Liquid-liquid phase separation of aqueous lysozyme solutions: effects

Theory for the Liquid–Liquid Phase Separation in Aqueous Antibody Solutions. Miha KastelicVojko Vlachy. The Journal of Physical Chemistry B 2018 122...
1 downloads 0 Views 676KB Size
2140

J . Phys. Chem. 1990, 94, 2140-2144

What causes the reversible photodegradation? The mere reversibility suggests a photophysical rather than photochemical effect. There are two major possible causes: (1) excited-state (singlet or triplet) conformational changes, e.g., some freezing-in of an excimer or dark trap configuration (linking together different chains or different parts of the same chain); (2) local heating (as suggested by one referee), possibly at dark trap or excimer sites, with some similar local morphological effects. We prefer either one of these speculations over the assumption of a reversible photochemical effect due to oxygen.

and the exciton decay kinetics for the dilute blends. The presence of photosensitive dark traps may lead to such anomalous degradation behavior, via excited-state conformational changes or local heating. However, no photochemical impurities are detected and no signs of discoloration are observed. Despite the fact that the heterogeneity exponents are highly dependent on the irradiation time, a consistent concentration dependence is still observed. Further investigations are necessary for understanding this kind of reversible photodegradation behavior and its morphological implications.

Summary and Conclusions Reversible photophysical degradation of PI VN/PMMA blends is observed via excimer spectra and exciton decay curves. The degradation occurs at 77 K upon prolonged irradiation. Thermal annealing can restore the original singlet excimer to monomer ratio

Acknowledgment. We thank Prof. Albert F. Yee for the use of the DSC equipment. This work was supported by the National Science Foundation under Grant DMR-83039 19. Registry No. P I V N , 25135-12-0; PMMA, 9011-14-7

Liquid-Liquid Phase Separation of Aqueous Lysozyme Solutlons: Effects of pH and Salt Identity Victor G. Taratuta, Andreas Holschbach,+George M. Thurston, Daniel Blankschtein,: and George B. Benedek* Department of Physics and Center for Materials Science and Engineering and Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 (Received: July 10, 1989)

We have studied the properties of liquid-liquid phase separation in aqueous solutions of the protein lysozyme. The system under investigation was a lysozyme solution in sodium phosphate buffer of high ionic strength and of pH far below the isoelectric pH of lysozyme. Liquid-liquid phase separation was marked by opacification of the solution as its temperature was lowered, at fixed concentration, to a well-defined value called Tcloud.Below this temperature the solutions separated macroscopically into two coexisting liquid phases. We have conducted measurements of Tcloudas a function of ionic strength of the sodium phosphate buffer, its pH, and the identity and concentration of added salts. We have observed the following: (1) Tcloyd was insensitive to changes in the total ionic strength of the sodium phosphate buffer within the range of the (rather high) ionic strengths studied (0.300-0.600 M), while the buffer pH was held constant. (2) However, when salts (NaCI, NaBr, KC1, KBr) were added to the sodium phosphate buffer keeping the total ionic strength of the buffer constant (by decreasing the was observed to increase roughly linearly with salt concentration. The contents of sodium phosphate correspondingly), Tcloud identity of the salt, at constant total ionic strength and pH of the buffer, had a marked effect on the magnitude of that increase. (3) Finally, Tcloudincreased approximately linearly with pH between pH values of 5.8 and 8.0 while the total ionic strength of the buffer was held constant. Our data suggest that the pH and ionic strength alone are not sufficient to describe the conditions for liquid-liquid phase separation in our solutions. Each ionic species present in the solution must be individually assayed for its effect on phase separation. A thermodynamic Gibbs free energy for hard spheres with an attraction is used to model the phase separation phenomenon in aqueous lysozyme solutions.

Introduction The existence of temperature-induced liquid-liquid phase separation in an aqueous protein-water solution was first reported, for a solution of the protein lysozyme in salt water, by Ishimoto and Tanaka.l They measured a concentration-dependent cloud-point temperature curve, which they identified as a coexistence curve for a liquid-liquid phase equilibrium in these solutions. Until now, such a coexistence curve has been measured only for one other2 protein, bovine y,,-crystallin. It is currently believed3that the liquid-liquid phase separation in a protein-water solution is closely related to the phenomenon of cold cataracts observed in the eye lenses of certain mammalian species, including calves, rats, and mice. In both cases, upon cooling the system under investigation at a constant concentration, a reversible opacification is observed when a well-defined temperature, Tcloudr is reached. A coexistence curve for liquid-liquid phase separation

in a binary protein-water solution describes the dependence of T+,d on protein concentration. At temperatures T > Tclwd,the mixture exists as a stable one-phase solution, while at T < Tcloud the solution separates into two coexisting liquid phases, which differ in protein concentration. If the correspondence between the liquid-liquid phase separation and cold cataracts is a valid one, then understanding the mechanisms that determine the location and shape of a coexistence curve of a binary protein-water solution is an essential first step toward understanding the biologically important phenomenon of cold cataracts. Lysozyme, while not a component of the lens, can serve as a useful model system in assessing the influence of important physicochemical factors, such as protein surface charge and ionic environment, on the location of the coexistence curve. However, the existence of a liquid-liquid phase separation phenomenon in a lysozyme-water mixture has ( 1 ) Ishimoto, C.; Tanaka, T. Phys. Reu.

*To whom correspondence should be addressed. Present address: Wetzlarerstrasse 81, 6300 Giessen/Klein Linden, West Germany. *Department of Chemical Engineering.

0022-3654/90/2094-2140$02.50/0

Leu. 1977, 39, 474. (2) Thomson, J. A.; Schurtenberger, P.; Thurston, G. M.; Benedek, G. B. Proc. Nail. Acad. Sci. U.S.A. 1987, 84, 7019. (3) Siezen, R. J.; Fisch, J. R.; Slingsby, C.; Benedek, G. B. Proc. Nail. Acad. Sci. U.S.A. 1985, 82, 1071.

0 1990 American Chemical Society

Liquid-Liquid Phase Separation of Lysozyme Solutions been the subject of an interesting controversy. Phillies reexamined4 the system used by Ishimoto and Tanaka.' He argued that the reversible opacity of lysozyme-water solutions was due to scattering from crystals rather than from thermally driven concentration fluctuations and was, in fact, a cryoprecipitation effect, not a critical point phenomenon. Clearly, the origin of this controversy lay in the fact that equilibrium coexistence of two liquid phases had not yet been observed in binary lysozyme-water solutions. In the course of our experiments on lysozyme-water solutions, we did observe, under a variety of solution conditions, a coexistence of two clear bulk liquid phases separated by a sharp meniscus. These two liquid phases are characterized by different concentrations of protein. Formation of solid particles, reported by Ishimoto and Tanaka' and by Phillies? was also detected in our solutions. We observed, however, that such formation could be minimized if extreme care was exercised to inhibit the processes of oxidation and denaturation of protein. Among the measures found effective were the use of thoroughly degassed buffers, centrifugation of the solutions to remove solid matter, and storing specimens under dry nitrogen. These measures are similar to those used in establishing a coexistence curve for the liquid-liquid phase separation in aqueous solutions of bovine y,I-crystallin,2 in which the formation of solid particles, in addition to the liquid-liquid phase separation, was also observed. In cases when the formation of macroscopic solid particles was allowed to proceed, their presence was made obvious through anomalous fluctuations in the intensity of light scattered from the solution. The noise signal corresponding to the intensity of the scattered light was monitored on a strip chart recorder. This noise was characterized by the presence of large spikes in the intensity due to the solid particles suspended in the solution, superimposed on typical noise produced by thermally driven concentration fluctuations in a crystal-free solution. We characterized the liquid-liquid phase separation by two independent measurements. One was the measurement of the opacification temperature of the solution, Tcloud, as a function of protein Concentration. In the other measurement, the concentrations of the two coexisting bulk liquid phases were determined at fixed temperatures. Both experimental procedures are described below. The results produced by these two different methods were found to be in good agreement with one another. In our view, these findings establish a close correspondence between the opacification and the formation of two coexisting bulk liquid phases in our solutions and confirm the existence of a liquid-liquid phase separation phenomenon in aqueous lysozyme mixtures reported by Ishimoto and Tanaka.' In the course of our investigation of liquid-liquid phase separation in lysozyme-water solutions, we first established the coexistence curves at the following three pH values of the buffer: 6.0, 6.5, and 7.0. Sodium phosphate buffer of constant ionic strength of 0.6 M was used in all three cases. The coexistence curves are shown in Figure 1. The point of maximum temperature of the coexistence curve is known as the critical point, which is characterized by the critical temperature, T,, and critical concentration, c,. Interestingly, it is observed that the value of the critical concentration, c,, of lysozyme in an aqueous solution was nearly identical in all the solutions studied. Our measurements indicated it to be 230 f 10 mg of protein per milliliter of solution. It is noteworthy that a quite similar c, value of 240 f 10 mg/mL was reported for aqueous solutions of bovine yI,-crystallin.2 Moreover, the shape and the width of the lysozyme solution coexistence curves did not change significantly with pH. On the other hand, the critical temperature for phase separation, T,, was extremely sensitive to changes in pH. Thus, the effect of changing the pH of the buffer has been mainly to translate the coexistence curve parallel to the temperature axis. We, therefore, postulate that a change in Tcloud measured at any protein concentration corresponds to an equivalent change in the theoretically important quantity, T,, as the buffer pH is changed. We, furthermore, report our investigations of the effects of ionic strength, pH, and identity (4) Phillies, G. D. J. Phys. Reu. Lett. 1985, 55, 1341.

The Journal of Physical Chemistry, Vol. 94, No. 5, 1990 2141 I I 7.5

I

I

tt

I

I

'

1

I

I

1

5.0

-5,0t / -7.5c

I 0

I

I IO0

1

I 200

I

l

300

l

400

l

l

500

I ysozyme concentration ( m g / m l )

Figure 1. Coexistence curves of the aqueous lysozyme solutions in sodium phosphate buffer: pH 7.0 (circles), pH 6.5 (squares), and pH 6.0 (triangles). The corresponding solid symbols represent the results of the temperature quenching experiments in which lysozyme concentrations in two coexisting bulk liquid phases are directly measured at a constant

temperature. The solid lines are drawn to help guide the eye. and concentration of added salts on liquid-liquid phase separation temperature in buffered lysozyme-water solutions a t a single protein concentration, which we have conveniently chosen to be 90 mg/mL. In addition, we present a thermodynamic Gibbs free energy for hard spheres with an attraction to model phase separation in aqueous lysozyme mixtures.

Materials and Methods Chicken egg white lysozyme is a tightly packed globular protein with molecular weight of 14 400 and is approximately ellipsoidal5 in shape with dimensions 45 X 30 X 30 A3. Its isoelectric pH is equal to 1 1.2, and at pH 7.0 it carries a net positive charge of about 7.5 electronic charges.6 The samples of lysozyme were prepared by using the following procedure. Lysozyme powder (Sigma No. 7001, twice crystallized; Sigma Chemical, St. Louis, MO) was dissolved in a sodium phosphate buffer of the appropriate pH and ionic strength as described below. The buffer was thoroughly degassed prior to the preparation of the sample. At room temperature, with stirring, the protein easily went into solution until a concentration of 30-50 mg/mL was reached. The solution was then centrifuged, to remove any undissolved material and air bubbles, and concentrated to 90 mg/mL under dry nitrogen atmosphere by use of an Amicon ultrafiltration unit equipped with a YM-10 membrane. Lysozyme concentrations were determined by UV absorption spectroscopy from a measurement of OD280after lOOOX dilution in buffer using a specific absorption coefficient = 26.4.s The solution was then dialyzed exhaustively against sodium phosphate buffer whose precise composition was determined by the solution conditions chosen for the specific measurement. The specific solution conditions were established as follows: ( i ) Dependence of Tcloudon the Sodium Phosphate Buffer Concentration. Sodium phosphate buffer consisting of a mixture of monobasic and dibasic sodium phosphate salts was used. The pH of the buffer was kept constant a t 6.8. The concentrations of the buffer salts were varied in such a way as to keep their relative proportions constant while changing their total concentration in the range corresponding to a total ionic strength between 0.3 and 0.6 M. ( i i ) Dependence of Tcloud on the Identity and Concentration of Added Salt. The buffer consisted of sodium phosphate with an addition of one of the four salts: NaCI, NaBr, KCI, or KBr. A pH of 6.8 was maintained. The total ionic strength, resulting from ( 5 ) Dickerson, I. E.; Geis, I. The Structure and Action of Proteinr; Harper and Row: New York, 1969. (6) Tanford, C.; Wagner, M . L. J . Am. Chem. SOC.1954, 76, 3331.

2142

The Journal of Physical Chemistry, Vol. 94, No. 5, 1990

both sodium phosphate and added salt, was maintained at 0.6 M. (iii) Dependence of Tcloud on p H . No added salts were used in this experiment. The pH of the sodium phosphate buffer was varied from 5.8 to 8.0 by properly balancing the composition of the buffer which consisted of monobasic and dibasic sodium phosphate salts. The ionic strength of the buffer was kept constant at 0.6 M. Tcloud was measured by laser light scattering. The solution, contained in a cylindrical scattering cell having a light path of 4 mm, was placed in a light scattering spectrometer equipped with a temperature-controlled sample holder. The temperature of the sample was controlled to an accuracy of f O . l OC and a stability of f0.05 "C. An argon ion laser (Spectra Physics Model 164 at the wavelength of 488 nm; Spectra Physics, Santa Clara, CA) was used for these measurements, although a low-power He-Ne laser would be sufficient to produce an adequate level of the scattered intensity in the solutions studied. The temperature was initially set well within the single-phase region and then lowered in small steps. The light scattered at an angle of 90" was detected with a photomultiplier [EM1 Model 9863A; EMI, Hayes, Middlesex, U.K.]. At each temperature setting, the scattered intensity was monitored on a chart recorder to verify that the solution had reached thermal equilibrium before the scattered intensity was recorded, and the next temperature decrement was attempted. In our solutions, the relaxation time needed varied between 5 and 20 min, depending on the proximity to Tcloud. As Tcloud was approached, the intensity of the scattered light increased dramatically. At a certain temperature, Tdlsappcar, the transmitted beam would disappear completely and the solution would become visibly cloudy due to the presence of domains of protein-rich phase in the matrix of protein-poor phase for the case of solutions having c < cc. These domains create strong inhomogeneities in the index of refraction of the solution, thus leading to very strong multiple scattering. At the onset of clouding, before macroscopic, gravity-driven separation into two liquid phases could commence, the temperature was increased until the solution clarified, the transmitted beam reappeared, and the scattered intensity was restored to its value observed prior to clouding. It was possible to reverse the phase separation due to the very large viscosity of the solution at the clouding point which made the macroscopic, gravity-driven phase separation proceed extremely slowly. The minimum temperature at which the solution clarified was recorded as Treappcar, and Tcloud was determined as the average between T d i , p r and, Tmppea'.The difference between Tdivlppar and Treappar is re ected in the indeterminacy in Tcloud. In our measurements this indeterminacy was estimated to be f0.1 "C. For the direct measurement of the concentrations of the two coexisting liquid phases, a solution of a given concentration was quenched in a centrifuge (Beckman L5-50B, SWSO.1Ti rotor) at a temperature below T,. After centrifuging for 2-3 h at a speed of 2000-4000 rpm, the solution would separate macroscopically into two coexisting clear liquid phases separated by a sharp meniscus. The concentration of each of the two phases was measured, and the results were compared to those obtained from the light scattering measurements. As can be seen from Figure 1, the two methods produce consistent results. The use of the centrifuge to separate the lysozyme-water solution macroscopically into two coexisting liquid phases was necessary due to the extremely large viscosity of the solution below the critical temperature. In solutions quenched without the use of a centrifuge, the macroscopic separation would typically take several days to reach an advanced stage, during which time the protein in the solution was susceptible to crystal formation.

Taratuta et al.

i

t

i

t- - 2

1 -3

c i-

-4

i1 1

,

I I 6 0.2 0.3 0.4 0 . 5 ionic s t r e n g t h o f b u f f e r ( M )

0

0. I

0.6

Figure 2. Dependence of Tcloud on the ionic strength of sodium phosphate buffer, pH 6.8.

0

"I

bU 5

0 0

: 0

0 0

0

1

I

i I

0

1

Experimental Results

Our experimental findings are illustrated in Figures 1, 2, 3, and 4. Figure I has been described above. In Figure 2 we show the on the ionic strength of the sodium phosphate dependence of T-&c,l buffer in the range from 0.3 to 0.6 M and at constant pH of 6.8. As can be seen from Figure 2, no change in Tcloud with ionic strength of the sodium phosphate buffer was observed within the range of ionic strengths studied.

(7) Verwey, E. J. W.; Overbeek, J . Th. G. Theory of the Stability of Lyophobic Colloids; Elsevier: Amsterdam, 1948.

Liquid-Liquid Phase Separation of Lysozyme Solutions

2.5

Figure 4 shows the dependence of T&,ud on buffer pH in the pH range between 5.8 and 8.0 and at total buffer ionic strength of 0.6 M. Within this range, a substantial increase in the cloud-point temperature was observed. This increase occurs between -6 and 0 OC and is approximately linear in pH.

-

0-

-

0

-

0

0

u

a

-2.5

0

0

0

t"

0

- 5.0-

0 0

0

-7.5

The Journal of Physical Chemistry, Vol. 94, No. 5, I990 2143

-

- 10.0 5

6

7

0

9

PH

Figure 4. Dependence of Tcbudon pH in sodium phosphate buffer, at the ionic strength of 0.6 M.

The effects of four different added salts on T&,,d are illustrated in Figure 3. For the four salts studied, we have observed an approximately linear increase in Tcloud with the concentration of the added salt. In these experiments we kept the pH constant at 6.8 and also kept the total ionic strength of the buffer (due to both sodium phosphate and the added salts) constant at 0.6 M. Note that the concentration range of the added salts used in this experiment was from 0 to 0.55 M for NaCl and NaBr salts and from 0 to 0.20 M for KCI and KBr salts. The difference in the maximum salt concentration used was necessitated by the greater tendency of the potassium salts, in comparison to the sodium salts, to promote crystallization of lysozyme in our solutions. The findings illustrated in Figures 2 and 3 demonstrate very clearly that the ionic strenth alone cannot be used to adequately characterize the physical factors that control the phase separation temperature. As can be seen from Figure 3, the chlorine salts, NaCl and KCI, produce similar changes in Tcloud. Likewise, the effects of the bromine salt, NaBr and KBr, on Tcloud are similar to one another in magnitude but are different from those produced by the chlorine salts. Thus, cation substitution (Na+ by K') produces little change in Tcioud, while the substitution of one anion by another (CI- by Br-) leads to a substantial change in Tcloud. The relatively weak dependence of Tcloud on the identity of the cation, as contrasted to that of the anion, might be associated with the fact that lysozyme at pH around neutral has a net positive charge and, therefore, tends to repel positive ions. Negatively charged ions, on the other hand, can approach a positvely charged lysozyme molecule more easily than the positively charged cations can and, therefore, can be expected to strongly interact with the charged groups on the protein surface. Such interactions, through contributing to the relative strength of proteinsolvent interactions and of screened protein-protein interactions, would thereby affect Tcloud. Note, however, that it has recently been demonstrated by Lesins and Ruckenstein8 that there can exist attractive electrostatic interactions between a protein and an adsorbent, even when the net charge of the protein is of the same sign as that of the adsorbent. In order to further investigate the effects of ions on the phase separation in protein solutions, it might be instructive to undertake a similar study above the isoelectric pH (PI) of the protein, where the protein would have a net negative charge. However, lysozyme is not a good candidate for this experiment due to its extreme PI of 11.2, at which the protein itself becomes unstable. (8) Lesins, V.; Ruckenstein, E. Colloid Polym. Sci. 1988, 266, 1187.

Discussion One of the interesting observations of the present study is the insensitivity of the critical concentration for phase separation in lysozyme-water solutions to pH changes in the experimental range. Furthermore, it is interesting to observe that another protein, bovine yII-crystallin, was shown2 to have a similar critical concentration (compare c, = 230 f 10 mg/mL for lysozyme and cc = 240 f 10 mg/mL for yI,-crystallin). The fact that two proteins, so different in their structure and function, exhibit very similar critical concentrations for phase separation suggests that the details of the surface structure of a protein may not be very important in determining this parameter. Instead, an effective hard-core repulsion, which ignores the fine scale details of the surface structure of the protein, may play a dominant role in setting the value of c,. In order to understand the properties of the measured liquidliquid phase boundaries and their dependence on solution conditions, it is useful to construct a Gibbs free energy model for a binary protein-water mixture. We adopt a model capable of predicting that the critical concentration is in the observed range. This model is similar to a thermodynamic theory originally developed to analyze phase separation phenomena in two-component colloidal solutions by Jansen et aL9 Its mathematical structure and associated parameters describe, at a phenomenological level, those water-mediated interactions between individual proteins that determine the properties of the observed phase separation. We adopt as a model a solution consisting of monodisperse hard spheres (proteins), each of volume flp, occupying a volume fraction 4, dispersed in a continuous medium (water). These assumptions are reasonable since lysozyme is a tightly packed globular protein whose volume is much larger than that of a water molecule. The Gibbs free energy, G, per unit volume, V, of such a solution is modeled as

where kBT is Boltzmann's constant times absolute temperature. The standard part of the Gibbs free energy, Go, which is linear in 4, reflects (i) the free energy change of the solution when a solvent molecule is added to pure solvent and (ii) the corresponding change when a single protein molecule is introduced into pure solvent. The second term represents the entropy of mixing of hard spheres, using the well-known Carnahan-StarlingIo expression. The third term describes pairwise attractions between the hard spheres in the mean-field approximation and is quadratic in the protein volume fraction. The quantity C is a dimensionless parameter which reflects water-mediated attractions between the hard spheres. Its magnitude reflects a suitable combination of protein-protein, protein-water, and water-water interactions. A full description of this model will be published elsewhere." We regard this form of a Gibbs free energy as a useful starting point in the description of aqueous binary mixtures of globular proteins which exhibit phase separation. Using the thermodynamic stability conditions at the critical point,12 we can compute the numerical values of 4 and C a t the critical temperature. These are found to be 4, = 0.13 and C, = 10.6. In comparison, the experimental value of the critical volume fraction in our solutions and in aqueous solutions of bovine yII-crystallin2was close to +c = 0.17. (9) Jansen, J . W.; deKruif, C. G.; Vrij, A. Chem. Phys. Lett. 1984, 107, 456. (10) Carnahan, N . F.; Starling, K. E. J . Chem. Phys. 1969, 51, 635. ( 1 1) Blankschtein, D. To be published. (12) Guggenheim, E. A. Thermodynamics; North-Holland: New York, 1961.

2144

The Journal of Physical Chemistry, Vol. 94, No. 5, 1990

In order to understand how pH and buffer composition influence the location of the coexistence curve in protein-water solutions, it would be necessary to have a molecular theory for the interaction parameter C. The required theory should take accurate account of the electrostatic interactions between different components of the mixture. The nature of such interactions is now the subject of a great deal of experimental and theoretical research for systems even much simpler than our own (see, for example, ref 7 and 8). In the absence of such a molecular theory, the specific effects of the salt identity on T, are difficult to account for in quantitative terms owing to the rich variety of phenomena that occur on adding ions to a protein solution. CI- and Br- ions in solution are characterized by different degrees of hydration. Generally, the smaller the bare radius of an ion, the more hydrated it is and the greater its hydrated radius.I3 The difference in hydrated sizes between chloride and bromide ions may provide a clue to understanding the quantitative difference in their effects on T,, as they yield different electrostatic interaction energies between that ion and the charged groups on the surface of the protein. Since such interactions are very important in influencing the overall protein-solvent interactions as well as the overall solvent-mediated protein-protein interactions, they must be carefully considered in investigating the effects of salts on T, in protein-water solutions. Similarly, different mechanisms could be invoked to explain the present pH data. We outline two possibilities below. First, the magnitude of the electrostatic interactions between protein surface residues and the surrounding solvent will be affected by the solution pH, since the surface charge distribution of the protein varies with pH. Charged amino acid residues on the surface of the protein molecule will have electrical double layers centered around them. As the charge of an individual surface residue changes, there will be a free energy change due to the formation or dissolution of the corresponding electrical double layer, with the magnitude of this free energy change being mediated by the proximity of other protein molecules. Since the proximity of another protein would mean that the surface residue is more completely surrounded by a low dielectric constant medium than when solvent alone is nearby, we expect that the double layer formation free energy will, in fact, favor the adjacency of protein and solvent. To the extent that this change in the free energy does favor such adjacency, it will tend to lower T,. It is known14 that the only lysozyme surface residue titrated in the experimental pH range is the single histidine residue 15. This residue changes its charge from 0 to +1 in lowering pH from 8.0 to 6.0. Corresponding to this, phase separation temperature decreases (see Figure 4) as the adjacency between the histidine residue and solvent becomes more favorable electrostatically. Thus, our experimental data are qualitatively consistent with the above interpretation. On the other hand, in lowering the pH from 8.0 to 6.0, we decrease the concentration of the monovalent (proton donor) (1 3) Edsall. J. T.; Wyman, J. Biophysicol Chemistry; Academic: New York, 1958. (14) Roxby, R.; Tanford, C. Biochemistry 1971, 10, 3348.

Taratuta et al. phosphate ion, H2P04-, by a factor of 3 and, at the same time, increase the concentration of the divalent (proton acceptor) phosphate ion, HP04*-, by a factor of 35. Thus, their relative concentration is changed by a factor of approximately 100 as buffer pH is changed by two pH units. These ions may each have a different effect on TCld of the solution. In contrast, the apparent independence of cloud-point temperature on concentration of these ions in Figure 2 could be related to the fact that their concentrations changed only by a factor of 2 in that experiment and that their relative concentrations did not change at all. In our experiments on the effects of NaCI, NaBr, KCI, and KBr salts, we have had an opportunity to observe just how important the ion identity can be in influencing the phase separation temperature in aqueous lysozyme solutions. Therefore, changing the concentrations of H2P04-and HP042-ions could produce an alternative mechanism for the change in Tcloud, perhaps, similar to that occurring in changing the concentrations of C1- and Br- ions. Thus, one of the mechanisms proposed involves change in the energy of the electrical double layer as the surface charge of the protein changes with buffer pH, while the other one deals with change in the ionic atmospheric with pH. Both of these effects could independently contribute to the change in the phase separation temperature. The experimental data reported in this paper raise a number of interesting issues, both theoretical and experimental, which have to be explored in order to gain a better understanding of the important phenomenon of phase separation in protein-water solutions and its dependence on the solution conditions. Our data clearly show that the buffer pH and ionic strength alone are not sufficient to fully describe the conditions for a liquid-liquid phase separation in aqueous lysozyme solutions. Instead, the specific ionic composition of the solution plays an important role in determining the magnitude of Tcloud. To our knowledge, there does not yet exist a satisfactory molecular theory for the microscopic origin of the interaction parameter C, which determines the temperature for phase separation in protein-water solutions. We regard the present paper as a first step in a systematic study of the solution factors that can influence the location of the liquid-liquid phase boundary in such protein-water solutions. We believe that such experimental and theoretical information is needed to help in the rational design of chemical agents capable of suppressing the phenomenon of phase separation which occurs in cold cataracts. Acknowledgment. The authors thank Dr. C. R. Middaugh and Dr. J. A. Thomson for helpful discussions of the experiments. This work was supported in part by the National Science Foundation under Grant 84-18718-DMR and by the National Institutes of Health under Grant EYO-5 127-07. Daniel Blankschtein acknowledges the donors of the Petroleum Research Fund, administered by the American Chemical Society, for partial support of this research and is grateful for the support by the TexacoMangelsdorf Career Development Professorship at M.I.T. Registry No. Cl-, 16887-00-6; Br-, 24959-67-9; lysozyme, 9001-63-2.