to act stabilizing on the helical structure Vila et al. (2000); Garcia and Sanbonmatsu (2002).

In this work we employ standard all-atom MD simulations to investigate the structure and (helix) stability of a shorter version of the salt-bridge forming “EK()” peptide with 12 aa and sequence Ace - AEAAAKEAAAKA-Nme. We systematically explore the (de)stabilizing action of four different salts (NaCl, NaI, KCl, KF) at large concentrations ( M) and analyze the molecular and structural details in relatively long s runs. Particular focus is given to peptide structure, role of salt bridges and backbone shielding, and the specific influence of solvent and ions thereon. An attempt is made to interpret peptide (de)stabilization by a preferential hydration/exclusion mechanism of solvent and ions.

Ii Methods and systems

MD simulations are performed using the parallel module sander.MPI in the simulation package Amber9.0 with the ff03 force-field for the peptides and the rigid and nonpolarizable TIP3P water for the solvent Case (2006). Parallel computing on 8-16 processors is executed on the supercomputer HLRBII of the Leibniz-Rechenzentrum München (LRZ). All simulated systems are maintained at a fixed pressure of bar and a temperature by coupling to a Berendsen barostat and Langevin thermostat Case (2006), respectively. The cubic and periodically repeated simulation box has edge lengths  Å  including approximately 1500 water molecules. Electrostatic interactions are calculated by particle mesh Ewald summation and all real-space interactions (electrostatic and van der Waals) have a cut-off of 9 Å. The peptide is generated using the tleap tool in the Amber package Case (2006).

Cations and anions in our study are modeled as nonpolarizable Lennard-Jones spheres with charge and interaction parameters as supplied by Dang Dang (1995) and summarized in Tab. I. The considered concentrations result from ion pairs in the simulation box. We investigate the structure and helix stability of the 12 aa long peptide with the acetyl (Ace) and amine (Nme) capped sequence Ace-AEAAAKEAAAKA-Nme forming three -helical turns in the fully folded state, where Glu2 and Lys6, and Glu7 and Lys11, are potentially able to form a salt bridge, respectively. We simulate at a temperature  K, relevant for comparison to experimental trends for the somewhat longer (17 aa) analog in Marqusee and Baldwin (1987) in NaCl. The influence on helix stability of large concentrations ( M) of the simple monovalent salts NaCl, KCl, NaI, and KF is investigated. We simulate each system for s requiring a total simulation time of s. All simulation snapshots are visualized using VMD Humphrey et al. (1996). Trajectory analysis is done using the ptraj tool in the Amber package Case (2006), where, in particular, secondary structure elements such as the helicity (i.e., helix percentage per peptide configuration) are identified using the DSSP method by Kabsch and Sander Kabsch and Sander (1983). In our MD we start with initially unfolded peptide configurations. Even the most stable helices feature typical folding times of at least 25-50 ns before the first folding event, an order of magnitude slower than in implicit solvent calculations of the 17 aa long “EK()”peptide Wang et al. (2007). We simulate further ns before gathering statistics.

We attempt to examine and interpret some of our data in the framework of preferential interactions Timasheff (1993). Briefly, the excess chemical potentials of a ternary mixture of solute (macromolecule) , cosolute (salt) , and water are intimately connected by Gibbs-Duhem-like relationships: and . The experimentally accessible preferential interaction parameters are defined by

(1)

where is the ratio of the number of cosolutes and water molecules in the bulk, and is the ratio of the number of cosolutes and water molecules in a local domain around the macromolecule. Thus, the excess chemical potential of the macromolecule’s configuration is directly related to the distribution of cosolutes in the solute’s vicinity. If the cosolutes are excluded (preferential hydration) from the macromolecule’s local domain, then . If cosolutes bind to the macromolecule, then (preferential dehydration). It is observed empirically that a strong preferential hydration (cosolute exclusion) typically stabilizes protein structures although it can not be used as a stability criterion, i.e., does not necessarily imply the stabilization of the native structure Arakawa et al. (1990). Whether the preferential interaction framework can be used as a criterion for protein stability is therefore questionable. The microscopic reason is fine balance between exclusion of the cosolute from and its specific binding to the protein surface in dependence of the chemical nature of the latter Arakawa et al. (1990); Schellman (2003). These favorable binding events are typically considered responsible for protein denaturation but it is in general unclear where the particular binding sites are or who the binding partner is. The detailed molecular origins that define are thus not well understood.

Iii Results and discussion

First of all, a comment should be made on the reliability of classical molecular force-fields used in MD simulations. We employ ion parameters that were introduced by Dang Dang (1995) as the default Amber parameters, e.g., Na and K from Åqvist Aqvist (1990) mixed with Cl from Dang, are known to be faulty at larger salt concentrations. They show a too strong ion pairing propensity and premature freezing at concentrations considerably lower than saturation Auffinger et al. (2007); Joung and Cheatham III (2008). We have tested the Dang parameters for a wide range of densities in the homogeneous case and always found a reasonable fluid structure as can be judged by looking at the radial distribution functions Auffinger et al. (2007). Their dependence on the particular water model has been found to be weak Auffinger et al. (2007); Patra and Karttunen (2004). However, macroscopic bulk properties such as the osmotic pressure for these force fields at large concentrations however, have not been systematically investigated to the best of our knowledge. The necessity of the explicit consideration of the polarizability of ions is still a matter of debate Jungwirth and Tobias (2006); Kunz (2006). Regarding the water model we use TIP3P parameters as it is the default water model in Amber and hopefully reasonably balanced with the solute force-fields, although it is known that other models (such as SPC/E) seem to perform better in describing water bulk properties Becker et al. (2001). Also intrapeptide potentials are typically not free from error when benchmarked to experiments or other more reliable calculations, see, e.g. references Sorin and Pande (2005); Hornak et al. (2006); Gnanakaran and Garcia (2005). Therefore, we do not claim to be quantitative in our work but focus on the discussion of effects and qualitative trends.

iii.1 Helix stability

Examples for peptide helicity vs. simulation time are plotted in Fig. 1 for a) the salt-free case, b) 3.7 M NaCl, and c) 3.5 M KCl. Without salt the helicity can be considerable () but two large-scale unfolding transitions are observed at and 900 ns, where the peptide is basically unfolded (helicity 50%) for 50 ns. The total helicity averaged over the whole trajectory is found to be , not far from the experimental value of about of the 17 aa long analog Marqusee and Baldwin (1987). Adding NaCl at a large concentration clearly leads to destabilization of the helical structure as shown in Fig. 1 b). States with more than helicity are rarely sampled and stable only on 50-200 ns time periods. The average helicity goes down to 37% for 3.7 M. The destabilizing trend is in agreement with experimental measurements of the longer analog (17 aa), where a continuously decreasing helicity with increasing NaCl concentrations was found with a value of 57% at M Marqusee and Baldwin (1987). In contrast to NaCl, we find only a slight destabilizing action when replacing sodium by the 30% bigger potassium ion (cf. Tab. I) as shown in Fig. 1 c). The average helicity of the system of 63% is close to the salt-free case even at a large concentration of M KCl. The average helicity of all investigated systems is summarized in Tab. II.

The helicity resolved by individual peptide residues for the same systems is plotted in Fig. 2. The helicity distribution for the salt-free case and 3.7 M KCl is highest in the center of the peptide and diminishes to the terminal ends as found before in studies of a similar peptide involving salt bridges Ghosh et al. (2003). Note that these ’finite-size’ or end-effects lead to a somewhat smaller total helicity in our MD study when compared to the experimental, longer analog with 17 aa. Further inspections of the data reveals that the helicity along the peptide is asymmetric with respect to the terminal ends, i.e., there is a distinct jump in helicity between the Glu2 and Ala3 residues. The reason is the preferable interaction of the carboxylate group of Glu2 with the N-terminus which was also observed in an implicit solvent study of the same sequence pattern Wang et al. (2007). It is known experimentally that ’N-capping’ by specific side chains strongly influences the helix distribution along a given peptide Vasquez et al. (1987); Serrano and Fersht (1989); Chakrabartty et al. (1993). Although negatively charged residues close to the N-terminus have been found to stabilize the helix locally by a preferential charge - helix dipole interaction we observe that Glu2 can be dragged to the N-terminus enabling Lys6 to interact with the Ala1 or Glu2 carbonyls, thereby destabilizing the Glu2-Lys6 salt-bridge and their backbone hydrogen bonds (leading to a local destabilization of the helix). We will return to this issue in the next section where the role of salt bridge formation is discussed in more detail. With 3.7 M NaCl the helicity per residue clearly diminishes for each residue as is also illustrated in Fig. 2.

For a better understanding of the equilibrium structure distribution of the investigated peptide it is instructive to calculate the probability by which a configuration root mean square deviates (rmsd) from the fully helical reference structure. The latter was chosen randomly from configurations with 100% helicity. We found that the following results did only slightly () depend on its particular choice. The according free energy vs. the rmsd reaction coordinate without salt, NaCl, and KCl is shown in Fig. 3. At zero salt concentration and with 3.7 M KCl there are two distinct minima at which the peptide is in the fully helical state with three helical turns [rmsd  Å, cf. Fig. 3 a)], or features two neighboring helical turns [rmsd  Å, cf. Fig. 3 b)]. The free energy barrier along our chosen reaction coordinate between these two states is about . In the salt-free case larger rmsd appear only with a much higher energy penalty, while for KCl also two small local minima are present at  Å  and 6.5 Å. The situation is qualitatively different with 3.7 M NaCl. The minimum corresponding to the full helical state disappears and a broad minimum occurs at larger rmsd values  Å. Detailed inspection of our MD data reveals that this broad minimum is occupied by mainly two distinct configurations, which can obviously not be resolved with our simple rmsd reaction coordinate. At a rmsd  Å  we observe long-lived configurations that feature two helical turns at the both ends of the peptide with an intermediate loop, see the snapshot in Fig. 3 c), while the rmsd  Å  mainly represents the coil state where helicity completely vanishes, see Fig. 3 d). The “loop”-states [Fig. 3 c)] in NaCl are long-lived and stabilized by specific sodium and water binding to the peptide backbone, an intriguing observation we will discuss later in more detail.

In order to test the influence of changing the anion we replaced NaCl by NaI and KCl by KF. We observed only little change in the results above for the potassium salts, but a noticeably larger destabilizing action for the iodide salt. In fact, NaI is known to be a strong denaturant Baldwin (1996). It seems however, that for the particular peptide and ions studied the difference in the cation charge density is the crucial parameter which determines the denaturation capability of the salt. Although comparable in size and often observed to be similar in their salting-in and salting-out actions on proteins Baldwin (1996); Nandi and Robinson (1972a, b), qualitatively different influence on the solubility of amino acids Koshkbarchi and Vera (1997), on the activity and stability of halophilic proteins Madern et al. (2000), and a striking difference in binding to protein carboxylates and carbonyls Vrbka et al. (2006); Uejio et al. (2008) have been observed experimentally. Unfortunately, no experimental data on KCl, KF, or NaI action on the “EK()” peptide for a more meaningful and verifying/falsifying comparison is available yet.

iii.2 Salt bridge formation and backbone shielding

The charged residues Glu2 and Lys6, and residues Glu7 and Lys11 are potentially able to form EK-type salt bridges, which may directly stabilize the helix by shifting the equilibrium to the helical, more compact state Marqusee and Baldwin (1987) and, additionally, may contribute to helix stability through backbone desolvation and shielding from water Vila et al. (2000); Garcia and Sanbonmatsu (2002); Ghosh et al. (2003). To elucidate these mechanisms in more detail we calculate the (normalized) probability distribution , where is the distance between the carboxylate carbon atoms on Glu and the amine nitrogen on Lys for the first (Glu2-Lys6) and second (Glu7-Lys11) salt bridge, averaged over the whole MD trajectory. Examples for the salt-free case and large NaCl and KCl concentrations are shown in Fig. 4 for the Glu7-Lys11 pair. The first (contact) peak for distances  Å  indicates a direct salt-bridge while the second peak  Å corresponds to an indirect – by one water molecule separated – salt bridge. Illustrating MD snapshots are shown at the top of Fig. 4. With those distribution at hand we estimate the probability , , of finding a direct or both, direct and indirect salt bridges in the course of our simulation by integrating up to or , respectively. In the salt-free case only by of the time a direct salt bridge is formed by one of the Glu-Lys pairs. This value is in accord with the results of Ghosh et al. Ghosh et al. (2003) on the similar “EK()” peptide who found that possible EK salt bridges spend only in contact configuration and thus probably do not play an immediate role in helix stabilization. Both, direct and indirect salt bridges however, have a higher probability 27% for the Glu2-Lys6 pair and in for the Glu7-Lys11 pair. As discussed previously the asymmetric stability behavior between the two salt bridges is the consequence of the interaction of Glu2 with the N-terminal. The probability values for forming direct and both, direct and indirect salt bridges for all investigated systems are summarized in Tab. II.

The distribution function depends on salt type and concentration as is also illustrated in Fig. 4. For NaCl the salt bridge formation probability is decreased by more than 50%. Obviously this strongly correlates with the average helicity in the system as in a non-helical state the charged side chains are too far apart to come into a contact state. For the large KCl concentration the distribution changes much less when compared to NaCl. For further analysis, in Fig. 5 we plot the radial distribution function (rdf) of water and cations (anions) around the Glu (Lys) headgroup, i.e., the Glu carboxylate carbon (Lys side chain nitrogen). Strikingly, and observable from the contact peak, there is strong affinity of sodium to the carboxylates over potassium as has been found in previous experimental studies and quantum mechanical calculations Vrbka et al. (2006); Uejio et al. (2008). The rdf of potassium on the other hand is comparable in magnitude to that of the water oxygen; to quantify, we estimate the coordination number of species in the first carboxylate solvation shell with

(2)

where Å  is chosen to be the extension of the first solvation shell and is the number density of species . While in the salt-free case 8-9 water molecules directly solvate the carboxylate, of them are replaced on average by sodium ions in the NaCl solution, indicating a strong specific binding of sodium. Note also the decreased contact value of the water-carboxylate rdf in the NaCl solution in Fig. 5. In KCl we find on average only ions which replace water molecules, suggesting that potassium is much weaker in breaking direct or indirect salt bridges. In contrast, sodium directly competes with water and Lys-nitrogen atoms in binding to the carboxylates and has a strong impact on salt bridge formation. The distributions of water and anions around the Lys nitrogen is plotted in the inset to Fig. 5: F Cl I is the preferred sequence of interaction as could have been expected from electrostatic considerations, i.e., there is a stronger attraction with larger ion charge density. Calculating coordinations numbers as above however, reveal that also only on average fluoride ions are able to replace a water molecule in the water solvation shell, still much less effective in replacing water than sodium around the carboxylates.

Another secondary-structure stabilizing effect has been proposed to be the local or nonlocal shielding of backbone hydrogen bonds from water by large side chains Vila et al. (2000); Garcia and Sanbonmatsu (2002); Ghosh et al. (2003). For an analysis we follow Ghosh et al. Ghosh et al. (2003) and plot in the inset to Fig. 4 the average number of water molecules in the first solvation shell of the backbone carbonyls resolved by amino acid residue. In the systems with large average helicity (salt-free and KCl) the backbone carbonyls 2 or 3, 6, 7 are distinctively less solvated (1.1-1.3 water molecules) than the others. These carbonyls indeed belong to or are in immediate vicinity of the two Glu and the central Lys side chains. By salt-bridging along the folded peptide, exactly their carbonyls are screened from the surrounding water and the hydrogen bond stability of the latter is very likely to be increased. Snapshots are shown at the top of Fig. 4, illustrating the screening effect of the Glu7-Lys11 direct and indirect salt bridges. From the inset to Fig. 4, we clearly see that adding salt leads to a larger hydration of the backbone, probably due to the induced unfolding and its exposure to the solvent. Interestingly, even for the large NaCl concentration salt bridge formation is not entirely suppressed, see Tab. II, and backbone carbonyl 7 remains least solvated.

An interesting conclusion from this analysis is that direct and indirect salt bridges – while they may not play a dominant immediate role – stabilize the helix by efficiently screening some of their backbone hydrogen bonds. The strong affinity of sodium over potassium to the carboxylates makes the former a powerful competitor to water in the binding to the Glu side chains, thereby decreasing the probability of forming direct or indirect salt bridges.

iii.3 Preferential hydration and ion binding

In order to check whether similar specific binding events as found in III.B. can also be detected for other parts of the peptide, we have calculated the rdfs between ions and the backbone amide oxygens and nitrogen atoms, averaged over the whole trajectories. Examples for the cation and anions are plotted in Fig. 6 a) and b), respectively. Analogous to the results in III.B. a strong attraction of sodium to the carbonyl oxygens is observed in striking contrast to a weak affinity of potassium. The anions however, interact with the backbone even weaker, cf. Fig. 6 b). Interestingly, iodide shows here the highest first peak compared to the other anions (I Cl F) conversely to its interaction with the Lys headgroup, see the inset of Fig. 5, where (F Cl I). Related to this, the strong affinity of sodium to the backbone is enhanced in the presence of iodide. We argue that these effects may have two possible origins: firstly, the bulk activity of NaI is larger than that of NaCl at the same concentration Robinson and Stokes (2002), in other words, transferring a sodium-iodide pair from a NaI solution to a reference solution costs less energy than from NaCl. Secondly, it was shown that the relatively large anions have an affinity to nonpolar surfaces Jungwirth and Tobias (2006); Kunz (2006); Horinek and Netz (2007) or hydrophobic solutes Kalra et al. (2001). To inspect this hypothesis for our (partly nonpolar) peptide, we plot the rdf between the carbon atom in the alanine side chain and cations or anions in the insets to Fig. 6 a) and b), respectively. Indeed we find that Iodide has the strongest affinity to the nonpolar side chains from all considered ions, giving rise to a relatively high affinity to the protein surface. These findings agree with MD of a toy model of a heterogeneous biomolecule, where it has been argued and demonstrated that fluoride and iodide binding strongly depends on the local surface characteristics Lund et al. (2008). They also support the perspective that the iodide propensity to nonpolar surfaces may indeed impact protein stability Jungwirth and Tobias (2006); Kunz (2006), a mechanism also proposed for other large ions such as Guanidium Mason et al. (2007).

In the following we attempt to examine and interpret this data in the framework of preferential interactions (see Methods). To look at the ionic distribution in detail we separately calculate the preferential interaction parameter for every ionic species , so that . We define the local peptide domain using an appropriate cut-off around every atom for counting water and salt particles. We find that our results do qualitatively not depend on the choice of , for which we tested values between 4 and 8 Å. In Fig. 7 we plot the results for  Å  – which roughly corresponds to the water first solvation shell (Figs. 5 and 6) – versus peptide helicity. We make following observations: firstly, for all ions, showing preferential hydration of the peptide in all configurations and for all salts. Secondly, in agreement with the preferential interaction picture the destabilizing salts show the largest affinity to the peptide, indeed in the order of their destabilizing action. Thirdly, the dependence of on the helicity is surprisingly weak given the broad conformational changes of the peptide and shows hardly structure. For NaI and NaCl, increases by 30% and 6%, respectively, when going from the helical to the coil state, while for the other salts the change is less than 5%. Our data gives thereby microscopic support of the statement that preferential hydration can not be used as a criterion for protein stability Arakawa et al. (1990). Overall, the large iodide has a surprisingly strong affinity to the protein, obviously due to the more attractive interaction with the nonpolar Ala side chains and backbone nitrogen atoms when compared to the other anions or potassium, see the discussion of Fig. 6 above.

It seems that the denaturation of the helical structure by NaCl and NaI can be attributed to the favorable interaction of sodium with the backbone carbonyls, amplified by weak anions such as iodide. Therefore it is instructive to calculate the number of ions and the preferential interaction parameter just for the first water solvation shell of the protein backbone which we define as , evaluated for every ionic species. From the the first minimum of the backbone-water rdfs in Fig. 6 we estimate the extension of the first solvation shell to be  Å, which we now take to define the local domain around the backbone. Results for the number of ions in the first backbone solvation shell are plotted in Fig. 8: while less than one potassium or anionic particle can be found on average close to the backbone, typically more than one sodium interacts with it, strongly depending on helicity. In the nonhelical state approximately 2 and 2.5 sodium ions are on average bound to the backbone for NaCl and NaI, respectively. If these numbers are translated into , see the inset of Fig. 8, we observe that changes sign for the nonhelical states for a helicity and for NaCl and NaI, respectively. Potassium (shown) and anions (not shown) do not exhibit this behavior. If is calculated including the second solvation shell (Å) a change of sign for sodium is not observed either. On the basis of this data one could speculate that preferential solvation only of the first backbone solvation shell , (1st solv. shell), by the destabilizing species (here sodium) may be a criterion for protein denaturation.

Finally, inspection of MD trajectories for NaCl and NaI reveal that the strong interaction of sodium with the backbone carbonyls can result in intriguingly long-lived protein configurations where sodium is bound and actively involved in the protein structure, as shown in Fig. 9. The central part of the peptide loops around a single sodium ion, thereby binding it with 3-4 backbone oxygens, while still a partly helical structure can be maintained, cf. Fig. 9 a). Sometimes an additional water molecule is captured by the backbone-ion complex and binds, as illustrated in Fig. 9 b) and c). These states are surprisingly stable on a long 10-20 ns time scale. We have not observed such long-lived states involving potassium or anions.

Iv Concluding remarks

In summary, by using all-atom MD simulations we have provided molecular insights into the structural stability of a short salt-bridge forming peptide. We find that specific structural mechanisms such as salt bridges and side chain shielding can stabilize a helical structure in accord with experiments. Although direct salt bridges are found only rarely, the action of indirect (by one water molecule separated) salt bridges must not be overlooked as they may contribute significantly to the shielding of local backbone hydrogen bonds. These stabilizing mechanisms are observed to be suppressed by the specific binding of sodium to carboxylates and backbone carbonyls, thereby shifting the equilibrium from helical to coil states. This specific binding was not observed for the larger potassium ion.

A surprisingly large affinity to the peptide for iodide is observed however, in particular to the nonpolar, hydrophobic parts of the protein, and as previously observed in studies and experiments of large anions at planar hydrophobic interfaces Jungwirth and Tobias (2006); Kunz (2006); Horinek and Netz (2007), toy model proteins Lund et al. (2008), and simple hydrophobes Kalra et al. (2001). The reasons are probably a mix of a larger bulk activity for halide salts with larger anions Robinson and Stokes (2002) and a specific water-assisted affinity to nonpolar residues what renders iodide itself a rather hydrophobic entity. Bulk effects and the high peptide affinity of iodide leads to an increased number of sodium ions close to the peptide, implying that iodide directly and indirectly increases peptide dehydration and thus promotes helix destabilization. These findings support the perspective that the iodide propensity to nonpolar surfaces may impact protein stability Jungwirth and Tobias (2006); Kunz (2006). They also highlight the highly synergetic action of the strongly electrostatically coupled co- and counterions, which can be attracted to the protein surface for different reasons.

In an attempt to interpret our findings in the preferential interaction framework we find indeed that for the strongest denaturant, NaI, the peptide is least hydrated and is maximal. It remains negative however, showing a preferential hydration. Resolving vs. helicity shows weak structure, possibly due to canceling contributions from polar and nonpolar parts of the proteins and has to be more thoroughly investigated. However, if is calculated for the destabilizing species (sodium) in the first backbone solvation shell only, a strong slope and a change of sign is observed at intermediate helicities. This preferential solvation of the backbone [(1st solv. shell)] may serve as an criterion for protein denaturation. An experimental verification of this hypotheses is probably hardly feasible but it may provide food for thought and inspire further (computational) investigation on this matter as the microscopic origins of are subtle.

The mechanism found may be of general importance to understand cosolute effects on protein secondary structure stability and further experiments probing systematically the salt-specific action on the -helical stability of this or similar short peptide are strongly encouraged. As we have shown in this work, molecular insights from MD simulations can provide valuable information to understand the intricate mechanisms in solvent-protein interactions and thereby protein stability and folding. For this however, accurate MD force-fields, in particular for ions at moderate to large concentrations, are an essential prerequisite to avoid possible artifacts Auffinger et al. (2007); Joung and Cheatham III (2008), and need to be benchmarked to experiments. Finally, we note that the novel long-lived ’loop’-configurations in the denatured/unfolded states in which sodium and water is bound and immobilized by the peptide backbone may be experimentally accessible by nuclear magnetic relaxation dispersion methods (NMRD) Denisov et al. (1996) or time-resolved FRET measurement Möglich et al. (2006).

V Acknowledgements

J. D. is grateful to D. Horinek and I. Kalcher for useful discussions, the Deutsche Forschungsgemeinschaft (DFG) for support within the Emmy-Noether-Program, and the LRZ München for computing time on HLRBII.

References

  • Baldwin (1996) R. L. Baldwin, Biophys. J. 71, 2056 (1996).
  • Tobias and Hemminger (2008) D. J. Tobias and J. C. Hemminger, Science 319, 1197 (2008).
  • Jungwirth and Tobias (2006) P. Jungwirth and D. J. Tobias, Chemical Reviews 106, 1259 (2006).
  • Horinek and Netz (2007) D. Horinek and R. R. Netz, Phys. Rev. Lett. 99, 226104 (2007).
  • Kunz (2006) W. Kunz, Pure Appl. Chem. 78, 1611 (2006).
  • Barthel et al. (1998) J. M. G. Barthel, H. Krienke, and W. Kunz, Physical Chemistry of Electrolyte Solutions: Modern Aspects (Steinkopff, New York, 1998).
  • Vila et al. (2000) J. A. Vila, D. R. Ripoli, and H. A. Scheraga, Proc. Natl. Acad. Sci. 97, 13075 (2000).
  • Vrbka et al. (2006) L. Vrbka, J. Vondrasek, B. Jagoda-Cwiklik, R. Vácha, and P. Jungwirth, Proc. Natl. Acad. Sci. 103, 15440 (2006).
  • Uejio et al. (2008) J. S. Uejio, C. P. Schwartz, A. M. Duffin, W. S. Drisdell, R. C. Cohen, and R. J. Saykally, Proc. Natl. Acad. Sci. 105, 6809 (2008).
  • Collins (1997) K. D. Collins, Biophys. J. 65, 65 (1997).
  • Collins (2004) K. D. Collins, Methods 34, 300 (2004).
  • Mason et al. (2007) P. E. Mason, J. W. Brady, G. W. Neilson, and C. E. Dempsey, Biophys. J. 93, L04 (2007).
  • Timasheff (1993) S. N. Timasheff, Annu. Rev. Biophys. Biomol. Struct. 22, 67 (1993).
  • Parsegian et al. (2000) V. A. Parsegian, R. P. Rand, and D. C. Rau, Proc. Natl. Acad. Sci. 97, 3987 (2000).
  • Timasheff (2002) S. N. Timasheff, Proc. Natl. Acad. Sci. 99, 9721 (2002).
  • Shimizu and Smith (2004) S. Shimizu and D. J. Smith, J. Chem. Phys. 121, 1148 (2004).
  • Smith (2006) P. E. Smith, Biophys. J. 91, 849 (2006).
  • Shulgin and Ruckenstein (2005) I. L. Shulgin and E. Ruckenstein, Biophys. Chem. 118, 128 (2005).
  • Ghosh et al. (2005) T. Ghosh, A. Kalra, and S. Garde, J. Phys. Chem. B 109, 642 (2005).
  • Thomas and Elcock (2007) A. S. Thomas and A. H. Elcock, J. Am. Chem. Soc. 129, 14887 (2007).
  • Dumetz et al. (2007) A. C. Dumetz, A. M. Snellinger-O’Brien, E. W. Kaler, and A. M. Lenhoff, Protein Science 16, 1867 (2007).
  • Dyer (1951) W. J. Dyer, J. Food Science 16, 522 (1951).
  • Lanyi (1974) J. K. Lanyi, Bacteriological Rev. 38, 272 (1974).
  • Karplus and McCammon (2002) M. Karplus and J. A. McCammon, Nat. Struct. Mol. Biol. 9, 646 (2002).
  • Marqusee and Baldwin (1987) S. Marqusee and R. L. Baldwin, Proc. Natl. Acad. Sci. 84, 8898 (1987).
  • Marqusee et al. (1989) S. Marqusee, V. H. Robbins, and R. L. Baldwin, Proc. Natl. Acad. Sci. 86, 5286 (1989).
  • Spek et al. (1999) E. J. Spek, C. A. Olson, Z. Shi, and N. R. Kallenbach, J. Am. Chem. Soc. 121, 5571 (1999).
  • Chakrabartty et al. (1994) A. Chakrabartty, T. Kortemme, and R. L. Baldwin, Protein Sci. 3, 843 (1994).
  • Scholtz et al. (1991) J. M. Scholtz, E. J. York, J. M. Stewart, and R. L. Baldwin, J. Am. Chem. Soc. 113, 5102 (1991).
  • Sorin and Pande (2005) E. J. Sorin and V. S. Pande, Biophys. J. 88, 2472 (2005).
  • Garcia and Sanbonmatsu (2002) A. E. Garcia and K. Y. Sanbonmatsu, Proc. Natl. Acad. Sci. 99, 2782 (2002).
  • Levy et al. (2001) Y. Levy, J. Jortner, and O. M. Becker, Proc. Natl. Acad. Sci. 98, 2188 (2001).
  • Gnanakaran and Garcia (2005) S. Gnanakaran and A. E. Garcia, Proteins: Structure, Function, and Bioinformatics 59, 773 (2005).
  • Fedorov et al. (2007) M. V. Fedorov, J. M. Goodman, and S. Schumm, Phys. Chem. Chem. Phys. 9, 5423 (2007).
  • Hummer et al. (2000) G. Hummer, A. E. Garcia, and S. Garde, Phys. Rev. Lett. 85, 2637 (2000).
  • Monticelli et al. (2005) L. Monticelli, D. P. Tieleman, and G. Colombo, J. Phys. Chem. B 109, 20064 (2005).
  • Wang et al. (2007) W.-Z. Wang, T. Lin, and Y.-C. Sun, J. Phys. Chem. B 111, 3508 (2007).
  • Makhatadze et al. (2003) G. I. Makhatadze, V. V. Loladze, D. N. Ermolenko, X. F. Chen, and S. T. Thomas, J. Mol. Biol. 327, 1135 (2003).
  • Olson et al. (2001) C. A. Olson, E. J. Spek, Z. Shi, A. Vologodskii, and N. R. Kallenbach, Proteins: Structure, Function, and Genetics 44, 123 (2001).
  • Ghosh et al. (2003) T. Ghosh, S. Garde, and A. E. Garcia, Biophys. J. 85, 3187 (2003).
  • Case (2006) D. A. Case (2006), software AMBER9.0, University of California, San Francisco.
  • Dang (1995) L. X. Dang, J. Am. Chem. Soc 117, 6954 (1995).
  • Humphrey et al. (1996) W. Humphrey, A. Dalke, and K. Schulten, J. Molec. Graphics 14, 33 (1996).
  • Kabsch and Sander (1983) W. Kabsch and D. Sander, Biopolymers 22, 2577 (1983).
  • Arakawa et al. (1990) T. Arakawa, R. Bhat, and S. N. Timasheff, Biochemistry 29, 1924 (1990).
  • Schellman (2003) J. A. Schellman, Biophys. J. 85, 108 (2003).
  • Aqvist (1990) J. Aqvist, J. Phys. Chem. 94, 8021 (1990).
  • Auffinger et al. (2007) P. Auffinger, T. E. Cheatham III, and A. C. Vaiana, J. Chem. Theory Comput. 3, 1851 (2007).
  • Joung and Cheatham III (2008) I. S. Joung and T. E. Cheatham III, J. Phys. Chem. B 112, 9020 (2008).
  • Patra and Karttunen (2004) M. Patra and M. Karttunen, J. Comp. Chemistry 25, 678 (2004).
  • Becker et al. (2001) O. M. Becker, A. D. MacKerell Jr, B. Roux, and M. Watanabe, Computational Biochemistry and Biophysics (Routledge, USA, 2001).
  • Hornak et al. (2006) V. Hornak, R. Abel, A. Okur, B. Strockbine, A. Roitberg, and C. Simmerling, Proteins: Struc. 65, 712:725 (2006).
  • Vasquez et al. (1987) M. Vasquez, M. R. Pincus, and H. A. Scheraga, Biopolymers 26, 351 (1987).
  • Serrano and Fersht (1989) L. Serrano and A. R. Fersht, Nature 342, 296 (1989).
  • Chakrabartty et al. (1993) A. Chakrabartty, A. J. Doig, and R. L. Baldwin, Proc. Natl. Acad. Sci. 90, 11332 (1993).
  • Nandi and Robinson (1972a) P. K. Nandi and D. R. Robinson, J. Am. Chem. Soc. 94, 1299 (1972a).
  • Nandi and Robinson (1972b) P. K. Nandi and D. R. Robinson, J. Am. Chem. Soc. 94, 1308 (1972b).
  • Koshkbarchi and Vera (1997) M. K. Koshkbarchi and J. H. Vera, Ind. Eng. Chem. Res. 36, 2445 (1997).
  • Madern et al. (2000) D. Madern, C. Ebel, and G. Zaccai, Extremophiles 4, 91 (2000).
  • Robinson and Stokes (2002) R. A. Robinson and R. H. Stokes, Electrolyte Solutions (Dover publications, 2002), 2nd rev.
  • Kalra et al. (2001) A. Kalra, N. Tugcu, S. Cramer, and S. Garde., J. Phys. Chem. B 105, 6380 (2001).
  • Lund et al. (2008) M. Lund, R. Vácha, and P. Jungwirth, Langmuir 24, 3387 (2008).
  • Denisov et al. (1996) V. P. Denisov, J. Peters, H. D. Hörlein, and B. Halle, Nat. Struc. Biol. 3, 505 (1996).
  • Möglich et al. (2006) A. Möglich, K. Joder, and T. Kiefhaber, Proc. Natl. Acad. Sci. 103, 12394 (2006).
atom Å (kJ/mol) charge
Na 2.584 0.418 +1
K 3.332 0.418 +1
F 3.120 0.753 -1
Cl 4.401 0.418 -1
I 5.171 0.418 -1
O 3.151 0.636 -0.834
H 0 0 +0.417
Table 1: Atom-atom Lennard-Jones parameters and and charge used in our study for the ion  Dang (1995) and TIP3P water oxygen (O) and hydrogen (H) interactions.
salt /M helicity
0.0 0.71 0.10 0.27 0.11 0.43
NaCl 3.7 0.30 0.03 0.07 0.06 0.18
KCl 3.5 0.63 0.02 0.11 0.10 0.33
NaI 3.3 0.23 0.04 0.10 0.09 0.24
KF 4.0 0.58 0.03 0.16 0.09 0.31
Table 2: Helicity and salt bridge probabilities of the peptide for different types of salt at concentration . and correspond to the probability of finding the Glu2-Lys6 and Glu7-Lys11 head groups within a distance , where we define  Å  and  Å  to be direct and indirect salt bridges, respectively.
Figure 1: Peptide helicity vs. time for a) no salt, b) 3.7 M NaCl, and c) 3.5 KCl. Local time averaging over a 2 ns window of the raw MD data has been performed.
Figure 2: Average helix probability (helicity) of each residue in the peptide for no salt and large NaCl and KCl concentrations. Lines are guide to the eye.
Figure 3: Free energy along the rmsd (root mean square deviation from a ’perfect’ helix formed by this peptide) reaction coordinate. The MD snapshots of the peptide backbone at the bottom exemplify configurations with high probability (low free energy) with an rmsd location indicated by the arrows.
Figure 4: Probability distribution of distances between the carboxylate carbon atom on the Glu7 side chain and the amine nitrogen on the Lys11 side chain. The MD snapshots on the top exemplify direct and indirect (separated by one water molecule) salt-bridge configurations corresponding to the first and second peak of , respectively. Just side chains of amino acids 7-11 are shown. Note the shielding of the Glu7 backbone carbonyl hydrogen bonding to the Lys11 nitrogen (blue dashed line). Inset: Average number of water oxygens in the first solvation shell of the backbone carbonyl atoms vs. amino acid number (aa #). Lines are guide to the eye.
Figure 5: Radial distribution function (rdf) between the carboxylate carbon of the Glu side chains and X, where X = O (water oxygen), Na, or K, plotted for no salt () and NaCl and KCl at concentrations  M. Note that at (black line) and (green line) are nearly indistinguishable. Inset: rdf between the nitrogen of the Lys side chain and X, where X = O (water oxygen), F,Cl, or I.
Figure 6: a) Radial distribution function (rdf) between the backbone oxygen and X, where X = Na, K, or water oxygen (O) plotted for different salts, see legend. Inset: rdf between alanine side chains and X. b) Rdf between the backbone nitrogen and X, where X = F,Cl, I, or water oxygen (O) for different salts, see legend. Inset: rdf between alanine side chains and X.
Figure 7: Preferential interaction parameter for every ionic species in the first peptide solvation shell (averaged over every peptide atom) vs. peptide helicity.
Figure 8: Average number of cations (Na and K, filled symbols) and anions (I, Cl, and F, open symbols) in the first solvation shell of the backbone amides vs. helicity for different salts at concentrations 3-4M. Inset: preferential interaction parameter for Na and K only for the first backbone solvation shell vs. helicity.
Figure 9: MD simulations snapshots of long-lived (10-20 ns) peptide configurations observed in NaCl or NaI solutions at large concentrations, where a), one Na-ion (blue sphere) or, b), c), Na and one water molecule (red and white spheres) are permanently hydrogen bonded and immobilized by the peptide backbone. Only backbone atoms of the peptide are shown.
Figure 10: TOC figure.
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
""
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
   
Add comment
Cancel
Loading ...
192794
This is a comment super asjknd jkasnjk adsnkj
Upvote
Downvote
""
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters
Submit
Cancel

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test
Test description