Thermodynamic analysis of the effect of concentrated salts on protein interaction with hydrophobic and polysaccharide columns

ARCHIVES OF BIOCHEMISTRY AND BIOPHYSICS Vol. 248, No. 1, July, pp. 101-105, 1986

Thermodynamic Analysis of the Effect of Concentrated Salts on Protein Interaction with Hydrophobic and Polysaccharide Columns TSUTOMU

ARAKAWA

Amgen, 1900 Oak Terrace Lane, Thousand Oaks, Calgornia Received December

91320

13,1985, and in revised form March 18,1986

An attempt was made to explain the effect of concentrated salts on protein interaction with hydrophobic columns. From the previously observed results of preferential interactions for salting-out salts with proteins, it was shown that the free energy of the protein is increased by addition of the salts and this unfavorable free energy is smaller for the proteins bound to the columns because of their smaller surface area exposed to solvent; i.e., the bound form of the proteins is thermodynamically more stable. This explains the protein binding to the hydrophobic columns at high salt concentrations and the elution by decreasing the salt concentration. The unfavorable interaction free energy was greater for Na2S04 or (NH&SO4 than for NaCl, which explains the stronger effect of the former salts on the protein binding to the columns. The observed favorable interaction between KSCN or guanidine hydrochloride and the proteins explains the decreasing effect of these salts on the protein binding to the hydrophobic columns. 0 1986 Academic

Press, Inc.

Hydrophobic interaction chromatography is one of the most powerful techniques for protein purification (l-6). In this technique, proteins are bound to hydrophobic columns such as phenyl-agarose in concentrated solutions of strong salting-out salts and eluted from the columns by decreasing the salt concentration, since the protein binding to these columns is weak without concentrated salts. It is believed that the protein binding in the hydrophobic interaction chromatography occurs through hydrophobic interaction; indeed, a thermodynamic study has shown that the interaction is driven by an entropy change, suggesting a major contribution of hydrophobic interaction (7). However, the effect of the salts on the protein binding to these columns has never been fully explained, although it has been suggested that the mechanism of protein binding to the hydrophobic columns in the concentrated salts is the same as for protein salting-out (2,6). However, the mechanism of protein salting-out has also been poorly understood

until recently. Melander and Horvath (8) have shown that the increased surface tension of water by the salts favors proteinprotein contacts in protein salting-out and protein-ligand contact in the hydrophobic interaction chromatography. However, some salts such as MgCls do not follow the surface tension effect, since they increase the surface tension of water as does (NH&SO4 (8), but increase the protein solubility (9) and do not enhance the protein binding to hydrophobic columns as much as expected from the surface tension increment (lo), in contrast to (NH&S04. Moreover, MgClz even weakened the protein binding (11). Extensive protein-solvent interaction studies by Arakawa and Timasheff (12,13) have indicated a strong correlation between the preferential interaction of proteins with salts and the effect of the salts on the protein solubility. These studies could also explain why certain salts do not follow the surface tension effect. The purpose of this paper is to extend the results of the preferential interaction mea101

0003-9861/86 $3.00 Copyright All rights

0 1986 by Academic Press, Inc. of reproduction in any form reserved.

102

TSUTOMU

surements to the analysis of the salt effect on the protein binding in hydrophobic interaction chromatography. It is also shown that protein interaction with polysaccharide columns, such as Sepharose 4B and cellulose, can be understood in terms of the same theory.

ARAKAWA

difference for the ligand. Equations [3] and [4] indicate that protein binding to a hydrophobic column is a function of the intrinsic affinity of proteins with the column and of the solvent interactions with the proteins and the column. For a given column and protein, the protein binding should depend only on the preferential interactions. THEORY Now, let us examine whether the preferential interaction of the proteins with Let us describe the binding of the protein the salts can explain the effect of the salts to the column by in hydrophobic interaction chromatography. First, the observed protein binding LtPrLP PI and elution in the hydrophobic column where L is the functional group (defined with such salts as phosphate, Na2S04, here as ligand) of the column, P is the pro(NH&S04, and NaCl must be explained. tein, and LP is the protein bound to the These salts are largely excluded from the column. The equilibrium constant, K,, in immediate domain of the proteins in conthe absence of salt, may be given by centrated solutions; i.e., the proteins are hydrated (9, 12-15). The -RT In K,., = AG, PI preferentially thermodynamic consequence of such inwhere AG, is the intrinsic free energy teractions is that the interactions between change of the protein binding to the column the salts and the proteins are thermodyand determined by an affinity between the namically unfavorable (9,16); i.e., addition protein and the ligand in the absence of of the salts increases the free energy of the salts. In the presence of salts, K is related proteins (n’Gs,l) as shown in Fig. 1. In adto the free energy change of the binding, dition, it was observed that this free energy increase is proportional to the surface area AG, by -RT In K = AG = AG, + AG,,,. [3] of the protein molecules (9, 12, 13). This leads to the consequence that the free enAG,,, is the free energy change resulting ergy increase due to the interactions with from the preferential interaction of the the salts would be smaller for the protein salts with the proteins and ligand and is molecules bound to the column ligand, since given by part of the surface of the protein molecules, when bound to the column, becomes not AG,,, = ‘rG sol - ?‘Go, + pGoJ accessible for the solvent and hence the unfavorable interaction with the salts is = CbLGso,+ bPGso,)- ?%c,, + tPGso,) reduced. This means that bPGSO1 < fPGsol, as [41 shown in Fig. 1. It is likely that a similar where LPGsolis the free energy due to the type of salt interactions also occurs with preferential interactions of the salts for the the column ligand and, hence, bLGsOl< %Gsol ligand-protein complex and &LGLIoland for the same reason as described for the protein. As a consequence, AG,,, becomes ff Go, are the respective free energies when they are unbound (free state). The paramnegative and K in Eq. [3] increases when eter LPGsol may be decomposed into two the salts are added. The elution of the proterms, the interaction free energies for the teins with decreasing salt concentration ligand (bLGs,,l)and the protein (bPGs,,l)in the can be explained from the fact that the unligand-protein complex. Therefore, bPGsol favorable interaction with the salts de- fpGsOlis the interaction free energy difcreases and hence fpGsOl(and possibly fLGs,1) ference for the protein between the free becomes less positive as the salt concenstate and a protein-ligand complex. bLGsol tration is decreased. This situation is de- fLG801is the corresponding free energy pitted in Fig. 1, indicating that AG,,, be-

PROTEIN

“0

salt

low salt

INTERACTION

high salt

SALT CONCENTRATION

FIG. 1. Schematic illustration of free energy changes due to preferential interactions as a function of salt concentration.

comes less negative at lower salt concentration, which results in weaker protein binding. Thus, the protein binding and elution are determined by a balance between a negative value of AG,,, and a positive value of AG,; i.e., for a given protein and column, the protein binding occurs at higher salt concentrations which render AG < 0 and the protein elution occurs at the lower concentrations which render AG > 0. This gives a clear explanation of how the salting-out salts function in the hydrophobic interaction chromatography. It is worth noting that if AG, is strongly negative, as may be true of highly hydrophobic ligands such as octyl groups, protein elution would not occur when the salt concentration was decreased. Next, the preferential interaction results should be able to explain why the protein binding is stronger in (NH&S04, Na2S04, and phosphate than in NaCl(17-19). It has been observed that the free energy increase of the proteins by addition of the salts is much greater for Na2S04 [and possibly phosphate and (NHI)SOJ than for NaCl(9, 12-15). Thus, Na2S04 [and phosphate and (NH&SO41 should lead to a more positive value of @GSOIand hence a more negative value of AG,,, than NaCl at the same salt concentration, which should result in a

WITH

COLUMNS

103

stronger effect of the former three salts on the protein binding. More rigorous examination may come from the results of hydrophobic interaction chromatography and preferential interaction for MgClz and MgS04, since their opposite effect on protein solubility has been one of the most difficult to explain. For example, the surface tension effect predicts both to be salting-out salts (8), while MgCl, has a salting-in effect (9). It was shown that MgC12 can elute a hydrophobic protein which is retained in the hydrophobic column even after a low-salt wash (11). In this case, AG, should be negative for this protein since it can bind to the column without the salting-out salts and AG,,, for MgC12 should be positive to induce the protein elution. It was shown that MgCl, exhibits no protein preferential hydration and binds to the proteins (13). Since the MgClz binding should be greater for the free protein molecules, the free state of the protein molecules is thermodynamically more favorable than the bound state, i.e., bPGsol> fPGsol and hence AG,,, > 0. This agrees with what is expected from the observed protein elution by addition of MgClz in the hydrophobic column (11). For MgSO*, I am not aware of hydrophobic interaction chromatography experiments performed with this salt. Therefore, a simple experiment was carried out by loading 10 mg of myoglobin in 40 mM Tris, pH 7.5, containing 2.5 M MgS04 onto a phenyl-Sepharose column (3 ml) equilibrated with the same buffer. All the loaded protein was retained on the column and could be eluted from the column by decreasing MgS04 concentration. This clearly shows that MgS04 enhances the protein binding to the column. The preferential interaction measurements for MgS04 showed the preferential exclusion from the proteins as for Na2S04 (9, 13), which should enhance the protein binding for the same reason as described for other salting-out salts. On the other hand, myoglobin in 40 mM Tris, pH 7.5, containing 3 M MgC12, flowed through a phenyl-Sepharose column equilibrated with the same salt solution, indicating no enhancement by 3 M MgCl, of protein binding to the hydrophobic col-

104

TSUTOMU

ARAKAWA

umn. The explanation described for the umns, such as Sepharose 4B and cellulose MgClz effect can also be applied to the pro- (23-25), just as in the hydrophobic columns. tein elution induced by addition of KSCN Although the polysaccharides have a poor guanidine hydrochloride (20), since these tential to form hydrogen bonds with the salts showed no preferential hydration or proteins, they are also known to be hydroeven a preferential binding to the proteins phobic (26-28). Since the protein binding (12, 21; T. Arakawa and S. N. Timasheff, behavior to the polysaccharide columns is similar to that to hydrophobic columns, the unpublished results), just like MgClz. Next, let us consider the mechanism of polysaccharide columns may be regarded fractionation of different proteins. This as less nonpolar hydrophobic columns, and will occur for the following two reasons. the mechanism of their functioning should First, proteins have different hydropho- be similar to that of hydrophobic columns. bicities and, hence, affinities to hydrophobic ligands. Those with a lower affinity ACKNOWLEDGMENTS (more positive value of AG,) would require higher salt concentration for binding, as I would like to thank Dr. P. Bui for correcting Enexpected from Eq. [3]. Second, the differences between proteins in chemical nature, glish, and Mrs. J. Bennett for typing the manuscript. such as polarity, and in physical nature, REFERENCES such as surface area, should lead to differences in their interactions with salts and, 1. GEREN, C. R., MAGEE S. C., AND EBNER, K. E. (1976) as a consequence, in different values of Arch Biochern. Biophys. 172,149-155. ‘rGSO1and AG,,. For those (possibly less 2. RIMERMAN, R. A., AND HATFIELD, G. W. (1973) hydrophobic) proteins which have weaker Science 182,1268-1270. interactions with the salts and less nega3. COMINGO, D. E., MIGNEL, A. G., AND LESSER, B. H. tive values of AGsol,binding would require (1979) Biochim. Biophys. Acta 563,253-260. higher salt concentrations. Thus, these two 4. DOELLGAST, G. J., AND FISHMAN, W. H. (1974) factors should lead to a stronger binding B&hem. J. 141,103-112. for more hydrophobic proteins at a given 5. RAHIMI-LARIDJANI, I., GRIMMINGER, H., AND LINsalt concentration. This is in general what GENS, F. (1973) FEBS I&t. 30,185-187. has been observed in hydrophobic inter6. MEMOLI, V. A., AND DOELLGAST, G. J. (1975) action chromatography (6,22). Biochem Bivphyx Rex Commun 66,1011-1016. 7. GELSEMA, W. J., BRANDTS, P. M., DELIGNY, C. L., It has been suggested that hydrophobic THEEUWES, G. M., AND ROOZEN, A. M. P. (1984) interaction chromatography has an adJ. Chrmnutogr. 295,13-29. vantage over reverse-phase chromatogra8. MELANDER, W., AND HORVATH, C. (1977) Arch phy (22) because the latter system employs B&hem.. Biophys. 183,200-215. organic solvents and low pH, which would 9. ARAKAWA, T., AND TIMASHEFF, S. N. (1985) in induce protein denaturation. On the other Methods in Enzymology (Wyckoff, H. W., Hirs, hand, hydrophobic interaction chromatogC. H. W., and Timasheff, S. N., eds.), Vol. 114, raphy employs salting-out salts, which pp. 49-77, Academic Press, Orlando, Fla. have been shown to even stabilize the pro- 10. RAYMOND, J., AZANZA, J.-L., AND FOTSO, M. (1981) .I Chnwnatogr. 212,199-209. teins (12,13). However, care has to be exercised when the proteins are eluted by 11. MILLER, N. T., AND KARGER, B. L. (1985) J. ChrcF matogr. 326,45-61. KSCN or MgClz, since it was suggested that these salts could destabilize or denature 12. ARAKAWA, T., AND TIMASHEFF, S. N. (1982) Bie chemtit~ 21,6545-6552. the proteins (12,13; T. Arakawa and S. N. T., AND TIMASHEFF, S. N. (1984) BioTimasheff, unpublished results). In fact, 3 13. ARAKAWA, chemistry 23,5912-5923. M NaSCN (similar to KSCN) was shown 14. AUNE, K. C., AND TIMASHEFF, S. N. (1970) Bicto induce conformational changes of prochemistry 9.1481-1484. teins (16). 15. TUENGLER, P.. LONG, G. L., AND DURCHSCHLAG, H. Finally, it is known that protein binding (1979) And Biodem 98,481-4&i. and elution occur in polysaccharide col- 16. LEE, J. C., GEKRO, K., AND TIMASHEFF, S. N. (1979)

PROTEIN

INTERACTION

in Methods in Enzymology (Hirs, C. H. W., and Timasheff, S. N., eds.), Vol. 61, pp. 26-49, Academic Press, New York. 17. PAHLMAN,

S., ROSENGREN, J., AND HJERTEN, S. (1977) J. Chromatogr. 131,99-108. 18. STROP, P. (1984) J. Chromatogr. 294,213-221. 19. KATO, Y., KITAMURA, T., AND HASHIMOTO, T. (1984)

J. Chromatogr. 20. NISHIKAWA,

B&hem

298,407-418.

A. H., AND BAILON,

P. (1975)

And

68,274-280.

21. LEE, J. C., AND TIMASHEFF, S. N. (1974) Biochemistrg 13, 257-265. 22. FAUSNAUGH, J. L., GUPTA, P. S., AND REGNIER, F. E. (1984) Anal. Biochem 137,464-472.

WITH

105

COLUMNS

23. FUJITA, T., SUZUKI, Y., YAMANTI, J., TAKAGAHARA, I., FUKII, K., YAMASHITA, J., AND HORIO, T. (1980) J. Bioch~m 87,89-100. 24. MEVARECH, M., LEICHT, W., AND WEBER, (19’76) Biochemdy l&2383-2387. 25. LEICHT, W., AND PUNDAK, 114, X36-192. 26. NEAL,

J. L., AND GORING,

S. (1981) And

Biochem

D. A. L. (1970)

J. Chem. 48,3745-3747. 27. MARSDEN, N. V. B. (1977) Nuturwissenscha&n 148-149. 28. JANADO,

M., AND NISHIDA,

Chem 10,489-500.

T. (1981)

M. M.

Canad.

64,

J. Solution