1460
J. Phys. Chem. B 2009, 113, 1460–1467
Liquid Expanded Monolayers of Lipids As Model Systems to Understand the Anionic Hofmeister Series: 2. Ion Partitioning Is Mostly a Matter of Size E. Leontidis* and A. Aroti Department of Chemistry, UniVersity of Cyprus, Nicosia 1678, Cyprus ReceiVed: October 24, 2008; ReVised Manuscript ReceiVed: NoVember 22, 2008
In the preceding paper of this series [Leontidis, E.; Aroti, A.; Belloni, L. J. Phys. Chem. B 2009, 113, 1447], we considered and modeled the increase of the surface pressure of dipalmitoyl phosphatidylcholine (DPPC) monolayers over electrolyte solutions of various monovalent sodium salts. The experimental results for salts with large, less hydrophilic anions can be successfully described by models treating ionic specificity either as specific partitioning in the interfacial lipid layer or as a result of ion-lipid dispersion interactions. However, the results for salts with more hydrophilic anions, such as chloride and fluoride, cannot be fitted by any of these models, while they clearly demonstrate the existence of a specific sodium-DPPC interaction. In the present paper, we first prove that the experimental results for sodium fluoride (NaF) can be fitted by a model that is based on simultaneous complexation of sodium ions with up to three lipid molecules, as suggested by recent molecular dynamics simulations. We then return to the experimental results of sodium salts with more hydrophobic anions, treated in the preceding paper, and prove that these can be fitted equally well with a complex model, which accounts for both sodium complexation with the lipid head groups and anion partitioning within the lipid monolayers. The partitioning parameters obtained from this more complete model correlate well with several measures of ion specificity, such as ionic volume, von Hippel chromatographic parameters, or viscosity B-coefficients. A model for these partitioning chemical potentials is created based on the competition of cavity and ion hydration terms. The model leads to an excellent correlation of the partitioning chemical potentials with a function of the ionic radius, suggesting that specific anion effects on this lipid model system are mostly a matter of ionic size. Two notable exceptions from this correlation are thiocyanate and acetate ions, the charge distribution of which is not spherically symmetric, so that they are expected to have orientational-dependent interactions with the water-lipid interface. The implications of the present results on ion specificity in general are discussed. 1. Introduction In the preceding paper of this series,1 we examined the possibility of using monolayers of the lipid dipalmitoyl phosphatidylcholine (DPPC) at the air-water interface as a model system to understand specific ion effects at membrane-water interfaces. We have used a variety of models to fit the increase of the surface pressure of DPPC monolayers over electrolyte solutions of various monovalent sodium salts with respect to that over pure water when the monolayer is in the liquid expanded (LE) phase. A major issue addressed by this modeling effort was whether the specific ion effect observed in the surface pressures of DPPC monolayers was a result of localized ion-lipid binding, partitioning of ions between bulk water and the lipid monolayer, or ion-specific dispersion (or other longrange) interactions. DPPC monolayers as a model system allowed us to exclude localized chemical binding as the source of ionic specificity but could not discriminate between the other two possibilities. In the meantime, it was realized that partitioning or dispersion force models, which relied on the exclusion of sodium from the lipid monolayers, could not explain the experimental data obtained with sodium salts of more hydrophilic anions, such as NaF. Many recent computer simulations of lipid bilayers show that Na+ and other cations actually interact with the zwitterionic lipids quite strongly.2-13 Sodium was found * To whom correspondence should be addressed. E-mail: psleon@ ucy.ac.cy.
to create complexes with one to four (or even five!12) lipid molecules simultaneously, interacting quite strongly with the carbonyl groups of the lipids. In this paper, we set out to improve the models of the preceding paper, taking into account the much more complex situation unveiled at lipid bilayer-electrolyte solution interfaces by computer simulations. From the two successful models described in the preceding paper, we have opted not to reconsider that based on dispersion interactions since this field approach would then be coupled to a chemical binding picture in a rather awkward way. It is conceivable that a dispersion force model augmented with a sodium-binding ansatz might work as a fitting algorithm. We believe however that it is more consistent to couple the sodium binding model to the anion partitioning model. In addition, partitioning chemical potentials can form as the basis of a deeper analysis of ion-lipid interface interactions, as will be discussed below. Because we want to place emphasis on the development of our ideas about lipid-electrolyte interactions, we have opted to give a nontraditional structure to this paper, which contains a sequence of theoretical sections alternating with sections discussing the fitting results of the theories. We first show in section 2 that the presence of lipid-sodium clusters at the monolayer surfaces is not sufficient to explain our data, unless we indeed allow the simultaneous presence of a range of sodium-lipid complexes with stoichiometries from 1:1 to 1:3. We then combine in section 3 this more realistic
10.1021/jp809444n CCC: $40.75 2009 American Chemical Society Published on Web 01/14/2009
Ion Partitioning Is Mostly a Matter of Size
J. Phys. Chem. B, Vol. 113, No. 5, 2009 1461
picture of sodium-lipid complexation with the partitioning model for anions and find that the resulting model provides a more general and satisfactory agreement with our data, covering both hydrophilic and chaotropic anions. The fitting exercise generates a fairly robust set of partitioning chemical potentials for anions between the lipid-water interface and the bulk water. These chemical potentials are modeled as sums of solvation and cavity terms, and a striking correlation is found between them and a function of the hydrated ionic size in section 4. We discuss the implications of this correlation in the final section of this paper and give our perspective of both the ion-lipid monolayer work and the current status of our general knowledge about ionic specificity. 2. Salts with Hydrophilic Anions. The Case of NaF 2.1. Theoretical Modeling. In the case of NaF or other salts with very hydrophilic anions, it is reasonable to assume that the anion does not interact with the lipids, and in fact, it avoids the water-lipid interface, in a manner observed in simulations of NaF at the electrolyte solution-air interface.14 The direct charge interaction between fluoride and the positive trimethylamino moiety of the phosphocholine head group is assumed to be weak. The observed surface pressure increase of the DPPC monolayers is here attributed to sodium binding. If it is assumed that each sodium ion binds simultaneously to m lipid molecules (an assumption that has been made in the past when treating (especially divalent) cation binding to lipid bilayers15-17), then the following set of equations applies
mL0 + Na+ f LmNa+
(1)
m[LmNa+] + [L0] ) [L]tot
(2)
( )
σ ) √8kΒTC∞ε0ε sinh
Km )
[LmNa+] [L0]m[Na+]
qeψ0 ) qeNAV[LmNa+] > 0 (3) 2kΒT
) (σ/qeNAV)
([L]tot - mσ/qeNAV)mC∞exp(-qeψ0 /kBT)
(4)
Here, [L0] and [LmNa+] are the surface concentrations of free lipids and sodium complexes with m lipid molecules, [L]tot is the total surface concentration of lipid molecules, and the rest of the notation was explained in the preceding paper. By assigning values to m and Km, one can combine eqs 3 and 4 and solve them for σ and ψ0. The surface pressure increase in this case is given by the Davies equation (eq 4 of the preceding paper1), reproduced below
( )
qeψ0 σ πsalt(aL) - πH2O(aL) ≡ ∆π(aL) ) 2kBT tanh (5) qe 4kBT The model behind eqs 1-4 however is in disagreement with recent computer simulations of DPPC bilayers, which have shown that sodium ions can form complexes with one, two, three, or even four lipid molecules at the DPPC-water interface, and all of these complexes with different stoichiometry coexist.2-13 A stepwise binding model, which corresponds more closely to
this idea, is based on the following chemical equilibria and material and charge balances
Na+(s) + L0 f NaL+
KNa )
[NaL+] ) [Na+]s[L0]
[NaL+] (6) [L0]C∞exp(-qeψ0 /kBT) +
NaL + L f 0
NaL2+
NaL2+ f L0NaL3+
KL )
KL )
[NaL2+] [NaL+][L0] [NaL3+]
[NaL2+][L0]
(7)
(8)
and so forth
[L0] + [NaL+] + 2[NaL2+] + 3[NaL3+] + ... ) [L]tot (9)
( )
σ ) √8kΒTC∞ε0ε sinh
qeψ0 ) qeNAV{[NaL+]+ 2kΒT [NaL2+] + [NaL3+] + ...} (10)
To simplify the calculations, it is assumed that the same constant KL applies to all complexation steps beyond the first. Assigning values to KNa and KL, one can solve the model equations numerically. After obtaining σ and ψ0, one can calculate the surface pressure increments using again the Davies equation. The disadvantage of this, physically more correct, model is that by necessity, it contains a minimum of two adjustable parameters (KNa and KL). 2.2. Fitting Results for NaF. In Figure 1a-c, we show the results of fitting the model of eqs 1-4 to the DPPC/NaF experimental data for the cases of sodium ions binding to 1, 2, or 3 lipid molecules. In all cases, the fits are poor, and the qualitative trends of the fits are wrong. From the preceding paper,1 it was known that the case m ) 1 (1:1 lipid-Na+ binding) does not fit the NaF data since the theoretical curves have a much higher slope than that observed in the experiments (see Figure 11 of the preceding paper1). For m ) 2, we observe a similar behavior, while for m ) 3, the theoretical curves may actually pass through a pronounced maximum and then decrease, in disagreement with the experiment. The model of eqs 6-10 was applied using a simple stepwise error minimization technique, in which a value for KNa is set first and KL is varied until the minimum least-squares difference from the experimental data is found. The procedure was repeated with a new value of KNa and so on, until a global error minimum was found, which was in fact rather shallow and broad. With the current NaF experimental data, the minimum fitting error was found for KNa ) (6.6 ( 0.4) × 10-4 m3 mol-1 and KL ) (3.2 ( 0.2) × 106 m2 mol-1. Figure 2 shows that this model fits the NaF experiment quite well, reproducing the relative insensibility of the DPPC surface pressure at 85 Å2 per lipid molecule to the NaF concentration in the subphase. Figure 3 shows the relative fraction of interfacial sodium ions bound to one, two, or three lipid molecules at each NaF concentration. The results are in very good agreement with the recent computer
1462 J. Phys. Chem. B, Vol. 113, No. 5, 2009
Leontidis and Aroti
Figure 3. Percentage of 1:1, 2:1, 3:1, and 4:1 lipid-Na+ complexes at the interface, as calculated by the complexation partitioning model with UF f ∞ (black bars). Shown are also analogous results from recent computer simulations7,12 (gray and dark gray bars).
or even four lipids.2-8,11,12 Since in our model we did not allow binding of four lipids to one sodium, it is gratifying to observe that the sum of the lipid fractions of the L3Na+ and L4Na+ complexes from the simulations11,12 is very close to the present model estimate for the L3Na+ complexes. Table 1 contains these interfacial sodium fractions and the corresponding values observed in computer simulations to illustrate the general agreement of these approaches.
Figure 1. Attempts to fit the surface pressure increment of DPPC monolayers at 295 K and 85 Å2 per molecule in the presence of NaF, using chemical binding models and assuming (a) 1:1 lipid-Na+ complexation, (b) 2:1 lipid-Na+ complexation, and (c) 3:1 lipid-Na+ complexation. The solid lines are theoretical fits with various binding constants.
3. Salts with Fewer Hydrophilic Anions. A More Complete Penetration Model 3.1. Theoretical Considerations. Since our experimental results for NaF and NaCl1 and also the previous modeling results and recent computer simulations support multiple sodium binding to the DPPC molecules, we now reconsider the partitioning model of the previous paper, which was quite successful for salts with more chaotropic anions. We remind the reader that that model was based on the idea of complete sodium exclusion from the lipid monolayer. The obvious question arises if a composite model, which considers sodium binding but also treats anion-lipid interactions in the spirit of the preceding work, might be similarly successful in fitting all of the experimental results presented before. The equations of the partitioning model require very few modifications to treat this more complex situation. When sodium binds to the lipids at x ) δ, the surface pressure increase of the lipid monolayer is given by
{
∆π ) φδσ + kBT
∫0∞
1 - εε0(∇ψ)2 2 qeψ qeψ + U+ +C∞ 1 + exp + kBΤ kBΤ
Figure 2. Fit of the surface pressure increment of DPPC monolayers at 295 K and 85 Å2 per molecule in the presence of NaF using a complexation model which assumes multiple lipid-Na+ complexation with coexistence of various complexes at the interface. Shown are results for total fluoride exclusion (solid line) and partial fluoride exclusion with UF ) +2.0kBT (dashed line) and UF ) +1.5kBT (dotted line).
simulations, which predict that a small fraction of sodium ions bind to a single lipid molecule while most bind to two or three
(
1-
[(
) (
) (
)
)
]
qeψ qeψ - Uexp - 2Θ(x - δ) kBΤ kBΤ
}
dx (11)
Equation 11 is, in fact, a small modification of eq (A.II.3) of the preceding paper. The only new term is the surface charge density term φδσ, which usually appears when a mathematically sharp monolayer of surface charge is introduced at the boundary between two phases. Equation 11 applies to the more general
Ion Partitioning Is Mostly a Matter of Size
J. Phys. Chem. B, Vol. 113, No. 5, 2009 1463
TABLE 1: Fractions of Lipid Molecules That Are Free or Bound in LNa+, L2Na+, L3Na+, and L4Na+ complexes
present work Lee et al.a Gurtovenkob a
free lipids
lipids in LNa+ complexes
lipids in L2Na+ complexes
lipids in L3Na+ complexes
lipids in L4Na+ complexes
0.581 0.605 0.560
0.009 0.005 0.010
0.064 0.040 0.070
0.346 0.150 0.250
0.200 0.110
From ref 12. b From ref 7.
TABLE 2: The Two NaF Parameter Sets That Have Been Used to Compute All the Anionic Partitioning Parameters That Appear in Table 3 total fluoride exclusion from lipid monolayer partial F- exclusion
case that the cation can also penetrate into the lipid layer. By setting U+ f +∞, we disallow this possibility within the framework of this model. Equations 6-10 hold again with ψ0 replaced by ψδ and allow the calculation of the surface charge density due to sodium adsorption at x ) δ. 3.2. Fitting Results for All Sodium Salts. The initial step toward applying the complex anion partitioning-sodium binding model is to return to the NaF results discussed in section 2. We assign various values to the partitioning chemical potential of F-, and for each value, we compute the optimal values of KNa and KL that provide the best fit to the NaF results. In Figure 2, it can be observed that reasonable fits can be obtained by assuming positive values of UF from infinity to +2.0kBT. For UF values that are smaller or even negative, the results cannot be properly fitted by any reasonable values of the sodium binding parameters, and the best-fit curves have a clearly steeper slope than the one observed experimentally. There is something rather special about the value UF ) +2kBT; this is just enough exclusion to counterbalance a positive electrostatic potential of roughly +50 mV (generated by the adsorption of sodium ions) and hinder fluoride buildup at the interface. The conclusion is that even such a multiparameter model suggests exclusion of fluoride ions from the lipid-water interface. We have applied this full-blown model to the data of all other sodium salts for two sets of parameter values (UF, KNa, and KL), shown in Table 2, which represent the possible range of these parameters. The first set assumes complete exclusion of F- from the lipid monolayer (UF f +∞), while the second set assumes the less stringent exclusion dictated by UF ) +2.0kBT. As we shall see below, it makes little difference which set of NaF parameters will be chosen. Keeping these parameters fixed, we have fitted the experimental results of the DPPC monolayers with U_ as the only adjustable parameter in each case, exactly as in the preceding paper. We have thus obtained new optimal values for the partitioning chemical potentials of all anions in the lipid monolayers. In Figure 4a, we see that this new and more complex model actually provides a much improved fit for the NaCl results (dashed line) over the original model of the preceding paper (solid line), which ignores sodium binding. The new model reproduces the slow rise of ∆π at lower salt concentrations. This suggests that the apparent anomaly of the NaCl data is indeed a “signature” of sodium-lipid interactions. In Figure 4b, we see that for SCN-, a more chaotropic anion, assuming no sodium binding or multiple sodium binding to the lipids makes little difference to the quality of the fit. Table 3 contains a list of all of the partitioning parameters obtained using the two NaF parameter sets and allows a comparison with the partitioning chemical potentials reported in the previous paper. It appears that accounting for sodium binding actually leads to a decrease of the anionic partitioning parameters by 0.5-0.7kBT.
UF/kBT
KNa/m3 mol-1
+∞ +2.0
6.6 × 10-4 1.4 × 10-4
KL/m2 mol-1 3.2 × 106 8.0 × 106
Reliable PF6- numbers could not be obtained with the complex models because the experimental data for this salt were obtained at only two relatively low concentrations (less than 50 mM), where lipid-Na+ interactions play a very important role and do not allow reliable extraction of U_ values. The PF6- values in Table 3 are derived from extrapolation from the lines of Figure 5, in which the partitioning parameters of the two complex models are plotted versus the partitioning parameters of the sodium-exclusion case, derived in the previous paper.1 This figure shows that the U_ values in the presence and absence of sodium binding are highly correlated. All U_ values have good correlations with typical measures of ion specificity, such as the viscosity B-coefficients,18 the von Hippel chromatographic partitioning parameters,19 or the transfer free energy from water to organic solvents.20 These correlations are provided in Figure S1 as Supporting Information.
Figure 4. Fits of the complexation partitioning model with UF f ∞ (dashed line) and the simple partitioning model of the previous paper1 (solid line) for DPPC monolayers at 295 K and 85 Å2 per molecule in the presence of (a) NaCl and (b) NaSCN.
1464 J. Phys. Chem. B, Vol. 113, No. 5, 2009
Leontidis and Aroti
TABLE 3: Sets of Partitioning Chemical Potentials for 11 Anions Obtained with the Two NaF Parameter Sets of Table 2 and Also with the Assumption of Total Sodium Exclusion of the Preceding Paper ion
no Na+ binding
FClCH3COOBrNO3ClO3IBF4ClO4SCNPF6-
-0.70 ( 0.10 -1.40 ( 0.05 -1.78 ( 0.05 -2.50 ( 0.05 -2.90 ( 0.05 -3.15 ( 0.10 -3.30 ( 0.10 -3.70 ( 0.05 -4.23 ( 0.10 -4.50 ( 0.10
4. A Model for the Partitioning Chemical Potentials The correlations in Figure S1 (Supporting Information) suggest that the partitioning parameters U_ may be generally useful as “Hofmeister parameters” to quantify anionic specificity. It is necessary to understand the nature of these parameters more deeply. A crude, preliminary step in this direction is outlined in Figure 6, in which we try to transfer to the water-DPPC interface what has recently become known about the interaction of ions with the water-air interface.21-25 In the partitioning picture, we model the lipid-water interface as an “interphase”, that is, as a different solvent. U_ is then a partitioning chemical potential or free-energy difference when an anion is transferred from bulk water to this interface. This free-energy difference consists of at least two terms, a contribution from short-range ion-solvent hard-core repulsions and a contribution from attractive interactions, in a phenomenological thermodynamic approach, which can be traced back to the ideas of Eley on solvation splitting,26 the solvophobic theory of Sinanoglu,27 and the WCA theory of the statistical mechanics of liquids.28,29 This type of approach has been used so often in the past in biophysics and physical chemistry that it has become a standard way to describe solvation in a solvent or the change of solvation upon transfer into another solvent.29-37 The short-range attraction between ions and water dipoles is assumed to provide a positive contribution that tends to keep the ion in the water bulk because the strong ion-dipole interactions are partly lost when an ion goes into the lipid layer. We assume that this term can be modeled by a classical Born-type approximation, although much literature exists on elaborate modifications of this solvation term. Since the actual dielectric environment of an ion immersed between the lipid head groups or in contact with the hydrophobic part of the lipid monolayer is not known, we simply write
∆Gwflip ≈ pol
(
)
(ziqe)2 1 1 1 ∝ (12) 8πε0(Ri + Rw) εlip εw Ri+Rw
This is a term inversely proportional to the hydrated ion radius involving the supposedly tightly held first solvation shell. The assumption is that most of the difference will occur beyond the first solvation shell and that all ions feel “a similar” dielectric environment within the monolayer. The charge is omitted in the end since, in this work, we consider only monovalent anions. The second contribution to the transfer free energy involves van der Waals and hard-core interactions. Water-water interactions may play an important role in this term. A negative cavity contribution will arise when the ions are transferred toward the lipid interface, and there is a free-energy gain when some water molecules of hydration are liberated and can return to the bulk.
Na+ binding, UF ) 2.0kBT 2.0 -0.20 ( 0.05 -0.95 ( 0.10 -1.45 ( 0.05 -1.85 ( 0.05 -2.50 ( 0.10 -2.95 ( 0.10 -2.80 ( 0.05 -3.20 ( 0.05 -3.90 ( 0.10 -4.05
Na+ binding, UF ) +∞ -0.02 ( 0.05 -0.75 ( 0.10 -1.30 ( 0.05 -1.65 ( 0.05 -2.35 ( 0.05 -2.90 ( 0.05 -2.65 ( 0.10 -3.05 ( 0.05 -3.80 ( 0.05 -3.95
This contribution is modeled, following the established precedent, as proportional to solute area or volume.29-37
∆Gwflip ≈ (γlip - γw)4π(Ri + Rw)2 ∝ -(Ri + Rw)2 cav
(13)
where again it is assumed that all ions sample roughly the same cavity environment in bulk water and at the interface. This picture is of course oversimplified. The “wall” of the ionic cavity inside of the lipid layer will consist partly of water molecules and partly of lipid segments and must be quite inhomogeneous, while the cavity radius in the lipid layer may well be different from that in bulk water. Terms that do not depend on ionic size might also be present in the free-energy difference as well as highly ion-specific terms for ions that have special interactions with the lipids. In the picture that we draw here, most of the anion-lipid interaction can be explained as a composite size effect, in the sense that the size affects the direct ion-solvent interactions and the solvent-solvent interactions in the periphery of the ionic cavity. Ignoring for the moment all other complications, we make the ansatz
U- ) ∆Gwflip + ∆Gwflip ∝ f(Ri + Rw) ) pol cav -(Ri + Rw)2 +
c′ (14) Ri + Rw
We anticipate that a plot of U_ versus the right-hand side of eq 14 should be roughly linear. The constant c′ can be fixed by assuming that for Cl-, which is thought to be very weakly drawn toward the lipid interface,2-14 the solvation and cavity terms roughly cancel each other. Using thermochemical ionic radii collected by Roobottom et al.,38 we obtain a value of c′ equal to 29.2 Å3. Figure 7 contains a plot of U_ versus f(Ri + Rw), according to eq 14. The correlation is remarkable given the simplicity of the model, superior to correlations of the U_s with the free energy of hydration, the polarizability, or the viscosity B-coefficient. Ionic specificity does exist however, beyond that covered by eq 14, as exemplified by the two important outliers, acetate and thiocyanate (we believe that the deviation of BF4may be mostly due to a nonprecise value of the ionic radius38). Acetate and thiocyanate possess significant dipole moments and have large asymmetric polarizabilities.39 In addition, these ions have a distinctly “linear” shape, and acetate has a hydrophobic methyl group, which creates a local hydration asymmetry for the ion in bulk water40 and would force it to adopt a very special orientation in the lipid interface considered here. Alternatively, the position of acetate in the correlation might be related to its hardness. While this is definitely possible, acetate associates
Ion Partitioning Is Mostly a Matter of Size
J. Phys. Chem. B, Vol. 113, No. 5, 2009 1465
Figure 7. Plot of the partitioning parameter, U_, versus the function of ionic size defined in eq 14. The set of U_ obtained for UF ) +2.0kBT is chosen here to highlight the position of F- in the correlation. Figure 5. Correlation between the ionic partitioning chemical potentials obtained from the partitioning model that does not assume lipid-Na+ binding and those obtained from the complexation partitioning model for UF f ∞ (empty circles) and UF ) +2.0kBT (filled circles).
ClO4- is often thought to be more chaotropic than SCN- and should have a more negative U_ value. Several experimental investigations of DPPC bilayers using many different techniques (chromatography, electrophoresis, NMR, EPR) concluded that the interaction of SCN- with phosphatidylcholine interfaces is in fact “weaker” than that of ClO4-.43-48 On the other hand, SCN- has been reported to strongly affect DPPC bilayers, leading to lipid interdigitation.49 From Figure 7, it appears very plausible that the extra specificity of some ions, which appears to go beyond the effect of size (or charge density), may be the result of a nonspherical charge distribution. 5. Closing Discussion and Outlook: Toward a Model of Ionic Specificity
Figure 6. Scheme for ion partitioning between bulk water and the surface lipid monolayer, showing (a) the partial dehydration of the ion upon transfer to the monolayer and (b) the aqueous cavity reduction when water molecules are replaced by lipid segments.
with the interface more strongly than the softer chloride, although the choline group is soft and the opposite would be expected. In addition, the surface pressure (at 85 Å2) versus concentration plot for this ion is compatible with a partitioning rather than with a binding scenario (see preceding paper1). Acetate lies above the straight line in Figure 7, implying a somewhat weaker association with the interface than that expected on the basis of size only. Thus, we favor the solvation asymmetry rather than the binding picture in the present system. SCN- was also recently reported to prefer an orientation either parallel to the surface at the air-water interface41 or tilted with respect to the surface by roughly 40° (see the vibrational sum frequency generation (VSFG) spectroscopy results of ref 42).
We feel that the results obtained in this work shed light on the old problem of the lyotropic or Hofmeister series. The relative success of the crude model of the previous section for the partitioning parameters implies that the chaotropic side of the Hofmeister series can be understood to a large extent by considering the indirect effect of ionic size on ion-water and water-water interactions close to the ionic cavities. This picture is in broad agreement with the computer simulation work of Karlstro¨m50 and Dill.51 Hydrophilic anions are expelled from the lipid surfaces, and the monolayer model system then makes “visible” the interaction between the lipids and sodium. This model system is therefore capable of discriminating between the two ions of an electrolyte. Actually, this discrimination goes deeper; cations interact with the lipids mostly through binding interactions, while anions mostly partition within available space in the lipid monolayer. Thus, this interface appears to establish a different interaction mechanism toward the two ions of an electrolyte. An important experimental question must be asked here: Is NaCl really the “close-to-neutral” salt, for which the anion and the cation effects roughly cancel each other? As mentioned earlier, recent computer simulation work has revealed that K+ ions interact much more weakly than Na+ with the head groups of zwitterionic lipid bilayers.7,8,11,12 A perfectly “neutral” salt for the DPPC monolayer might probably be KF; both ions, being highly hydrophilic, would hardly interact with the lipid layer. Our results for Cl- are consistent with the ideas of Collins and others,52,53 who consider Cl- to be the dividing anion between the kosmotropes and the chaotropes; its partition coefficient toward the lipid monolayer of the present work is very close to unity (U_ very close to zero). What is the precise environment of anions at the water-lipid interface? The formalism of the partitioning model is based on the picture of Figure 8a, in which a uniform monolayer
1466 J. Phys. Chem. B, Vol. 113, No. 5, 2009
Leontidis and Aroti
Figure 8. Two pictures for the structure of the lipid-water interface; (a) a two-phase system with two homogeneous phases, as assumed by the present complexation partitioning model; (b) a more open monolayer structure with pockets of low lipid density, as suggested by recent computer simulation results.2-13
environment acts as a different solvent for the ions. The recent computer simulation work however appears to be more compatible with the “Swiss-cheese-like” interfacial picture of Figure 8b, where it is shown that the multiple binding of sodium on the lipid head groups and the resulting lipid clustering leads to density inhomogeneities of the LE phase. Anions may in fact use these “comfortable”, less dense pockets at the interface as solubilization sites. However, this molecular detail can only be unveiled by detailed, targeted molecular dynamics simulations or by advanced spectroscopic methods, such as VSFG or SHG. Our investigation has also revealed that very special ionic effects that defy simple explanations always exist, such as the effect of NaSCN on DPPC monolayers. It is interesting that the partitioning model appears to fit the NaSCN data quite well (Figure 4b), but the U_ parameter obtained is probably too large by at least 1kBT. Really specific interactions are unavoidable in most interfaces, but the main message of the present work is that some order is possible “in most systems in most cases”. Outliers in particular systems can be easily detected by plotting results of completely different experiments against each other, as we did in Figure S1 (Supporting Information) to show that SCN- is indeed behaving in an irregular way in our monolayer model system. There is still some way to go to completely understand specific salt effects. The combination of cations and anions and difficult interfaces, such as protein surfaces, creates some hard challenges, which have just started to be addressed by computer simulation.54-56 Furthermore, the strong specific salt effects on protein denaturation57 and enzymatic reactions58,59 or on systems without clear interfaces (a good example is the salting-out effect of organic molecules60,61) are still rather poorly understood. However, working with good model systems has indeed led to progress and revealed unexpected analogies between widely different systems. Take, for example, the present monolayer model, in which sodium ions bind to the lipids through a complexation mechanism with carbonyl groups while large polarizable anions can in fact partition within the “open” lipid monolayer and enhance the sodium effect. Similar conclusions were recently drawn by Dzubiella,62 who examined salt effects on R-helix formation by oligopeptides. This coincidence shows that we are indeed approaching a much better understanding of specific salt effects on biological and physicochemical systems. What we believe that we have learned from the now extensive
Figure 9. Ionic parameters that directly determine ion specificity and derived ionic properties, which play important roles in several ionspecific phenomena: A chart of interconnections.
work at the air-water interface14,21,22 and the present investigation of lipid monolayers is summarized in Figure 9. According to this scheme, ionic size is synonymous with charge density, when ions of the same charge are considered. Charge density is the major ionic property in processes where ion pairing plays an important role. This is the case in many biological systems, as pointed out by Collins52,53 and supported by the recent peptide and protein computer simulations by Dzubiella62 and Jungwirth.54,55 In addition, ionic size determines the cavity size of the solute, which is always important, especially when ion pair formation is not dominant. There are situations where both effects of the size must be considered, as in the case of our monolayer work or in the case of the interaction of salts with PNIPAM reported by the group of Cremer.63,64 Finally, the size is the major determinant of the molecular polarizability, which fine-tunes intermolecular interactions, and may play unexpectedly important roles in a situation of a near-balance of forces, such as that encountered for ions at the surface of electrolyte solutions.14,21,22 Besides size and charge density, the actual shape of the charge distribution may be expected to play a role at interfaces, where space is no more isotropic. Such effects are seen (we believe) in our monolayer work (see the behavior of NaSCN and
Ion Partitioning Is Mostly a Matter of Size CH3COONa in Figure 7) and may constitute a major class of exceptions or anomalies of experimental lyotropic series reported in many situations. Acknowledgment. The corresponding author would like to thank the University of Cyprus for a generous three-year internal research grant (2005-2007) that enabled him to carry out this work. We would also like to thank Prof. Luc Belloni (CEA Saclay, France) for discussions leading to eq 11. We also thank Prof. Pavel Jungwirth and Robert Vacha of the Institute of Organic Chemistry of the Academy of Sciences of the Czech Republic for useful discussions about their molecular dynamics simulations of lipid bilayers. Supporting Information Available: A figure showing correlations of the partitioning chemical potentials with several measures of ionic specificity, as described in the main text of the paper. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Leontidis, E.; Aroti, A.; Belloni, L. J. Phys. Chem. B 2009, 113, 1447. (2) Pandit, S. A.; Bostick, D.; Berkowitz, M. L. Biophys. J. 2003, 84, 3743. (3) Pandit, S. A.; Bostick, D.; Berkowitz, M. L. Biophys. J. 2003, 85, 3120. (4) Berkowitz, M. L.; Bostick, D. L.; Pandit, S. Chem. ReV. 2006, 106, 1527. (5) Bo¨ckmann, R. A.; Hac, A.; Heimburg, T.; Grubmu¨ller, H. Biophys. J. 2003, 85, 1647. (6) Bo¨ckmann, R. A.; Grubmu¨ller, H. Angew. Chem., Int. Ed. 2004, 43, 1021. (7) Gurtovenko, A. A. J. Chem. Phys. 2005, 122, 244902. (8) Gurtovenko, A. A.; Vattulainen, I. J. Phys. Chem. B 2008, 112, 1953. (9) Lopez, C. F.; Nielsen, S. O.; Klein, M. L.; Moore, P. B. J. Phys. Chem. B 2004, 108, 6603. (10) Sachs, J. N.; Nanda, H.; Petrache, H. I.; Woolf, T. B. Biophys. J. 2004, 86, 3772. (11) Cordomi, A.; Edholm, O.; Perez, J. J. J. Phys. Chem. B 2008, 112, 1397. (12) Lee, S.-J.; Song, Y.; Baker, A. A. Biophys. J. 2008, 94, 3565. (13) Kaznessis, Y. N.; Kim, S.; Larson, R. G. Biophys. J. 2002, 82, 1731. (14) Jungwirth, P.; Tobias, D. J. J. Phys. Chem. B 2001, 105, 10468. (15) Grasdalen, H.; Eriksson, L. E. G.; Westman, J.; Ehrenberg, A. Biochim. Biophys. Acta 1977, 469, 151. (16) Hauser, H.; Hinckley, C. C.; Krebbs, J.; Levine, B. A.; Phillips, M. C.; Williams, R. J. P. Biochim. Biophys. Acta 1977, 468, 364. (17) Altenbach, C.; Seelig, J. Biochemistry 1984, 23, 3913. (18) Jenkins, H. D. B.; Marcus, Y. Chem. ReV. 1995, 95, 2695. (19) von Hippel, P. H.; Peticolas, V.; Schack, L.; Karlson, L. Biochemistry 1973, 12, 1256. (20) Czapkiewicz, J.; Czapkiewicz-Tutaj, B. J. Chem. Soc., Faraday Trans. I 1980, 76, 1663. (21) Jungwirth, P.; Tobias, D. J. Chem. ReV. 2006, 106, 1259. (22) Archontis, G.; Leontidis, E.; Andreou, G. J. Phys. Chem. B 2005, 109, 17957. (23) Pegram, L. M.; Record, M. T. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 14278.
J. Phys. Chem. B, Vol. 113, No. 5, 2009 1467 (24) Pegram, L. M.; Record, M. T. J. Phys. Chem. B 2007, 111, 5411. (25) Eggimann, B. L.; Siepmann, J. I. J. Phys. Chem. C 2008, 112, 210. (26) Ben-Naim, A. Molecular Theory of Solutions; Oxford University Press: Oxford, U.K., 2006. (27) Melander, W.; Horvath, C. Arch. Biochem. Biophys. 1977, 183, 200. (28) Weeks, J. D.; Chandler, D.; Andersen, H. C. J. Chem. Phys. 1971, 54, 5237. (29) Pratt, L. R.; Chandler, D. J. Chem. Phys. 1977, 67, 3683. (30) Paterson, Y.; Nemethy, G.; Scheraga, H. A. J. Solution Chem. 1982, 11, 831. (31) Rashin, A. A. J. Phys. Chem. 1990, 94, 1725. (32) Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T. J. Am. Chem. Soc. 1990, 112, 6127. (33) Honig, B.; Sharp, K.; Yang, A.-S. J. Phys. Chem. 1993, 97, 1101. (34) Hassan, S. A.; Guarnieri, F.; Mehler, E. L. J. Phys. Chem. B 2000, 104, 6478. (35) Gu, W.; Rahi, S. J.; Helms, V. J. Phys. Chem. B 2004, 108, 5806. (36) Jorgensen, W. L.; Ulmschneider, J. P.; Tirado-Rives, J. J. Phys. Chem. B 2004, 108, 16264. (37) Zhu, J.; Alexov, E.; Honig, B. J. Phys. Chem. B 2005, 109, 3008. (38) Roobottom, H. K.; Jenkins, D. B.; Passmore, J.; Glasser, L. J. Chem. Educ. 1999, 76, 1570. (39) Serr, A.; Netz, R. R. Int. J. Quantum Chem. 2006, 106, 2960. (40) Robertson, W. H.; Diken, E. G.; Johnson, M. A. Science 2003, 301, 320. (41) Petersen, P. B.; Saykally, R. J.; Mucha, M.; Jungwirth, P. J. Phys. Chem. B 2005, 109, 10915. (42) Viswanath, P.; Motschmann, H. J. Phys. Chem. C 2007, 111, 4484. (43) McLaughlin, S.; Bruder, A.; Chen, S.; Moser, C. Biochim. Biophys. Acta 1975, 394, 304–313. (44) Hauser, H.; Hinckley, C. C.; Krebs, J.; Levine, B. A.; Phillips, M. C.; Williams, R. J. P. Biochim. Biophys. Acta 1977, 468, 364–377. (45) Tatulian, S. A. Biochim. Biophys. Acta 1983, 736, 189–195. (46) Rydall, J. R.; Macdonald, P. M. Biochemistry 1992, 31, 1092–1099. (47) Bartucci, R.; Belsito, S.; Sportelli, L. Chem. Phys. Lipids 1996, 79, 171–180. (48) Hu, W.; Haddad, P. R.; Hasebe, K.; Mori, M.; Tanaka, K.; Ohno, M.; Kamo, N. Biophys. J. 2002, 83, 3351–3356. (49) Cunningham, B. A.; Lis, L. J. Biochim. Biophys. Acta 1986, 861, 237–242. (50) Karlstro¨m, G.; Hagberg, D. J. Phys. Chem. B 2002, 106, 11585. (51) Hribar, B.; Southall, N. T.; Vlachy, V.; Dill, K. A. J. Am. Chem. Soc. 2002, 124, 12302. (52) Collins, K. D. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 5553–5557. (53) Collins, K. D.; Neilson, G. W.; Enderby, J. E. Biophys. Chem. 2007, 128, 95. (54) Vrbka, L.; Vondrasek, J.; Jagoda-Cwiklik, B.; Vacha, R.; Jungwirth, P. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 15440. (55) Vrbka, L.; Jungwirth, P.; Bauduin, P.; Touraud, D.; Kunz, W. J. Phys. Chem. B 2006, 110, 7036. (56) Thomas, A. S.; Elcock, A. H. J. Am. Chem. Soc. 2006, 128, 7796. (57) Carrea, G.; Pasta, P.; Vecchio, G. Biochim. Biophys. Acta 1984, 784, 16. (58) Pinna, M. C.; Salis, A.; Monduzzi, M.; Ninham, B. W. J. Phys. Chem. B 2005, 109, 5406. (59) Salis, A.; Bilanicˇova`, D.; Ninham, B. W.; Monduzzi, M. J. Phys. Chem. B 2007, 111, 1149. (60) Long, F. A.; McDevit, W. F. Chem. ReV. 1952, 51, 119. (61) Grover, P. K.; Ryall, R. L. Chem. ReV. 2005, 105, 1. (62) Dzubiella, J. J. Am. Chem. Soc. 2008, 130, 14000. (63) Zhang, Y.; Furyk, S.; Bergbreiter, D. E.; Cremer, P. S. J. Am. Chem. Soc. 2005, 127, 14505. (64) Chen, X.; Yang, T.; Kataoka, S.; Cremer, P. S. J. Am. Chem. Soc. 2007, 129, 12272.
JP809444N
