|Home | About | Journals | Submit | Contact Us | Français|
Gadolinium (Gd3+) vibronic side band luminescence spectroscopy (GVSBLS) is used to probe, as a function of added Hofmeister series salts, changes in the OH stretching frequency derived from first shell waters of aqueous Gd3+ and of Gd3+ coordinated to three different types of molecules: i) a chelate (EDTA), ii) structured peptides (mSE3/SE2) of the lanthanide-binding tags (LBTs) family with a single high affinity binding site; and iii) a calcium binding protein (calmodulin) with four binding sites. The vibronic side band (VSB) corresponding to the OH stretching mode of waters coordinated to Gd3+, whose frequency is inversely correlated with the strength of the hydrogen bonding to neighboring waters, exhibits an increase in frequency when Gd3+ becomes coordinated to either EDTA, calmodulin or mSE3 peptide. In all of these cases, the addition of cation chloride or acetate salts to the solution increases the frequency of the vibronic band originating from the OH stretching mode of the coordinated waters in a cation and concentration-dependent fashion. The cation dependence of the frequency increase scales with charge density of the cations giving rise to an ordering consistent with the Hofmeister ordering. On the other hand water Raman shows no significant change upon addition of these salts. Additionally, it is shown that the cation effect is modulated by the specific anion used. The results indicate a mechanism of action for Hofmeister series ions in which hydrogen bonding among hydration shell waters is modulated by several factors. High charge density cations sequester waters in a configuration that precludes strong hydrogen bonding to neighboring waters. Under such conditions anion effects emerge as anions compete for hydrogen bonding sites with the remaining free waters on the surface of the hydration shell. The magnitude of the anion effect is both cation and Gd3+-binding site specific.
Proteins are not only a category of essential versatile biological macromolecule but also a class of polymeric molecules with significant biomedical and biotechnical potential. A required step needed in order both to understand protein function in vivo and to harness their potential application as biomaterials requires a molecular level understanding of how external and environmental factors influence the stability, dynamics and functionality of proteins. Osmolytes influence protein stability and reactivity both in vivo and in vitro. Hofmeister noted that cation and anion osmolytes can be ordered based on their effect on the solubility of proteins1. The so-called Hofmeister series ordering has been further extended to include the effect of ions on a wide range of solution properties2. Since its initial description there have been considerable interests and activities directed at exposing the underlying mechanism behind the Hofmeister series ordering of ion specific effects; nevertheless, the origin of this ordering remains uncertain and controversial3,4.
One of the proposed mechanisms for the Hofmeister ordering is through ion specific alterations in the hydrogen bonding network of water2,5,6. According to whether the ions are strongly hydrated or weakly hydrated, ions have been classified as either kosmotropes (structure makers) or chaotropes (structure breakers) respectively2,5–7. Though supported by indirect thermodynamic and macroscopic experiments2,6–16, the prevailing mechanism based on ion induced alterations in water interactions have been re-evaluated and challenged in recent studies3,4 that indicate a lack of direct effect of ions on bulk water17–22. Molecular simulation studies have provided contradictory descriptions based on chosen models and initial parameters22–24.
A possible reconciliation among these two different views and sets of results relating to the origin of the Hofmeister series might emerge from recent findings that ions accumulate and perturb local water structure at surface interfaces including protein/water interfaces. The degree of ion accumulation appears to scale with their ordering within the Hofmeister series3,25. It is in this critical interface between macromolecules such as proteins and the bulk solvent that one finds hydration shell waters. It is becoming ever more apparent that protein stability, dynamics and function are directly tied to hydration shell properties26–31. Thus, potentially, ions that alter the hydration layer surrounding a protein can impact protein properties without significant perturbation on water interactions in the bulk solvent regime.
The recent studies that have lead to a revision in thinking regarding the origin of the Hofmeister ordering have all focused on Hofmeister anions because of its large ionic radii and close distance to water molecules. In contrast, the present study focuses on the interplay between Hofmeister cations and anions and how this interplay is modulated by specific environments. The conceptual framework for analyzing the data is based on a simplistic model in which water structure is inhomogeneous on a nanoscopic level and water properties result from perturbations influencing the balance between two interconverting microstructural components of water: high-density clusters and low density clusters32–44. In this scheme, the difference in the properties between the two clusters arises largely from hydrogen bonding among waters within the high density clusters being weaker than in the low density clusters. It has also been proposed that distinctions between bulk and hydration water layers originate from a different fractional composition of two types of water clusters within these two solvent regimes36.
The effect of added salts on water interactions is probed using a novel spectroscopic technique that provides the vibrational frequencies of waters in the hydration shell of Gd3+. The Gd3+ cation is an especially suitable probe in that it readily replaces the similarly sized Ca2+ in chelates and calcium binding sites in proteins and peptides without disruption of the native structure45–48. The technique allows for probing of how a series of added Hofmeister salts influence hydrogen bonding among hydration waters both for the free Gd3+ in solution and for Gd3+ coordinated to a series of biologically relevant calcium binding sites.
The specificity and utility of this technique is based on earlier findings that Gd3+ vibronic side band luminescence spectroscopy (GVSBLS) can provide an infrared-like vibrational spectrum derived exclusively from molecules surrounding Gd3+, including those of coordinated first shell waters49–51. Changes in the vibrational frequencies from the first shell water molecule reflect changes in their hydrogen bonding to outer sphere solvent molecules52–54. As such, GVSBLS is potent molecular-level probe of hydrogen bonding within the hydration shell of Gd3+. GVSBLS is derived from the Gd3+ luminescence spectrum in which there are weak vibronic (vibration plus electronic) side bands (VSBs) associated with pure electronic transition from the lowest excited electronic state to the ground electronic state50,55 (Figure 1). These VSBs correspond to transitions which couple a change in both the electronic state of the Gd3+ and the vibrational state of nearby molecules. Figure 1a and 1c show that the ~273 nm (36630 cm-1) excited luminescence spectrum of Gd3+ in pure aqueous solution arises overwhelmingly from a single electronic transition, 6P7/2→ 6S7/2 at ~311 nm (32154 cm−1). The separation in frequencies between the electronic transition and that of the vibronic transitions yields the energy of the vibrational transition for molecular species in the immediate environment of Gd3+ (Figure 1b). GVSBLS technique has been successfully used to probe the Gd3+ associated ligands when the Gd3+ cation is coordinated to calcium binding proteins, phospholipids, and DNA51 as well as alcohols, crown ethers, carbonate, phosphate and several other assorted organic molecules56. The technique has also been applied to probe the waters within porous sol-gels and glasses57.
In the present study, changes in the O-H stretching frequencies from the first shell waters of the Gd3+ cation are monitored as a function of added salts. This frequency is inversely correlated with the strength of the hydrogen bonding to second shell waters. Thus, through VSBLS, the water coordinated to the Gd3+ cation becomes a probe of how added salts perturb hydrogen bonding between the “probe” waters on the surface of Gd3+ cation and those in the adjacent layer. This approach is applied to aqueous hydrated Gd3+ and to Gd3+ coordinated to three different binding environments: EDTA, calmodulin58 and peptides from an optimally developed family of molecularly engineered lanthanide-binding tags (LBTs)59–63. In most instances the chloride and acetate salts of the tested cations are used to eliminate effects due to different anions when comparing different cations. Additionally, different halide salts of magnesium, potassium and sodium are compared to evaluate the anion specificity of the cation effect.
All of these Gd3+ binding molecules have well characterized calcium binding sites. There are three to four coordinated water molecules in the single binding site of Gd3+ coordinated EDTA that are hydrogen bonded to the waters in the outer hydration shell51,64,65. Calmodulin63 has four binding sites, of which I and II binding sites have similar higher affinity for Gd3+. In each of the higher affinity binding sites, one water is found coordinated to the bound metal ions66. The complexity due to the mixed stretch frequency behaviors of either the multiple first hydration waters of coordinated Gd3+ in EDTA or the single first hydration water from multiple binding sites in calmodulin is addressed by employing 17 amino acid LBTs peptides that have a single EF hand calcium binding motif similar to those seen in typical calcium binding proteins such as troponin c59–61. Relatively modest site-specific changes in a troponin c-based EF hand motif have produced LBT’s with extremely high affinity for lanthanides59–61. Two LBT’s are chosen: mSE3 that has single water coordinated to a bound lanthanide and SE2 with no water in the inner hydration sphere of the coordinated lanthanide59–61 (and unpublished results from Langdon Martin in the B. Imperiali Lab at MIT. See Supplemental Table 1 for the amino acid sequences of mSE3 and SE2). The single coordinated water in mSE3 allows for a more straightforward comparison of the effect of added Hofmeister cations on the hydrogen bonding between the first shell water of the bound lanthanide ion and adjacent waters.
The results show a clear pattern in which for chloride or acetate anions added cations weaken the hydrogen bond between the first and second shell waters and that the degree of weakening follows ordering consistent with both the Hofmeister series and the charge density of cations.
SE2 and mSE3 peptides were obtained as a generous gift from Dr. Imperiali (Dept of Chemistry, MIT). All other materials were purchased from Aldrich-Sigma and used without further purification. Calmodulin from bovine testes was purchased as BioUltra grade.
All aqueous samples were prepared using distilled deionized water. Aqueous unbound GdCl3 samples for titration of Hofemeister series salts were made by mixing GdCl3 stock solution with suitable aliquots of salts stock solution to achieve final concentration of 0.5 M GdCl3 with desired amount of salts.
Solutions of 100 mM EDTA, 1 mM calmodulin, 1 mM apo-mSE3 and 1 mM apo-SE2 were prepared in 10mM HEPES buffer at pH 7.0 with 80 mM, 1 mM and 1 mM GdCl3, respectively. The extremely high affinity of Gd3+ for all of the binding sites in these calcium binding materials ensures that under experiment conditions all Gd3+ coordinates to the binding sites at the concentrations used (see Supplemental Table 2 for detailed binding constants). Aliquots of salts stock solutions were added to these samples to compare ions specific effects on the hydration shell waters of coordinated Gd3+. The relatively weak affinity of added cations other than Gd3+ prevents the substitution of coordinated Gd3+ with Hofmeister series cations used in the present studies (see Supplemental Table 2). Trehalose derived glass matrices doping with GdCl3 and salts were prepared by dissolving 0.25 M GdCl3 and 0.5 M salts with 100 mg trehalose in 1 mL solution. Aliquots of the mixture solution were placed on the horizontally positioned surface of glass sheets and dried in the vacuum in a desiccator until a transparent glass film was formed. The glass films were put in the oven at 50 °C for 1 hour for the further drying.
The Raman spectrum of the OH stretching mode of bulk water and the GVSBLS were acquired on a QuantaMaster Model QM-4/2000SE enhanced performance scanning spectrofluorometer (Photon Technology International, Lawrenceville, NJ). All of the solution samples were contained in quartz cuvettes and probed using 90° excitation geometry while glass samples were probed using front face geometry (around 45°). The Raman spectra were generated using a 350 nm excitation. For GVSBLS the excitation wavelength was around 273 nm where Gd3+ has a relatively enhanced absorption cross section. Continuous wave excitation to generate GVSBLS was used for samples of unbound Gd3+ and of Gd3+ coordinated EDTA. Time gated GVSBLS was used for Gd3+ coordinated calmodulin and mSE3/SE2 samples to eliminate short lived fluorescence signals from the protein or peptides (see Supplemental Figure 1). The gated GVSBLS were accumulated from a time point starting from 100 µs through 1000 µs after the pulsed excitation source. The luminescence spectrum was scanned and recorded from 290 nm to 400 nm. To rule out the possibility that the observed weak signals were not due to Raman scattering (for the un-gated signals) or non-Gd3+-derived fluorescence, excitation profiles were routinely conducted for observed peaks in emission spectra. The emission monochromator slits were adjusted in order to obtain suitable signal to noise ratios. The luminescence spectra were plotted as intensity vs. frequency shift (cm−1) from the electronic peak at 311 nm (3,2154 cm−1) to represent the energy difference that is equal to the vibration frequency of coordinated ligands. The spectra were processed and analyzed using GRAMS/32 AI (6.00) for background subtraction, smoothing and peak analysis.
The accuracy of the high resolution fluorometer achieves ±0.1 nm, which equals to ±8 cm−1 in the region of the vibronic side band by converting spectra from wavelength (nm) to wavenumber (cm−1) (Calculation: Δν(cm−1)=107×(Δλ(nm)/ λ1?λ2). The vibronic side band is at around 346 nm. Therefore: Δλ=±0.1 nm, λ1≈λ2=346 nm. Δν~±8 cm−1). Another contributing uncertainty in the peak position assignment is the random error that arises from the difference from one trace to the next, which can be controlled by averaging multiple repeated traces. The number of averaged traces is based on the signal to noise associated with the individual trace. For protein-free samples where the signal to noise is very high, an average of three traces is typical whereas for samples with weak signals, averaging can consist of as many as one hundred scans. Ordering of small peak shifts are verified by directly overlapping the normalized average spectra and comparing vibronic bands from samples with different additives. The intense electronic band at ~ 311 nm which does not shift under these conditions serves as a wavelength/frequency reference point. Under these conditions the entire peak band profile contributes to the evaluation (as opposed to just comparing peak positions) permitting facile determination if an entire band has shifted relative to another band given the line shape/band width does not change. It is consequently possible to detect shifts smaller than the resolution associated with a given wavelength, similarly to difference spectra analysis. The ordering with respect to which samples shift more than others is therefore still valid even the actual magnitude of the small shifts could be less than 10 cm−1 and claims of specific shifts in excess of 10 cm−1 are precisely accurate. Most of the measurements have been repeated on different samples on different occasions yielding consistent results and the trends are very clear and do not rest upon the exact values associated with the smallest shifts or small differences among many samples. The issues are similar to those associated with many FTIR and Raman studies
The frequency of the VSB arising from the water stretching mode (~ 3300 cm−1) measures hydrogen bonding between the first shell waters and outer group50,67. A decrease or increase of this frequency is correlated with an increase or decrease in hydrogen bond strength respectively.
Figure 2 shows a plot of the increase in the frequency of the OH stretching of hydration shell waters derived from the GVSBLS of free Gd3+ as a function of added MgCl2 and NaCl. It can be seen that with added salts the OH stretching frequency increases. At any given concentration of added salts, MgCl2 is seen to be significantly more effective in causing a frequency increase than NaCl. The figure also shows that the OH stretching frequency observed in the Raman spectrum from the same samples shows essentially no change with added salts. Whereas the GVSBLS probes only those waters in the hydration shell of Gd3+, the Raman probes all waters.
Figure 3 shows the GVSBL spectra of EDTA coordinated Gd3+ in the presence and absence of 4.0 M MgCl2. In addition to the VSB corresponding to the OH stretching mode, there are several other observable bands. The intense VSB at 1425 cm−1 and 1600 cm−1, which do not appear for free aqueous Gd3+ sample are the previously assigned 51 symmetric and anti-symmetric stretching mode of the EDTA derived C=O groups coordinated to bound Gd3+. The small shoulder at around 3000 cm−1 has been attributed 51 to the stretching of C–H groups in close proximity to Gd3+. Similar vibronic side bands are also observed for calmodulin (Supplemental Figure 2) or mSE3 coordinated Gd3+ (Supplemental Figure 3). Both calmodulin and the peptide manifest an OH stretching VSB. In contrast, the peptide SE2 known not to have any water coordinated to Gd3+ does not manifest an OH band in the GVSBLS (data not shown).
The comparison of the OH VSB spectra of free Gd3+ and Gd3+ coordinated to EDTA, calmodulin or mSE3 reveals a similar pattern in that coordination results in an increase in the OH stretching frequency. The OH stretch frequency increases by 40 cm−1, 30 cm−1 and 120 cm−1 in going from free Gd3+ to EDTA coordinated Gd3+, to calmodulin coordinated Gd3+ and to mSE3 coordinated Gd3+, respectively (see Supplemental Table 3). This result is consistent with previous findings derived from parvalbumin coordinated Gd3+ where an increase in the OH stretching frequency was also observed upon coordination51. It is also observed that the peak bandwidth substantially increases for protein and peptide bound sample (see Supplemental Figure 2 and Supplemental Figure 3).
Addition of MgCl2 and NaCl results in a concentration dependent increase in the frequency of the water stretching mode in the VSB spectrum from EDTA-coordinated Gd3+ (Figure 3), calmodulin-coordinated Gd3+ (Supplemental Figure 4) and the peptide-coordinated Gd3+ (Supplemental Figure 3). The pattern for the concentration and cation dependence is similar to what is observed for the OH stretching VSB from free Gd3+ (Figure 1). In contrast to the OH stretching band, the VSB frequencies for the C=O and C–H bands do not respond to the added salts as seen for EDTA-coordinated Gd3+ (Figure 3), calmodulin-coordinated Gd3+ (Supplemental Figure 4) and the peptide-coordinated Gd3+ (Supplemental Figure 3).
Figure 4 compares for free and EDTA coordinate Gd3+, the VSB OH frequencies as a function of cation for a series of added chloride and acetate salts at the same 1.0 M concentration. The plot is ordered with respect to charge density of the added cations. It can be seen that the magnitude of the frequency increase scales with charge density of the added cation. The increase in the frequency of the OH stretching band occurs for salts with either counter anion. The frequency increase relative to the salt-free Gd3+ solution increases in the following order: Cs+<K+<Na+<Li+<Ca2+<Mg2+<Al3+. The values for these frequencies and frequency shifts are summarized in Supplemental Table 4. This ordering roughly matches that of Hofmeister series ordering reflected in the value of the Setchenow constant that directly indicates the relative salting out ability of a given ion68 (see Supplemental Figure 5).
Figure 4 demonstrates that the cation dependent VSB OH frequency increase as a function of added cation salts is similar for EDTA coordinated Gd3+ and free Gd3+. A similar pattern on a less extensive set of salts is also seen for the protein and peptides coordinated Gd3+. Although the general patterns are similar for all of the Gd3+ complexes, the actual magnitude of the response to the added salts is dependent on both the particular environment of the Gd3+ and the specific cation as shown in Figure 5. The values of the cation dependent frequencies and shifts for the coordinated samples are provided in Supplemental Table 5.
Included in Figure 5 is the frequency for GdCl3 powder that has been dried for several hours. This degree of drying is associated with a VSB OH frequency that has undergone the largest increase relative to the initial unheated hydrated powder. Continued heating beyond this point resulted in a loss of the OH VSB, indicative of complete loss of first shell hydration waters. The OH VSB band reappears when the powder is exposed to a humid atmosphere. Similarly the initial heating-induced increase in OH frequency is attributed to a progressive loss of second shell waters resulting in an absence of hydrogen bonding interactions for the first shell waters. The similar frequencies for the OH VSB for mSE3 peptide coordinated Gd3+ in the presence of 4 M MgCl2 and the heated powder indicates that under these conditions, there is very weak or no hydrogen bonding between the single first shell water of the peptide coordinated Gd3+ and surrounding waters.
The above cation specific effects were all established using chloride salts. Acetate salts added to the free Gd3+ yield results closely resembling those obtained with chloride as shown in Figure 4. In Fig. 6a is shown the OH stretching frequency for free Gd3+ and EDTA coordinated Gd3+ as a function of added magnesium salts. There is minimal effect due to the added fluoride salt; however, the very low solubility of this salt does not allow for a valid comparison with the two other shown magnesium salts which are at the 1 M level. Additionally, attempts to compare the iodide salt of magnesium were also thwarted by extreme low solubility. There is slight suggestion of a frequency increase in going from the chloride to the bromide with both salts causing a large frequency increase relative to the salt free solution. In Fig. 6b, it can be seen that none of the potassium salts are effective in significantly changing the frequency relative to the salt-free samples. There is a suggestive pattern of increasing frequencies for the following sequence of anion salts of potassium: F−, Cl−, Br−, and I−. The anion dependence is significantly more dramatic for the sodium cation as seen in Fig. 6c. It can be seen both that the fluoride actually decreases the OH stretching frequency relative to the salt free solution and the frequency increases in going from F− to Cl− to Br− to I−. (It is the classic Hofmeister signature with a change in sign of the ion effect at the strength of water-water interactions.) Whenever an anion effect can be compared as a function of the specific cation, the effect increases in going from potassium to sodium to magnesium. Table 6 in the supplemental material provides the actual frequencies associated with each of these solutions.
As will discussed in the next section, the emerging model that accounts for the change in OH stretching frequency in the hydration shell of the Gd3+ includes competition between free mobile waters and anions for hydrogen bonding partners of the first shell waters around the Gd3+. The expectation is that in an environment that reduces the population of mobile waters, there will be an exaggerated cation-dependent anion effect. Water molecules contained within dry glassy matrices derived from sugars such as trehalose, are essentially sequestered by being incorporated into an extended hydrogen bonding network. As a result there are fewer waters available to participate in strong hydration shell interactions associated with added dopants within the glass. Under this condition, anions may become more competitive vis a vis water for sites in the second shell of solvation layer. Fig 6d shows the effect of added salts on the frequency of the OH stretching mode in the vibronic spectrum of Gd3+ within a dry glass made from trehalose. The addition of the different salts to the Gd3+ doped glass elicits a similar but more exaggerated pattern to what is seen in solution. The chloride salt of sodium has very minimal impact on the OH stretching frequency as is seen in solution for the free Gd3+. In contrast, the reduction of the OH stretching frequency seen in solution for free Gd3+ with the addition of NaF is duplicated in the glass but in an exaggerated fashion. Similarly, the MgCl2 induced frequency increase seen in solution is also seen in the glass but with a dramatically enhanced magnitude. The shifts and frequencies are included in Table 6 in the supplementary materials.
The presented results show that the addition of salts can modulate hydrogen bonding between waters in the first and second hydration layers surrounding Gd3+. For a given salt the magnitude of the effect increases with increasing concentration of the added salt. For added salts with low charge density anions such as chloride, bromide, iodide and acetate, there is weakening effect that is a function of the specific cation. The magnitude of the weakening, as reflected in the increase in the OH stretching frequency in the vibronic spectrum, scales with the charge density of the cation for a given concentration of added salts. The higher the charge density of the cation, the greater is the weakening of the hydrogen bonding for a given anion. The overall observed pattern of anion and cation effects in this study is similar to a pattern derived from a simulation23 in that both studies show opposite patterns for cations and anions with respect to the dependence of hydrogen bonding on charge density.
The results derived from the direct comparison of cations with triple, double or mono valence under the same salt concentration are potentially complicated because the counter anion concentrations are significantly different. Figure 2 shows that when comparing the Mg2+ and Na+ cations under conditions of the same chloride concentration, the Mg2+ cation still manifests the larger effect on the OH stretching frequency. This issue was further addressed by conducting comparisons among cations under the condition of high KCl concentration (4 M). The K+ cation has little or no effect on the OH stretching frequency and the chloride concentration is high enough to minimize differences in the chloride concentration when comparing the effect of cations at the same concentration. It is observed (Supplemental Figure 6 and supplemental table 7) that at the high chloride concentration, the magnitude of the OH frequency shift still tracks with the charge density of the cation (Al3+>Mg2+>Na+).
The comparison of the hydrogen bond weakening effect for the magnesium, sodium and potassium halide salt series also indicates that for both free and coordinated Gd3+ the magnitude of the effect for a specific cation is anion sensitive. For low charge density anions such as chloride, acetate, bromide and iodide there are only slight anion dependent differences. The overall pattern indicates that for a given cation, the weakening of the hydrogen bonding within the hydration layer increases with increasing charge density of the anion as reflected in the progression: iodide> bromide >chloride> >fluoride. The overall pattern is consistent with the model, to be discussed below, in which the anion effect is modulated by the cation with the effect being greatest for high charge density cations.
The proposed model is depicted in the Figure 7 schematic that shows key elements that account for the vibronic spectra results. The changes in the OH stretching frequency in the vibronic spectra reports on the strength of interaction H between the waters in the first hydration layer and the solvent molecules (water or anions) in the second shell hydration layer. There are three potential influences perturbing interaction H: Influence 1 due to the cation and its cluster of waters that comprise its hydration shell; influence 2 due to the anion and influence 3 due to the population of free waters not directly interacting with either the anion or cation. It has been shown using GVSBS that influence 3 can be modulated with the addition of osmolytes such as sugars and glycerol that enhance the formation of hydrogen bonding networks which in turn increases H69.
The cation water clusters are expected to have an indirect effect on H. This follows because the weakening effect by cations is not likely due to a direct polarity effect between the two positively charged cations that can be expected to be separated by at least several layers of water under the conditions of these measurements.
There are two possible indirect ways that the high charge density cations influence interaction H. Both effects arise from the formation of a tight cluster of waters around the high change density cation. The first effect is through the cation-induced enhancement in the relative population of high density water through disruption of hydrogen bonding in the vicinity of the high charge density cation (influence 1). This effect originates from the ordering of water in the hydration shell of the added cations that inhibits the formation of the optimum configuration needed to form low density water7,23. As a consequence there is a further build up of the relative concentration of high density water. The waters participating in a high density water cluster do not hydrogen bond to the hydration waters as effectively as those in low density water clusters (influence 3 in the schematic).
The second effect is due to the sequestration of water in the tight hydration shell of the high charge density cations70. With increasing sequestration of water, there is a greater probability of an anion participating in the hydrogen bonding interactions associated with the hydration shell of the Gd3+. The presented results indicate that chloride, bromide and iodide are not as effective as free (non-sequestered) water with respect to hydrogen binding to the waters in first hydration shell. In contrast, the results are indicative of fluoride ion being able to strongly participate in or enhance, through some indirect mechanism, the hydrogen bonding among hydration layer waters. This difference between anions may arise from several factors including steric limitations influencing how well different sized anions integrate into the second hydration shell, or a balance between electrostatics and orientation factors 23.
The above model is further supported from the results from the glassy matrices derived from trehalose. In the glass, the population of waters capable of participating as “normal” solution phase mobile waters in the hydrogen bonding within the hydration shell of the Gd3+ is reduced due to the recruitment of waters into the hydrogen bonding network of the glass. Under these conditions, the addition of cations that further sequester mobile waters will further deplete the population of waters capable of participating in the normal hydrogen bonding pattern within the hydration shell of the Gd3+. As a result, the hydrogen bonding between the first and second shell solvent molecules has contributions both from waters that can not adopt the optimal hydrogen bonding orientations and from the added anions which in the case of chloride does not act as an effective replacement for “normal” mobile waters. Magnesium has been shown to be highly effective at sequestering waters (about 6 waters tightly bound to Mg2+) and thereby reducing the relative population of mobile waters69,70. The addition of MgCl2 to the glass results in a very substantial weakening of the hydrogen bonding within the hydration shell of the Gd3+. The effect is not seen for NaCl but is also seen for LiCl (not shown) although not to the same extent as with MgCl2. The pattern fits with the model in that increasing charge density enhances the ability of the cation to sequester water and thus decrease the population of mobile waters that can strongly hydrogen bond to other waters. The use of the glass matrix also allows for a much clearer demonstration relative to solution phase that fluoride is effective in enhancing the hydrogen bonding within the hydration shell.
The ordering of the Hofmeister series salts with respect to weakening of the hydrogen bonding within the hydration shell of both free and coordinated Gd3+ has implications for the effect of these salts on protein properties. It has been argued that osmolytes such as urea can predispose proteins towards unfolding by enhancing the probability that waters enters into the hydrophobic protein interior71. The presence of internal water has been reported to cause peptide hydrogen bonds to lengthen72, a process consistent with a loosening of the structure. The proposed mechanism for this effect is through a shift in the balance between the entropic forces that favor water entering all accessible volumes within the protein and the enthalpic forces due to hydrogen bonding among waters that favor keeping water in the bulk phase69. Weakening of the hydrogen bonding among the external waters would enhance the entropically driven occupancy of internal sites within the protein with water molecules. The present results indicate that for low density anion salts, cations should increase the probability for water occupancy within the hydrophobic interior to a degree that scales with increasing charge density of the added cation.
The magnitude of the salt induced weakening can be estimated through the application of empirical Bauer-Badger rule which predicts an enthalpy shift of 0.16 kJ/cm−1 73,74. This relationship results in a 3 kJ/mol decrease in enthalpy for breaking the local hydrogen bonding in the hydration layer of free Gd3+ induced by 1 M MgCl2. In a 4 M MgCl2 solution, the estimated decrease in the enthalpy penalty for disrupting the hydration layer water around the metal binding site is 13 kJ/mol. Considering that most globular proteins are marginally stable with ΔGfolding of about −40 kJ/mol75–79, it is likely that the osmolyte dependent stability of the hydration shell waters will contribute to conformational properties as is seen for the osmolyte-induced quaternary structure transition in a dimeric hemoglobin80.
Proteins must be dynamically active to function. There is growing recognition that the wide varieties of dynamics that drive and modulate protein function are highly responsive to the surrounding solvent26–28,81,82. The emerging picture is that protein dynamics can be grouped hierarchically based on which category of solvent motion they are slaved to27–30,81,82. There are several basic concepts behind the solvent slaving models. First and foremost is the idea that a given protein molecule even at equilibrium can access a large number of conformational substates that are energetically comparable in overall stability83,84. These substates differ in the details of side chain packing and orientation although the overall global tertiary structure is similar for the members of the distribution. As a result of these differences, the substates can be functionally distinct with only some fraction of the distribution having measurable rates of reactivity. In the absence of solvent motions the barrier between substates is very high and as a consequence there is very slow inter-conversion among substates resulting in phenomena such as kinetic hole burning85–92. When only a limited set of substates are functionally active, the slow down of the inter-conversion of substates will result in a shut down in activity since the majority of the population will not be able to access the active substate conformations. The solvent slaving concept asserts that solvent motions reduce the enthalpic barrier between substates and allowing a quick entropic search among the many substates for the reactive species. This is the molecular basis for the idea of solvent as a lubricant for protein motions. The solvent controlled barrier between substates now controls the overall enthalpy for the protein process in question whereas the rate for the observed functional process also depends upon the duration of the entropic search, i.e. the number of substate transitions needed to access the functionally active species. The present results suggest that Hofmeister osmolytes can control protein reactivity by enhancing the barrier between substates as in the case of sugars, polyols that favor increased hydrogen bonding or decreasing the barrier by weakening the hydrogen bonding between hydration shell waters using the appropriate salts.
The initial VSB studies showed that the hydrogen bonding within the hydration shell of Gd3+ is stronger than that among waters in the bulk phase57,93. The present study shows that coordination to a chelate, a peptide or a protein causes a weakening of the hydrogen bonding within the hydration shell of the Gd3+. In these cases where the Gd3+ is coordinated, the weakening of hydrogen bonding is most plausibly explained in terms of steric factors. The confining environment surrounding the coordinated Gd3+ cation likely limits the ability of the hydration shell waters to adopt the geometry among the surrounding waters needed to create the low density water clusters36,51,69 that are associated with the strongest hydrogen bonding. The more constrained the environment, the weaker is the hydrogen bonding network, as the case of a Gd3+ coordinating peptide mSE3. This proposed weakening of the hydrogen bonding pattern associated with waters in the environment surrounding a first shell hydration water on a nanoscale-rough surface, should result in an increase in the relative concentration of high density water in that hydration layer. It has also been claimed that the equilibrium between the two species of water clusters solvating nonpolar groups is shifted toward the high density clusters when polar group dominates, as in the case for calcium binding sites on the surface of proteins36,94. Therefore at least two factors should lead to the weakened hydrogen bonding among waters surrounding the coordinated Gd3+: 1) steric constraint limiting the formation of the hydrogen bonding network need to form low density water clusters and 2) polar groups that disrupt already distorted hydrogen bond patterns which also favors higher density clusters.
Weakening of the hydrogen bonding among hydration shell waters due to steric constraints on the protein surface has implications for the osmolyte effects. If through the steric factors, there is an increase in high density water on the protein surface, then osmolytes such as Mg2+ and urea that do not readily integrate into the hydrogen bonding network of low density water clusters should have a higher occupancy in the hydration layer vis a vis the bulk solvent. Thus the rough surface of the protein lowers the enthalpic penalty of moving cations such as Mg2+ into the hydration layer and raises the enthalpic penalty for osmolytes such as glycerol that participate in hydrogen bonding networks69. This explanation may also contribute to the apparent discrepancy between the Raman signal from the bulk solvent and the local signal from the GVSBLS in that the effect on water should be enhanced where the local concentration of the cation is enhanced.
The present study supports a mechanism for the Hofmeister ordering of salts based on a disruption of hydrogen bonding among hydration shell waters (but not bulk water). The observation of combined anion and cation specific hydration shell effects raises the prospect for a better understanding of how in vivo levels of specific osmolytes modulate protein structure and dynamics. Furthermore, these studies provide a basis for manipulation of protein properties through a universal mechanism based on tuning the hydrogen bonding within the hydration shell waters of proteins.
Comparison and data of several Hofmeister series ions induced OH GVSBLS shifts of free Gd3+ or Gd3+ coordinated to calmodulin or mSE3; affinity of calmodulin or SE2/mSE3 peptides for cations; and SE2/mSE3 peptides sequence.
This work was supported through funding from National Institutes of Health Grant P01HL071064. The authors wish to acknowledge and thank Professors Barbara Imperiali (MIT) and Karen Allen (Boston University) for the information, advice and materials they generously provided.