J. Phys. Chem. B 2007, 111, 5411-5417
5411
Hofmeister Salt Effects on Surface Tension Arise from Partitioning of Anions and Cations between Bulk Water and the Air-Water Interface Laurel M. Pegram*,† and M. Thomas Record, Jr.*,†,‡ Departments of Chemistry and Biochemistry, UniVersity of Wisconsin-Madison, Madison, Wisconsin 53706 ReceiVed: January 11, 2007; In Final Form: February 21, 2007
We apply a recently developed surface-bulk partitioning model to interpret the effects of individual Hofmeister cations and anions on the surface tension of water. The most surface-excluded salt (Na2SO4) provides a minimum estimate for the number of water molecules per unit area of the surface region of 0.2 H2O Å-2. This corresponds to a lower bound thickness of the surface region of ∼6 Å, which we assume is a property of this region and not of the salt investigated. At salt concentrations j1 m, single-ion partition coefficients Kp,i, defined relative to Kp,Na+ ) Kp,SO42- ) 0, are found to be independent of bulk salt concentration and additive for different salt ions. Semiquantitative agreement with surface-sensitive spectroscopy data and molecular dynamics simulations is attained. In most cases, the rank orders of Kp,i for both anions and cations follow the conventional Hofmeister series, qualitative rankings of ions based on their effects on protein processes (folding, precipitation, assembly). Most anions that favor processes that expose protein surface to water (e.g., SCN-), and hence must interact favorably with (i.e., accumulate at) protein surface, are also accumulated at the air-water interface (Kp >1, e.g., Kp,SCN- )1.6). Most anions that favor processes that remove protein surface from water (e.g., F-), and hence are excluded from protein surface, are also excluded from the airwater interface (Kp,F- ) 0.5). The guanidinium cation, a strong protein denaturant and therefore accumulated at the protein surface exposed in unfolding, is somewhat excluded from the air-water surface (Kp,GuH+ ) 0.7), but is much less excluded than alkali metal cations (e.g., Kp,Na+ ≡ 0, Kp,K+ ) 0.1). Hence, cation Kp values for the air-water surface appear shifted (toward exclusion) as compared with values inferred for interactions of these cations with protein surface.
1. Introduction The Hofmeister series provides a qualitative ranking of salt anions and cations, originally based on their effectiveness (typically at molar concentrations) as protein precipitants, and subsequently found to be applicable to many other protein processes in which water-accessible surface area (ASA) changes significantly (folding, assembly).1,2 Considering these processes in the direction in which protein ASA is reduced (e.g., folding, subunit assembly, crystallization, precipitation) the general ranking of ions, in descending order of effectiveness in driving these processes, is as follows:3
SO42- > F- > Ac- > Cl- > Br- > NO3- > I- > ClO4- > SCNK+ ) Na+ > Li+ > NH4+ . GuH+ Very similar rankings of these ions are observed in their effects on the surface tension of water (see Figure 1),2,4 in their effects on the solubility of benzene and other hydrocarbons,5 in their retention relative to tritiated water on a recycling polyacrylamide column,6 and in their effects on other properties of, and processes in, aqueous solutions. This subject has been reviewed by Baldwin, who concluded that the effects of Hofmeister salts on biopolymer processes result from ion-specific interactions with both amide and nonpolar surface.2
Figure 1. (A) Representative surface tension data for selected Hofmeister salts. Open symbols represent sodium salts,48 while the corresponding filled symbol denotes the guanidinium salt of the same anion.35 The symbols for the various anions are as follows: sulfate (triangle), chloride (square), and bromide (circle). Error bars for the sodium salts are taken from the source; precision estimates provided for the guanidinium salts seem abnormally high ((∼0.35) and are not shown. (B) Representation of all STIs gleaned from the literature. The average STI for each electrolyte, as presented in Table 1, is shown as a straight line between 0 and 1 m. Electrolytes with a common anion are depicted by lines of the same color.
*To whom correspondence should be addressed. E-mail:
[email protected] (L.M.P.),
[email protected] (M.T.R.). † Department of Chemistry. ‡ Department of Biochemistry.
Salts of ions from the extremes of the Hofmeister series are widely used at high concentrations in biochemistry (e.g., ammonium sulfate for precipitation and crystallization; guani-
10.1021/jp070245z CCC: $37.00 © 2007 American Chemical Society Published on Web 04/14/2007
5412 J. Phys. Chem. B, Vol. 111, No. 19, 2007 dinium chloride or thiocyanate for denaturation), but the molecular basis of their ion-specific effects remains controversial. The goal of our research is to interpret or predict these salt effects in terms of the amount and type of biopolymer ASA buried or exposed in the process, by quantifying the individual thermodynamic contributions of interactions of anions and cations with different types of biopolymer surface. For salts, precedent for this approach is provided by the recycling chromatographic studies of von Hippel and colleagues,6 who calculated single-ion amide binding constants based on the different elution times (relative to THO) of different Hofmeister salts from a polyacrylamide column. The thermodynamic description of Hofmeister effects, and all other effects of electrolyte and nonelectrolyte solutes, on biopolymer processes is most fundamentally based on preferential interaction coefficients which quantify the thermodynamic consequence of the local accumulation or exclusion of the solute from the vicinity of the hydrated biopolymer surface as a function of solute concentration.7,8 Initial steps have been taken to quantify preferential interactions of electrolytes and uncharged solutes with biopolymer surfaces with use of a recently developed solute-partitioning model, in which a partition coefficient quantifies the distribution of an ion or uncharged solute between the bulk water and a local environment (i.e., water of hydration) at the surface of a biopolymer.9-11 Early evidence for differential partitioning of salt cations and anions at the air-water interface was provided by the observation that the surface potential varies widely for different salts of the same valence.12,13 Within the past 5 years, additional evidence that the cation and anion of a salt partition differently between the surface and bulk solution, and that some anions are actually accumulated at the surface, has been provided by MD simulations, surface spectroscopic measurements, and electrospray ionization mass spectrometry:14-20 cf. recent reviews by Petersen and Saykally21 and Jungwirth and Tobias.22 We recently developed a solute-partitioning model for individual ions between bulk solution and the air-water interface, and applied this model to analyze the effects of atmospherically relevant ions on the surface tension γ of water, as quantified by surface tension increments (STI; dγ /dm2, where m2 is molal salt concentration).23 In the present paper, we extend the analysis to the complete spectrum of the Hofmeister series and use the literature surface tension data (cf. Figure 1) to determine partition coefficients of Hofmeister anions and cations between the bulk solution and the air-water interface. Comparison of single-ion partition coefficients for the air-water and various types of biopolymer surface will yield insight into the thermodynamic and molecular origins of Hofmeister ion effects on biochemical processes. For example, analysis of literature data on the effects of Hofmeister salts on the solubility of hydrocarbons (model compounds for nonpolar surface in biopolymers24) by using an analogous model to that used here to analyze salt effects on surface tension reveals that both the amount of local water per unit area of surface and the partition coefficients for individual ions are similar to those obtained here (LMP and MTR, manuscript in preparation). 2. Thermodynamic Analysis and Background 2.1. Analysis of Salt Effects on Surface Tension with Use of the Solute Partitioning Model and Guggenheim Treatment of Interfaces. The Guggenheim treatment of surface tension considers the surface (denoted by σ) as a microphase of small but finite thickness at equilibrium with the bulk phase (b).25 For this model, Guggenheim developed the thermodynamic analysis of the STI for both uncharged and electrolyte solutes.26
Pegram and Record For electrolytes, this formal analysis in terms of the chemical potential of the electroneutral component precludes the attainment of ion-specific information. Although evidence for unequal surface/bulk concentration ratios of individual ions has existed for some time,21,27 a single-ion partitioning analysis has not previously been utilized to predict or interpret the effects of unequal distributions of electrolyte ions (Hofmeister cations and anions) on surface tension. For a two-component solution of any binary electrolyte (Mν+Xν-), our extension of the Guggenheim analysis yields eq 1, valid over a range of low electrolyte concentrations, to interpret surface tension increments of electrolytes in terms of partitioning of individual electrolyte ions between surface and bulk regions:23
[(
) ]
RTbσ1 ν ν+Kp,+ + ν-Kp,dγ )(1 + b() -1 • dm2 ν m 1
(1)
In eq 1, the individual ion partition coefficients Kp,i are defined as ratios of molal concentrations of the ions in the surface and bulk regions: Kp,i ≡ mσi /mbi = mσi /νim2. The term (1 + b() enters from the conversion of an activity derivative to a concentration derivative and is defined as (1 + b() ≡ (1 + dlnfb(/d lnm2) ) ν-1 (dOsm/dm2). Also in eq 1, bσ1 ≡ nσ1 /A is the number of water molecules per unit surface area in the surface phase, ν ≡ ν+ + ν- is the number of ions per formula unit of the salt, and m•1 is the solvent molality (55.5 mol/kg for H2 O). The ion partition coefficients (Kp,+, Kp,-) are related to the partition coefficient obtained by treating the salt as an electroneutral component (Kp,2):
νKp,2 ≡ ν+Kp,+ + ν-Kp,-
(2)
Although in principle the partition coefficients (Kp,+, Kp,-) of the salt ions as well as the nonideality factor (1 + b() and other aspects of the partitioning process underlying eq 1 may be concentration dependent,11,27 surface tension data as a function of inorganic electrolyte concentration in most cases are linear up to 1.0 m or higher (cf. Figure 1),28 indicating the extended applicability of the limiting behavior described by eq 1. (This may in part reflect compensations between the various possible concentration-dependent terms.) Since the m2 dependence of (1 + b(), though typically not large in the range of interest here (0.1-1.0 m), is one possible source of an m2-dependence of the STI, it is also useful to analyze the derivative of surface tension with respect to osmolality; from eq 1
[(
) ]
RTbσ1 ν+Kp,+ + ν-Kp,dγ -1 )- • dOsm ν m 1
(3)
In practice, analyses of extant surface tension data using either eq 1 or eq 3 give equivalent results (see below). The simplest interpretation of the observation that STI are relatively independent of m2 is that the Kp,i behave like macroscopic equilibrium partition coefficients and are relatively independent of m2, and that the thickness of the surface layer is also independent of m2. We limit the analysis and discussion in most cases to the concentration range 0.1-1.0 m where STI are concentration independent. Most inorganic acids reduce surface tension (i.e., dγ/dm2 < 0); application of eq 1 indicates that Kp,H+ ≈ 1.5.23 Accumulation of H+ in the air-water surface layer explains the reduction of the surface tension for most inorganic acids; only for H2SO4 does the effect of exclusion of SO42- (Kp = 0) outweigh the
Hofmeister Salt Effects on Surface Tension
J. Phys. Chem. B, Vol. 111, No. 19, 2007 5413
TABLE 1: Average dγ/dm2 Values28,30,35,48-57 electrolyte
dγ/dm2 ( s.d.
no. of sources
dOsm/dm2a
Na2SO4 Na2CO3 NaF NaOH NaCl NaBr NaNO3 NaI NaAc NaClO3 NaSCN NaClO4
2.77 ( 0.09 2.70 ( 0.07 1.81 ( 0.02 1.86 ( 0.10 1.73 ( 0.17 1.47 ( 0.21 1.21 ( 0.01 1.14 ( 0.09 0.93 ( 0.03 0.75 ( 0.15 0.50 ( 0.11 0.22 ( 0.06
4 3 3 1 6 5 2 5 1 2 1 1
1.93 ( 0.02 2.14 ( 0.02 1.74 ( 0.03 1.90 ( 0.04 1.85 ( 0.01 1.89 ( 0.01 1.70 ( 0.04 1.94 ( 0.01 1.96 ( 0.01 1.75 ( 0.01 1.91 ( 0.05 1.81 ( 0.01
K2SO4 K2CO3 KF KOH KCl KBr KNO3 KI KAc KClO3
2.57 ( 0.11 3.01 ( 0.10 1.80 ( 0.11 1.88 ( 0.10 1.59 ( 0.13 1.35 ( 0.02 1.07 ( 0.04 1.15 ( 0.16 0.76 ( 0.09 0.45 ( 0.36
1 1 1 2 6 2 2 5 1 1
1.95 ( 0.06 2.34 ( 0.01 1.83 ( 0.01 1.95 ( 0.04 1.78 ( 0.01 1.80 ( 0.01 1.54 ( 0.05 1.83 ( 0.01 1.99 ( 0.04 1.56 ( 0.06
Li2SO4 LiOH LiCl LiBr LiNO3 LiI LiAc LiClO4
3.13 ( 0.45 1.65 ( 0.12 1.65 ( 0.17 1.31 ( 0.12 1.23 ( 0.12 0.78 ( 0.12 0.84 ( 0.05 0.27 ( 0.06
2 1 5 1 1 1 1 1
2.30 ( 0.02 1.71 ( 0.02 1.98 ( 0.02 2.00 ( 0.05 1.95 ( 0.04 1.87 ( 0.02 1.89 ( 0.02 2.08 ( 0.06
(NH4)2SO4 NH4Cl NH4Br NH4NO3 NH4I
2.29 ( 0.12 1.39 ( 0.16 1.28 ( 0.06 1.08 ( 0.08 0.74 ( 0.12
2 3 1 2 1
1.91 ( 0.05 1.78 ( 0.01 1.80 ( 0.01 1.65 ( 0.03 1.81 ( 0.01
Cs2SO4 CsCl CsAc
3.02 ( 0.07 1.57 ( 0.05 1.12 ( 0.10
1 2 1
2.12 ( 0.02 1.70 ( 0.01 2.01 ( 0.05
(GuH)2SO4 GuHCl GuHBr
0.94 ( 0.16 0.75 ( 0.24 0.63 ( 0.18
1 1 1
1.83 ( 0.25 1.62 ( 0.04 1.63 ( 0.04
0.49 ( 0.07 -0.28 ( 0.01 -0.48 ( 0.12 -0.80 ( 0.04 -1.82 ( 0.33
4 2 1 2 2
2.07 ( 0.05 2.01 ( 0.05 2.07 ( 0.06 1.92 ( 0.03 2.02 ( 0.05
H2SO4 HCl HBr HNO3 HClO4
a The errors reported are given by the 95.4% confidence range (two standard deviations) estimated by the fit residuals in IgorPro 5.03.
accumulation of H+ and result in a positive STI. For inorganic salts, dγ/dm2 > 0, so, from eqs 1 and 2, values of Kp,2 are less than unity. In this paper, we apply eqs 1 and 2 to literature surface tension data and obtain Kp,i values for the individual ionic components (relative to a reference excluded salt). 3. Analysis Shown in Table 1 are the averages and standard deviations29 for the surface tension increment (STI, dγ/dm2) for 46 electrolyte (salts, acids, bases) solutions. (See the Supporting Information for the method of determination of each STI reported in Table 1: least-squares analysis of tabulated data, slope reported, calculated from a graph, etc.) Values of STI presented in Table 1 are compiled from literature data in the concentration range 0.1-1.5 m. All data sets were obtained at a constant temperature, in the range 15-30 °C. Although some temperature dependence of the STI is expected from eq 1 (arising from the direct
dependence (RT) and any temperature dependence of Kp,i or bσ1 ), within this temperature range, small systematic differences between different studies appear to have more significant effects on dγ/dm2 than differences in temperature. Outliers were defined as values of dγ/dm2 more than two standard deviations away from the average for a particular electrolyte. Also presented in Table 1 are the nonideality factors, (dOsm/ dm2) ≡ ν(1 + (), determined from the reported molal osmotic coefficients.30-35 Electrolyte osmolality was plotted versus solute molality and fit linearly in the concentration range of interest (0.1-0.7 m).36 We find a linear dependence of surface tension on osmolality (eq 3) for those solutes where tabular data are available, but neither is there enough tabular data nor is the experimental uncertainty small enough to determine whether γ is significantly more linear as a function of Osm than as a function of m2. Equation 1 predicts that the nonideality-corrected STI, (dγ/ dm2)/(1 + (), is determined by the thickness of the surface region (related to bσ1 ≡ nσ1 /A, in units of solvent molecules/Å2) and by the sum of the stoichiometrically weighted partitioning terms, νi(Kp,i - 1), for the individual ions of the electrolyte component. Values of the composite intrinsic molecular thermodynamic quantity, (Kp,2 - 1)bσ1 , where Kp,2 is the component partition coefficient (eq 2), are tabulated in the Supporting Information. Negative values of (Kp,2 -1)bσ1 indicate electrolytes for which the total interfacial concentration of ions is less than the total bulk concentration, even if the interfacial cation and anion concentrations differ. A value of (Kp,2 - 1)bσ1 ) 0 signifies an electrolyte with a total interfacial ion concentration equal to that in the bulk, and positive values of (Kp,2 - 1)bσ1 indicate net accumulation at the interface. Clearly, all of the salts have negative values, but there is a large range, from -0.016 for NaClO4 to -0.194 for Na2SO4. The acids exhibit an equally large range of mostly positive values, from +0.122 (HClO4) to -0.032 (H2SO4). To obtain composite partition coefficients Kp,2 of electrolytes (eqs 1 and 2), we assume that the thickness of the surface water layer is independent of the concentration and nature of the electrolyte and use the salt with the largest surface tension increment to obtain a lower bound value of bσ1 . The most excluded solute in this data set, Na2SO4, for which (Kp,2 - 1)bσ1 equals -0.194, is assumed to be completely excluded from the surface layer (i.e., the component partition coefficient, Kp,2, is set to zero), equivalent to the assumption that both sodium and sulfate ions are fully excluded from the interfacial region (Kp,Na+ ) Kp,SO42- ) 0).37 Values of Kp,2 were then analyzed to obtain individual ion Kp,i values by using eq 2 and the assumption that Kp,Na+ ) Kp,SO42- ) 0. The sulfate data sets were used to generate partition coefficients for each cation, with the exception of H+ and GuH+. To avoid possible complications arising from incomplete dissociation of H2SO417 and ion pairing in (GuH)2SO4 solutions,38 we used the average chloride partition coefficient (Kp,Cl- ) 0.69) to determine the partition coefficients for H+ and GuH+. The cation partition coefficients obtained in this way were then used to generate the remaining anion partition coefficients. 4. Results and Discussion 4.1. Overview of Surface Tension Increments (STI) of Hofmeister Salts and Other Electrolytes. All Hofmeister salts (as well as all strong bases) investigated have positive STI, whereas acids (with the exception of H2SO4) have negative STI (cf. Table 1). Representative data with error bars are shown in Figure 1A, and the entire collection of surface tension data is
5414 J. Phys. Chem. B, Vol. 111, No. 19, 2007
Pegram and Record
TABLE 2: Composite Kp,2 Values for Electrolytes at the Air-Water Interface 2-
SO4 CO32FHOClBrNO3IAcClO3SCNClO4-
Na+
Cs+
K+
Li+
NH4+
GuH+
H+
0.000 ( 0.045 0.122 ( 0.031 0.277 ( 0.008 0.319 ( 0.037 0.350 ( 0.066 0.459 ( 0.079 0.507 ( 0.006 0.591 ( 0.033 0.670 ( 0.013 0.704 ( 0.055 0.818 ( 0.040 0.915 ( 0.023
0.008 ( 0.025
0.083 ( 0.048 0.105 ( 0.030 0.315 ( 0.056 0.329 ( 0.038 0.377 ( 0.052 0.480 ( 0.006 0.519 ( 0.022 0.563 ( 0.060 0.734 ( 0.032 0.799 ( 0.161
0.053 ( 0.136
0.167 ( 0.047
0.642 ( 0.078
0.835 ( 0.025
0.328 ( 0.049 0.418 ( 0.061 0.544 ( 0.043 0.561 ( 0.044 0.710 ( 0.045 0.691 ( 0.019
0.456 ( 0.062 0.505 ( 0.023 0.546 ( 0.033 0.715 ( 0.046
0.678 ( 0.103 0.731 ( 0.077
1.097 ( 0.004 1.161 ( 0.041 1.288 ( 0.013
0.357 ( 0.021
0.612 ( 0.036
shown in Figure 1B, where the slopes of the lines are the averages from Table 1. As seen in Figure 1B, the differences between the salts are dominated by the anions, with a smaller cation modulation. Large differences in STI are observed in comparisons of salts with a common cation and anions from the opposite ends of the Hofmeister series (e.g., Na2CO3 (+2.70 dyn/cm) and NaF (+1.81 dyn/cm) vs NaClO3 (+0.75 dyn/ cm) and NaSCN (+0.50 dyn/cm)), with intermediate values of STI for anions in the middle of the Hofmeister series (e.g., NaCl (+1.73 dyn/cm), NaBr (+1.47 dyn/cm), and NaI (+1.14 dyn/ cm)). As previously noted,2,39-41 the resulting anion series generally follows the Hofmeister series; the largest values of STI are observed for those salts (sulfates, carbonates, fluorides) that favor processes that reduce the exposure of protein surface to water and solute (protein folding, precipitation, aggregation, etc), while the smallest values of STI correspond to salts that favor the exposure of protein surface (guanidinium salts, thiocyanates, perchlorates). The acetate salts are a notable exception; though acetate is toward the stabilizing (folding, precipitating, assembling, etc) end of the Hofmeister series, its STI (average ∼0.9 dyn cm-1/molal for NaAc, KAc, LiAc, and CsAc) is similar to that of iodide (∼1.0 dyn cm-1/molal), which is near the destabilizing end (solubilizing, disassembling, etc) of the Hofmeister series. In contrast with the behavior of the anion series, values of STI observed in series of salts with different cations and a common anion (e.g., chlorides or sulfates) are tightly grouped (Figure 1B), with the exception of H+ and GuH+. The separation of GuH+ from other cations (values of STI of GuH+ salts are generally less than half as large as those of the corresponding alkali metal salts) agrees well with the relative placement of GuH+ and alkali metal cations in the Hofmeister series. For any choice of anion, only minor differences in STI are observed in comparisons of salts of the Group I cations (Li+, Na+, K+, and Cs+) with one another and with the ammonium cation. Sodium salts generally increase the surface tension to the greatest extent; ammonium salts increase surface tension the least; and lithium, potassium, and cesium salts are indistinguishable and intermediate. 4.2. Single-Ion Partition Coefficients (Kp,i) for Cations and Anions of Hofmeister Salts. 4.2.1. Kp,i Values of Cation and Anions Are Independent and AdditiVe. Individual ion Kp,i were obtained by using eqs 1 and 2 and the assumption that Kp,Na+ ) Kp,SO42- ) 0. The cation and anion partition coefficients are presented in Tables 3 and 4, respectively. The second column in Table 4 lists the averages for each anion (determined for salts of different cations) along with the standard deviation. Additivity is validated by the similar values for the anions of different cations. For example, values of the single-ion partition coefficient for Cl- (Kp,Cl-) obtained by dissection of the surface tension increments for NaCl, CsCl, KCl, LiCl, and NH4Cl are
0.910 ( 0.020
1.627 ( 0.115
TABLE 3: Cation Partition Coefficients cation +
Na Cs+ K+ Li+
average Kp,+
cation
average Kp,+
0.00 ( 0.05 0.01 ( 0.04 0.12 ( 0.08 0.08 ( 0.21
+
0.25 ( 0.07 0.67 ( 0.21 1.50 ( 0.04
NH4 GuH+ H+
0.70, 0.70, 0.63, 0.76, and 0.66, respectively, yielding an average Kp,Cl- ) 0.70. This cation-invariant value of the Cl- partition coefficient (with a standard deviation of 0.04 and no apparent trend across the table) indicates that the cation and anion of these 1:1 salts partition independently, and confirm the prediction of eqs 1 and 2 that the STI of an electrolyte is composed of additive contributions of the cation and anion. Nearly all of the results in Table 4 are consistent with the assumption that bσ1 is a physical characteristic of the surface region of water.42 The Group I cations are all strongly excluded from the surface (average Kp ) 0.05 ( 0.06, i.e., local concentration 1/20 of the bulk), NH4+ is less excluded (Kp ) 0.25), and GuH+ is only slightly excluded (similar to Cl-, with a Kp ) 0.67) (Table 3). 4.2.2. Comparison of Surface Tension and Biopolymer Hofmeister Series of Anions and Cations. The relative rankings of the cation and anion partition coefficients in Tables 3 and 4 generally agree qualitatively with the Hofmeister series for ion effects on biopolymer processes. Figure 2 is a comparison of single-ion partition coefficients for air-water surface with qualitative rankings summarizing the effects of various salts on protein processes. The columns on the left contain quantitative single-ion coefficients, describing partitioning between the airwater surface and the bulk, and those on the right represent the qualitative rankings of ions gleaned from experimental data on the effectiveness of the ions in favoring or disfavoring processes that remove protein surface from water.3 For these protein processes, changes in KCl or NaCl concentration (in the molar range) are generally observed to have little to no effect (i.e., Kp,2 ∼1), and so we assume therefore that both the cation (Na+, K+) and the anion (Cl-) of these salts partition approximately equally between the protein surface involved in the process and bulk water.43 Figure 2 shows that STI-derived anion partition coefficients are quite consistent in both rank order and approximate magnitude with the biopolymer Hofmeister series of anions. Relative to the highly excluded sulfate anion, the anions range from almost complete exclusion from the air-water surface (CO32-) to weak exclusion (F-, HO-) and from an approximately uniform distribution (Br-, NO3-, I-) to moderate surface accumulation (ClO3-, SCN-, ClO4-). Thiocyanate (SCN-) is an example of the good agreement between the STI-derived partition coefficient and the inferred biopolymer partition coefficient; a strong protein denaturant (and thus accumulated at protein surface), SCN- is also one of the most accumulated
Hofmeister Salt Effects on Surface Tension
J. Phys. Chem. B, Vol. 111, No. 19, 2007 5415
TABLE 4: Anion Partition Coefficients anion SO42CO32FHOClBrNO3IAcClO3SCNClO4a
average ( s.d.
Kp,Na+ ) 0.00
0.22 ( 0.15 0.53 ( 0.02 0.58 ( 0.04 0.69 ( 0.04 0.86 ( 0.08 0.98 ( 0.09 1.18 ( 0.12 1.30 ( 0.05 1.44 ( 0.03 1.64 1.77 ( 0.04
0.00 ( 0.05 0.37 ( 0.13 0.55 ( 0.05 0.64 ( 0.09 0.70 ( 0.14 0.92 ( 0.16 1.01 ( 0.05 1.18 ( 0.07 1.34 ( 0.05 1.41 ( 0.12 1.64 ( 0.09 1.83 ( 0.06
Kp,Cs+ ) 0.01
0.70 ( 0.06
1.21 ( 0.08
Kp,K+ ) 0.12 0.07 ( 0.18 0.51 ( 0.13 0.53 ( 0.11 0.63 ( 0.13 0.83 ( 0.06 0.91 ( 0.09 1.00 ( 0.14 1.34 ( 0.10 1.47 ( 0.33
Kp,Li+ ) 0.08
0.58 ( 0.23 0.76 ( 0.24 1.01 ( 0.22 1.04 ( 0.22 1.34 ( 0.22 1.30 ( 0.21
Kp,NH4+ ) 0.25
0.66 ( 0.14 0.76 ( 0.09 0.84 ( 0.10 1.18 ( 0.11
Kp,GuH+ ) 0.67
Kp,H+ ) 1.50
(0.60 ( 0.26)a
(0.51 ( 0.12)b
0.80 ( 0.26
0.82 ( 0.09 1.07 ( 0.05
1.74 ( 0.21
1.75 ( 0.23 38
b
-
The Kp,SO4 value calculated from the (GuH)2SO4 STI is anomalous and could result from ion pairing Since HSO4 is the predominant sulfate species (∼90%) in the concentration range of interest, we calculated an estimate for Kp,HSO4- of 0.14. It is plausible that HSO4- would be more surface-accumulated than SO42-, although 0.14 is most likely a lower bound given the dehydration free energy of HSO4- (see Discussion). 2-
Figure 3. Schematic of the partial dehydration of a negatively charged ion that likely occurs as it is transferred from the bulk solution to (A) the air-water surface and (B) the protein surface. Specific ion-protein interactions for weakly hydrated ions are likely to provide the largest discrepancy between the two ion partitioning processes.
Figure 2. Comparison of single-ion partition coefficients Kp,i (Tables 3 and 4) for partitioning of ions between bulk water and the air-water surface with relative Hofmeister rankings of anions and cations for processes which expose protein surface to water (e.g., unfolding).
ions at the air-water surface (Kp ) 1.64). The lone exception, CH3COO- (i.e., Ac-), generally inferred to be moderately excluded from biopolymer surface, has an enhanced surface concentration relative to the bulk (Kp ) 1.30), presumably due to the poor bulk solvation of the methyl group and the favorable contribution of the hydrophobic effect to the thermodynamics of transferring it to the surface. For cations, Figure 2 shows the same rank order in both series but a significant shift in magnitude of partition coefficients. For example, alkali metal cations and GuH+ are widely separated in both series, but the almost complete exclusion of the Group I cations from the airwater surface is in stark contrast to their inferred neutral position in the biopolymer series. 4.3. Insights into the Origin of Hofmeister Effects. What are the molecular implications of the above comparisons of STIderived partition coefficients with the relative and absolute positions of ions in the Hofmeister series of effects on
biopolymer processes? Figure 3 depicts the two relevant processes: the transfer of an ion in bulk water either to the surface of water (path A) or the surface of a protein (path B). Both processes presumably involve partial dehydration of the ion, although the extents may differ. The interactions with the local water at each surface may also vary, but the largest difference between the thermodynamics of the processes in paths A and B (which dictate the Kp of the ion) is expected to originate from any ion-specific interactions with functional groups on the protein (and accompanying changes in hydration), for which there is no air-water surface analogy. The thermodynamics of ion transfer from the bulk to the surface of the aqueous solution are a relevant part of the interpretation of noncoulombic (Hofmeister) salt-protein interactions (and Hofmeister salt effects on protein processes) if the water at both the air and protein surfaces is similar and if (for molecular ions) the same face is presented at either surface. In situations where the interactions of ions with the surface (either air or protein) are not the major determinant of the partition coefficient Kp, the STI-based series deduced here should match the Hofmeister series. For the majority of the anions investigated, the agreement of the STI series with the Hofmeister rankings (Figure 2) suggests that the cost of partial dehydration of the ions may be the dominant factor for ion specificity.
5416 J. Phys. Chem. B, Vol. 111, No. 19, 2007
Figure 4. Linear correlation between Kp,- and e-a∆Godehyd./RT (squares). Free energies of anion dehydration are taken from ref 45. Also shown are anion surface-affinity data (fX-, triangles) from the electrospray ionization mass spectrometry experiments on nanodrops of Colussi and co-workers.20 The normalized affinities are scaled by one for clarity. The parameter a is described in the text and is 0.014 for the surface tension data and 0.056 for the nanodrop data. The filled square (CH3COO-) is considered an “outlier” from this correlation (see text) and was not included in the fit.
Indeed, dehydration was recently proposed by Colussi and co-workers20 as a major source of specificity for partitioning of anions between the bulk and the air-water surface. The authors measured the signal intensities of seven anions in an electrospray ionization mass spectrometer, where the normalized intensities reflect the anion-specific affinity for the surface of fissioning nanodrops. Normalized surface affinities (fX- ≡ IX-/ Itotal) were observed to increase exponentially with increasing anionic radii and decreasing dehydration free energies (i.e., Br< NO3- < I- < SCN- < BF4- < ClO4-). Figure 4 plots both (1 + fX-) and Kp,- (from Table 4) against an exponential function of scaled dehydration free energies (e-a∆G°dehyd./RT; free energies from ref 41). The correlations are quite good. The scaling parameter a,44 which represents most simply the fractional extent of dehydration of the salt ion at the surface, is quite small for both the surface tension (a ) 0.014) and nanodrop (a ) 0.056) data. The source of the nonspecific (dehydration-independent) favorable free energy of partitioning to the surface (ca. -5 kJ from our data) has yet to be identified, although Colussi and co-workers suggest it may originate in the depression of water permittivity in the interfacial region. Numerical values of the nanodrop surface affinities and STIbased anion partition coefficients cannot be readily compared, although ratios of fX- values would be expected to correspond to ratios of Kp,- values if the bulk phase were the same in both types of experiments. However, the STI-based partition coefficient analysis predicts ClO4- to be two times more surfaceaccumulated than Br-, while Colussi and co-workers oberve a 19-fold difference in surface affinity for these anions. A possible explanation for this is the small droplet size (µm-nm radii) in the experiments. In addition to their high degree of curvature, droplets of this size effectively contain little if any “bulk” solution; analysis of surface tension increments with the ion partitioning model applies only to the case where the volume of the surface region, relative to the bulk, is insignificant.
Pegram and Record For the cations, it is striking that no correlation exists between free energies of dehydration and surface-bulk partition coefficients, given the good correlation for the anions. For example, Na+ and Cl- have similar dehydration free energies (375 and 347 kJ, respectively45) but very different partitioning behavior at the air-water surface. Our analysis, which agrees qualitatively with MD calculations,16 predicts that Na+ is completely excluded and Cl- is only moderately excluded. In addition, the large shift between cation STI-derived partition coefficients and inferred placements in the biopolymer Hofmeister series (as shown in Figure 2) suggests that (as compared to the anions) the cations interact relatively more favorably with protein surface than the air-water surface. These two observations suggest that the thermodynamics of transferring cations and anions from the bulk to a surface are dominated by different interactions. Nevertheless, the additivity and counterion independence of the single-ion partition coefficients observed here should allow us to extend this analysis to obtain partition coefficients and ion-partitioning thermodynamics for the analysis of high-concentration (Hofmeister) salt effects on biopolymer stability, and to predict Hofmeister salt effects based on the change in accessible surface area. Acknowledgment. We thank Pete von Hippel for his helpful comments on the manuscript. This work was supported by NIH grant GM47022. Supporting Information Available: The methods of STI determination from the literature data and the composite thermodynamic quantity, (Kp,2 - 1)bσ1 , for each salt. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Hofmeister, F. Arch. Exp. Pathol. Pharmakol. 1888, 24, 247-260. (2) Baldwin, R. L. Biophys. J. 1996, 71, 2056-2063. (3) von Hippel, P. H.; Wong, K. Y. Science 1964, 145, 577-580. (4) Weissenborn, P. K.; Pugh, R. J. Langmuir 1995, 11, 1422-1426. (5) Long, F. A.; McDevit, W. F. Chem. ReV. 1952, 51, 119-169. (6) von Hippel, P. H.; Peticolas, V.; Schack, L.; Karlson, L. Biochemistry 1973, 12, 1256-1264. (7) Timasheff, S. N. AdV. Protein Chem. 1998, 51, 355-432. (8) Record, M. T., Jr.; Zhang, W.; Anderson, C. F. AdV. Protein Chem. 1998, 51, 281-353. (9) Record, M. T., Jr.; Anderson, C. F. Biophys. J. 1995, 68, 786794. (10) Courtenay, E. S.; Capp, M. W.; Anderson, C. F.; Record, M. T., Jr. Biochemistry 2000, 39, 4455-4471. (11) Felitsky, D. J.; Record, M. T., Jr. Biochemistry 2004, 43, 92769288. (12) Frumkin, A. Z. Phys. Chem. 1924, 109, 34-48. (13) Randles, J. E. B. Discuss. Faraday Soc. 1957, 24, 194-199. (14) Liu, D.; Ma, G.; Levering, L. M.; Allen, H. C. J. Phys. Chem. B 2004, 108, 2252-2260. (15) Ghosal, S.; Hemminger, J. C.; Bluhm, H.; Mun, B. S.; Hebenstreit, E. L. D.; Ketteler, G.; Ogletree, D. F.; Requejo, F. G.; Salmeron, M. Science 2005, 307, 563-566. (16) Jungwirth, P.; Tobias, D. J. J. Phys. Chem. B 2001, 105, 1046810472. (17) Mucha, M.; Frigato, T.; Levering, L. M.; Allen, H. C.; Tobias, D. J.; Dang, L. X.; Jungwirth, P. J. Phys. Chem. B 2005, 109, 7617-7623. (18) Petersen, P. B.; Saykally, R. J.; Mucha, M.; Jungwirth, P. J. Phys. Chem. B 2005, 109, 10915-10921. (19) Brown, E. C.; Mucha, M.; Jungwirth, P.; Tobias, D. J. J. Phys. Chem. B 2005, 109, 7934-7940. (20) Cheng, J.; Vecitis, C. D.; Hoffmann, M. R.; Colussi, A. J. J. Phys. Chem. B 2006, 110, 25598-25602. (21) Petersen, P. B.; Saykally, R. J. Annu. ReV. Phys. Chem. 2006, 57, 333-364. (22) Jungwirth, P.; Tobias, D. J. Chem. ReV. 2006, 106, 1259-1281. (23) Pegram, L. M.; Record, M. T., Jr. PNAS 2006, 103, 14278-14281. (24) Spolar, R. S.; Livingstone, J. R.; Record, M. T., Jr. Biochemistry 1992, 31, 3947-3955. (25) Guggenheim, E. A. Trans. Faraday Soc. 1940, 36, 397-412.
Hofmeister Salt Effects on Surface Tension (26) Guggenheim, E. A. In Thermodynamics; North Holland Physics Publishing: Amsterdam, The Netherlands, 1988; pp 293-297. (27) Randles, J. E. B. Phys. Chem. Liq. 1977, 7, 107-179. (28) Jarvis, N. L.; Scheiman, M. A. J. Phys. Chem. 1968, 72, 74-78. (29) For electrolytes with only one dγ/dm2 determination, the error presented is the standard deviation from the line of best fit (weighted by the experimental uncertainty for each surface tension measurement). The standard deviation from the line of best fit was also used for the nonideality data presented in Table 1. (30) Robinson, R. A.; Stokes, R. H. In Electrolyte Solutions; Butterworths: London, 1959; pp 483-490. (31) Bonner, O. D. J. Chem. Thermodyn. 1976, 8, 1167-1172. (32) Bonner, O. D. J. Chem. Eng. Data 1976, 21, 498-499. (33) Goldberg, R. N. J. Phys. Chem. Ref. Data 1981, 10, 671-764. (34) Sarbar, M.; Covington, A. K.; Nuttall, R. L.; Goldberg, R. N. J. Chem. Thermodyn. 1982, 14, 695-702. (35) Kumar, A. Fluid Phase Equilib. 2001, 180, 195-204. (36) As there is slight curvature in these plots, the concentration range of all electrolytes was lowered to the solubility limit of K2SO4 and KClO3 to ensure consistency. (37) From the surface tension increments for an analogous series of uncharged solutes,46 where sucrose is the most surface-excluded solute, one obtains a lower bound value for bσ1 of 0.186 water molecules/Å2. (38) Mason, P. E.; Dempsey, C. E.; Neilson, G. W.; Brady, J. W. J. Phys. Chem. B 2005, 109, 24185-24196. (39) Melander, W.; Horva´ th, C. Arch. Biochem. Biophys. 1977, 183, 200-215. (40) Arakawa, T.; Timasheff, S. N. Biochemistry 1982, 21, 6545-6552. (41) Arakawa, T.; Timasheff, S. N. Biochemistry 1984, 23, 5912-5923. (42) If surface water has the same density as bulk water, then the value of bσ1 determined from Na2SO4 corresponds to an interfacial region with a thickness of approximately 6 Å (or ∼2 water layers). This agrees well with the estimates from molecular dynamics simulations of a surface layer of 4-7 Å.16,47
J. Phys. Chem. B, Vol. 111, No. 19, 2007 5417 (43) Courtenay, E. S.; Capp, M. W.; Record, M. T., Jr. Protein Sci. 2001, 10, 2485-2497. (44) For our data, an estimate for a was determined by plotting lnKp vs -∆G°dehyd./RT; this estimate was further refined by obtaining the best fit for the data plotted in Figure 4, to allow the inclusion of Kp,SO42- ) 0.) For the data of Colussi et al., a ) 0.056 was determined from the slope of lnfX- vs -∆G°dehyd./RT. (45) Marcus, Y. In Ion Properties; Marcel Dekker: New York, 1997; pp 117-135. (46) Auton, M.; Ferreon, A. C. M.; Bolen, D. W. J. Mol. Biol. 2006, 361, 983-992. (47) Gopalakrishnan, S.; Jungwirth, P.; Tobias, D. J.; Allen, H. C. J. Phys. Chem. B 2005, 109, 8861-8872. (48) Washburn, E. W., Ed. In International Critical Tables of Numerical Data, Physics, Chemistry, and Technology, 1st electronic ed.; Knovel: Norwich, NY, 2003; pp 463-466. (49) Weissenborn, P. K.; Pugh, R. J. J. Colloid Interface Sci. 1996, 184, 550-563. (50) Aveyard, R.; Saleem, S. M. J. Chem. Soc., Faraday Trans. I 1976, 22, 1609-1617. (51) Hey, M. J.; Shield, D. W.; Speight, J. M.; Will, M. C. J. Chem. Soc., Faraday Trans. I 1981, 77, 123-128. (52) Johansson, K.; Eriksson, J. C. J. Colloid Interface Sci. 1974, 49, 469-480. (53) Matubayasi, N.; Tsunetomo, K.; Sato, I.; Akizuki, R.; Morishita, T.; Matuzawa, A.; Natsukari, Y. J. Colloid Interface Sci. 2001, 243, 444456. (54) Myhre, C. E. L.; Nielsen, C. J.; Saastad, O. W. J. Chem. Eng. Data 1998, 43, 617-622. (55) Suggitt, R. M.; Aziz, P. M.; Wetmore, F. E. W. J. Am. Chem. Soc. 1949, 71, 676-678. (56) Neros, C. A.; Eversole, W. G. J. Phys. Chem. 1941, 45, 388-395. (57) Livingston, J.; Morgan, R.; Davis, C. E. J. Am. Chem. Soc. 1916, 38, 555-568.
