these conditions the gel usually shrinks up to 50% [Demiroglou & Jennissen,. 1990]. ...... Fausnaugh, J.L., Kennedy, L.A and Regnier, F.E. (1984) Comparison of ...
IJBC, 2000, Vol. 5(2), pp. 131-163 Reprints available directly from the publisher Photocopying permitted by license only
© 2000 OPA (Overseas Publishers Association) Amsterdam N.V. Published by license under the Harwood Acadentic Publishers imprint, part of the Gordon and Breach Publishing Group. Printed in Malaysia
Hydrophobic Interaction Chromatography: the Critical Hydrophobicity Approach HERBERT P. JENNISSEN* Institut fUr Physiologische Chemie, Universitat-GHS-Essen, Hufelandstr. 55, 0-45122 Essen, Germany (Received January 17, 2000; In final form March 21, 2000)
1. INTRODUCTION
Even now over 20 years after the introduction of hydrophobic interaction chromatography the method has failed to gain the same foothold in the methodological repertoire of protein chemistry as has affinity chromatography, although a large number of proteins has been successfully purified by this method [Eriksson, 1989; Hjerten, 1981; Hofstee and Otillio, 1978; Mohr and Pommerening, 1986; Ochoa, 1978; Shaltiel, 1984; Yon, 1977]. In fact, a recent paper comes to the conclusion that certain "classical" commercial hydrophobic adsorbents are inadequate for an ideal downstream processing because of their high hydrophobicity [Oscarsson et al., 1995]. The criticism of these authors is essentially correct. The major problem encountered on such hydrophobic gels is that proteins can be very effectively adsorbed but an elution in a native state is often practically impossible. The problem therefore is to find a hydrophobic matrix with such a low hydrophobicity as to allow a binding and an elution of the intact protein. This problem is not new. The possibility of synthesizing low hydrophobicity gels has been present from the very beginning of hydrophobic interaction chromatography. In fact, such a "weak-hydrophobic chromatography system" in
* Mailing Address: Prof. Dr. H.P. Jennissen, Institut ftir Physiologische Chemie, Universitat-GHS-Essen, Hufelandstr. 55, D-45122 Essen, Germany Tel.: 49-201-723-4125, Fax: 49-201723-4694. 131
International Journal of Bio-Chromatography
132
HERBERT P. JENNISSEN
which the protein (phosphorylase b) eluted as a retarded peak on Seph-C 4 at a high salt concentration was demonstrated many years ago [Jennissen, 1981a]. The question is therefore: Can a generally applicable method be devised for the synthesis of gels with an optimal low hydrophobicity for proteins? As will be shown below we think that such a method is feasible. Non-covalent interactions of the hydrophobic type have been described as "the unusually strong attraction between non-polar molecules and surfaces in water" [Israelachvili, 1985]. The hydrophobic attraction energy for two contacting methane molecules is ca. 6-fold higher in water than the van der Waals interaction energy in vacuum and has been estimated to ca. -8.5 kJ morl [Ben-Nairn et at., 1977]. Hydrophobic interactions, which occur only in the presence of water, are based on the extrusion of a monomolecular layer of ordered water molecules covering two adjacent hydrophobic surfaces into less-ordered bulk water with a concomitant increase in entropy. Under the condition of ,1.H « T ,1.S the Gibbs free Energy is negative (-,1.G "" T ,1.S) constituting an entropy driven reaction. This entropy driven attraction between non-polar groups in water [Israelachvili 1985; Kauzmann, 1959; Lewin, 1974; Tanford, 1973; von Hippel and Schleich, 1969] is the basis for hydrophobic interaction chromatography as summarized in several fundamental reviews [Arakawa and Narhi, 1991; Halperin et at., 1981; Hjerten, 1981; Hofstee and Otillio, 1978; Oscarsson, 1997; Jennissen, 1988; Mohr and Pommerening, 1986; Ochoa, 1978; Shaltiel, 1984; Yon, 1977]. These reviews also contain first-hand information on the synthesis of hydrophobic supports and the separation methodology. For additional methodological information see [Hermanson et aI., 1992].
2. METHODS
Basic methodology underlying the chromatographic experiments referred to in this review can be found in the following papers [Jennissen, 1976a; Jennissen, 1976b; Jennissen, 1986; Jennissen and Botzet, 1979; Jennissen and Heilmeyer, 1975]. Fibrinogen was isolated from bovine blood according to [Blombaeck and Blombaeck, 1956; Mahn and Mueller-Berghaus, 1975] and further purified according to [Laki, 1951]. The clottablility was measured according to [Jacobsson, 1955] as modified by [Mahn and" Mueller-Berghaus, 1975] and was found to lie between 96-100% for the preparations employed. Human blood plasma was freshly prepared by centrifugation of whole blood containing 20 mM trisodium citrate. Protein (standard: bovine serum albumin, BSA) was determined on an AutoAnalyzer I (Technic on, Tarrytown, N.Y.) according to [Lowry et at., 1951].
HYDROPHOBIC INTERACTION
133
Uncharged alkyl-NC-Sepharoses (for nomenclature see [Zumbrink et al., 1995]) were prepared according to the carbonyl diimidazole method of Bethell et al. [Bethell et aI., 1979] in which uncharged alkyl carbamates are formed [Zumbrink et al., 1995]. For butyl- (SephC 4) and pentyl-agaroses (Seph-C s ) 14C-ethylamine (New England Nuclear, Boston) was employed as tracer for the determination of the degree of substitution [Jennissen and Demirolgou, 1992] and for hexyl-agaroses 14C-hexylamine (Amersham, Braunschweig) was employed [Demiroglou and Jennissen, 1990]. Uncharged butyl-SS-Sepharoses (for nomenclature see [Zumbrink et al., 1995]) of novel chemical structure [Demiroglou et al., 1994; Zumbrink et al., 1995] were prepared by activating with tresyl chloride and coupling to butanethiol in an alkaline medium [Demiroglou and Jennissen, 1990]. The degree of substitution was determined with 14C-butanethiol as described [Demiroglou and Jennissen, 1990]. Only fresh unregenerated gels were used in the experiments. Quantitative hydrophobic chromatography of fibrinogen [Jennissen and Demirolgou, 1992] was performed on small columns (0.9 x 12 cm) containing 2 ml of packed gel. The gel was washed and equilibrated with 20 vol. buffer A (50 mM TrisIHCI, 150 mM NaCl, 1 mM EGTA, pH 7.4]. As sample 1 mg purified fibrinogen was applied in a volume of 1 ml (in buffer A). The column was then washed with 15 ml buffer A and eluted either with 7.5 M urea or 1% SDS and fractions of 1.5 ml were collected. Protein was determined in the fractions as described above for the calculation of adsorbed and eluted protein. The experiments were performed at room temperature and the flow (40-50 ml/h) was achieved by gravity. The amount of fibrinogen adsorbed was determined by a balance calculation based on the various elution pools: run-through, 7.5 M urea-elutable fraction and 1 % SDS-elutable fraction. In general, the desired protein can either be measured specifically by an appropriate enzymatic activity test in a crude extract [Jennissen and Heilmeyer-Jr, 1975] or if the column is optimized utilizing a purified protein, by classical protein determinations [Jennissen and Demirolgou, 1992; Jennissen and Heilmeyer-Jr, 1975] as is the case here. The total yield of recovered fibrinogen on unsubstituted Sepharose 4B (control) was between 90-96%. The total yield for fibrinogen recovered by the various elution methods on the butyl-NC-Sepharoses (Seph-C 4) 4 - 47 Jlmollml packed gel was 89-92%. On the pentyl-NC-Sepharoses (Seph-Cs ) of 3 - 18 Jlmol/ml packed gel the yields were 89-94% and on highly substituted gels of 22 - 42 Jlmoll ml packed gel the yields were 77-85%. On the hexyl-NC-Sepharoses (Seph-C 6 ) of 2 - 42 Jlmollml packed gel the yields for fibrinogen were 77-85%. There appears to be a general tendency for the yield to decrease as the alkyl chain length is increased.
l34
HERBERT P. JENNISSEN
3. PRINCIPLES OF HYDROPHOBIC INTERACTION CHROMATOGRAPHY
The chromatographic separation of proteins by way of hydrophobic interactions was reported independently by two groups in 1972 [Er-el et al., 1972; Yon, 1972] and by three additional groups in 1973 [Hofstee, 1973a; Hjerten, 1973; Porath et at., 1973]. Historically the method then evolved in two distinct branches. The first branch was developed by Shaltiel, Hofstee and Yon [Er-el et al., 1972; Hofstee, 1973a; Yon, 1972] and centered on hydrophobic gels synthesized by the CNBr procedure, which may contain some residual charges .(hydrophobic non-charge-free gels, NCF-Gels) and on procedures run at low ionic strength. The second branch developed by Porath and Hjerten [Hjerten et al., 1974; Porath et ai., 1973] focused on gels synthesized by chemical methods which did not introduce charges (Hydrophobic charge-free gels, CF-Gels) and which were run at high ionic strength. In the first branch the name "hydrophobic chromatography" [Shaltiel and Er-el 1973] was coined, the dominant term for many years, whereas in the second the name "hydrophobic interaction chromatography" [Hjerten, 1973] predominated. In retrospect, although there is no doubt that uncharged hydrophobic gels are by virtue of displaying a single (pure) type of non-covalent interaction superior to the NCF-gels (see [Porath et al., 1973]), it appears that all groups involved in the development of this new method observed the binding and fractionation of proteins by hydrophobic interactions. In those cases of hydrophobic NCF-gels where the CNBr activated gels had been washed by the original procedure of [Porath et al., 1967] with 0.1 M NaHC0 3 (pH 7-8) or with 0.2 M Na2C03 (pH 9-10, [Jennissen, 1976bD prior to coupling with alkyl amines a large proportion of the cyanate esters was destroyed favoring uncharged coupling products [Kohn and Wi1chek, 1981]. Finally it was shown by Shaltiel in a careful study that in fact very similar results are obtained on hydrophobic CF-gels as on the original NCF-gels synthesized by the CNBr method (for review see [Halperin et al., 1981]). In addition it was shown by various groups [Hjerten, 1973; Hofstee ~nd Otillio, 1978; Jennissen, 1976b] that the charges introduced by the CNBr method were effectively quenched by salt in the range of 0.3-3.0 M so that the NCF-supports matched their uncharged counterparts in many of their hydrophobic properties. 3.1. The stationary phase
Important criteria for an optimal chromatographic separation of proteins by hydrophobic interactions are an extremely hydrophilic and a minimally interac-
HYDROPHOBIC INTERACTION
135
tive stationary phase. Long experience has shown that of the many stationary phases available charge free and chemical cross-link free agarose optimally meets both of these criteria in an ideal way. Besides being hydrophilic by virtue of its hydrogel structure (and thus an excellent starting point for hydrophobic modifications) it displays an extremely low non-specific adsorption of proteins. Agarose consists of a hierarchy of structures (for reviews see [Demiroglou et al., 1989; Jennissen, 1995]) leading to a paracrystalline structure with extensive areas for protein Adsorption [Demiroglou et al., 1989]. A beaded agarose gel of 4% agarose (Sepharose 4B, Pharmacia) contains ca. 5 x 106 beads [Jennissen & Heilmeyer, 1975] and ca. 30 mg of dry polysaccharide [Jennissen, 1976b] per ml packed gel. The agarose beads contain pores with an average pore radius of ca. 80-90 nm [Jennissen & Botzet, 1979; Demiroglou et al., 1989]. The total agarose surface area available for substitution with short chain alkyl residues in Sepharose 4B has been estimated to ca. 80 m2/ml packed gel (see [Jennissen, 1976b; Jennissen, 1981a]). For macromolecular solutes the individual specific surface area can be determined from adsorption isotherms. For the large protein molecule phosphorylase b (m = 197 kDa) the specific surface area of butyl agarose has been calculated to 8-9 m 2/ml packed gel [Jennissen, 1981a] indicating that only 10% of the total agarose surface is available to this protein. The above areas allow the calculation of surface concentrations of immobilized molecules [Jennissen, 1981a; Demiroglou & Jennissen, 1990]. Agarose can be maximally substituted with alkyl residues to ca. 50-100 Ilmollml packed gel. However under these conditions the gel usually shrinks up to 50% [Demiroglou & Jennissen, 1990]. Optimal degrees of substitution without gel contraction lie in the range of 1-30 Ilmollml packed gel. Present data strongly indicate that substitution procedures involving an initial activation step followed by a coupling step, as in the CNBr procedure, lead to a statistically uniform distribution of the immoblilized alkyl residues on the agarose gel [Amsterdam et al., 1974] with no indications for restricted accessibility [Jennissen and Demiroglou, 1982]. Finally, a coupling procedure (for comprehensive report see [Hermanson et al., 1992]) for the alkyl groups should be chosen which is simple, shows a statistical distribution, does not introduce charges, and shows no leakage of residues (for leakage see [Jennissen, 1979]).
3.2. The chain-length parameter A general systematic approach to the purification of proteins via hydrophobic interactions was initiated by Shaltiel [Er-el et al., 1972] who introduced the principle of the variation of the alkyl chain length of homologous alkylamines on NCF-gels which thus comprised a novel homologous series of hydrocarbon
136
HERBERT P. JENNISSEN
coated Sepharoses (Seph-Cn, [Shaltiel, 1974; Shaltiel, 1984]). They found that on a homologous series Seph-C 1 to SephC 6 the enzyme phosphorylase was excluded by Seph-C 1and Seph-C b retarded by Seph-C 3 and retained by Seph-C4 to Seph-C 6 . The major conclusion of these experimental results was that an increase in the chain length by -CH 2- units concomitantly increased the strength of protein binding from retardation to reversible binding to very tight binding ("irreversible" binding). In addition to this variation in affinity with the chain length the gels also changed in specificity for the adsorbed protein as was shown in later work (see [Shaltiel, 1984]). However the enzyme could only be eluted through reversible denaturation by a deforming buffer followed by renaturation to the active form. The experiments of Shaltiels group demonstrated the decisive influence of a "gradable" hydrophobic interaction between the matrix and a protein. Simultaneously the optimization of the gels allowed the purification of phosphorylase from a crude extract in one step. The principle applicability of these findings was confirmed on NCF-type hydrophobic alkyl gels by various groups [Hofstee, 1973c; Jakubowski and Pawelkiewicz, 1973; Jennissen and Heilmeyer, 1975; Jost et al., 1974; Raibaud et al., 1975] and also on neutral CF-type hydrophobic alkyl gels [Rosengren et al., 1975].
3.3. The surface concentration parameter 3.3.1. Critical surface concentration of immobilized residues
In 1975 however it could be shown that a second parameter is of equal if not greater importance for the binding of proteins to alkyl substituted gels. If, instead of the chain length, the suiface concentration of immobilized alkyl groups (i.e. density) is varied on Seph C 1 - Seph C4 protein adsorption is characteristically a sigmoidal function of the surface concentration of immobilized alkyl residues (Fig. 1) (suiface concentration series). In such a concentration series the strength of binding also increases from retardation to very tight binding as in the homologous series of Shaltiel. Fig. 1 illustrates the combined effect of chain elongation and surface concentration increase on the adsorption of the enzyme phosphorylase kinase to alkyl agaroses. Chain-elongation in a homologous series leads to a leftward shift of the sigmoidal curves of the concentration series and to a loss of sigmoidicity. Thus in the concentration series of alkyl agaroses a second general parameter for the variation of the gel hydrophobicity had been discovered which was equal to or more crucial for the binding of proteins than an increase in chain length alone [Jennissen and Heilmeyer, 1975]. Another important finding in this work was that a threshold value of the alkyl surface concentration, a "critical hydrophobicity can be defined at which a protein is adsorbed [Jennissen and
HYDROPHOBIC INTERACTION
137
Heilmeyer, 1975]. With a ratio of alkyl residues: positive charges in the gels of ca. 10:1 [Jennissen and Heilmeyer, 1975] the predominance of hydrophobic interactions as the basis for adsorption was convincing. Sigmoidal adsorption curves and critical surface concentrations were also obtained from binding data in the presence of 1.1 M ammonium sulfate (see Fig. 2) i.e. under charge quenching [Jennissen, 1978a] and a similar behavior has also been demonstrated on charge-free hydrophobic gels at low ionic strength [Jennissen and Demirolgou, 1992]. A sigmoidal binding of phosphorylase a to methyl-Sepharoses of increasing degree of substitution was also published several years later by Shaltiel's group [Shaltiel, 1978]. Rosengren et al. [Rosengren et al., 1975] later also showed the dependence of phycoerythrin adsorption on the density of immobilized alkyl groups in a series of uncharged Seph-C s to Seph-C]2 alkyl agaroses at high salt concentrations. Probably, due to the salting-out conditions and the long chain-length of the immobilized alkyl groups (~Cs), neither a pronounced sigmoidicity of binding nor a critical hydrophobicity were described. A prerequisite for the straightforward interpretation of the surface concentration series is a random distribution of the immobilized alkyl residues inion the agarose beads. This was shown to be the case by Shaltiel's group for CNBr modified gels [Amsterdam et al. 1994]. A non-random distribution, possibly as alkyl residue clusters, in gels synthesized from hexyl bromide (added in reaction-limited amounts) was assumed by [Caldwell et ai., 1976]. These authors coul.d not detect a sigmoidal dependence of adsorbed ~-amylase on the surface concentration of immobilized hexyl residues. Another important parameter, in the observation of sigmoidal adsorption curves is possibly the agarose structure itself [Jennissen, 1989; Jennissen, 1995] which may form sheets thus providing two-dimensional areas for the adsorption of proteins. This may not be the case for supports based on a different chemical structure such as dextrans. A very similar approach to the generation of graded hydrophobicity surfaces as in the concentration series (Fig. 1) of alkyl agaroses has been reported in the synthesis of so-called "hydrophobicity gradients" on glass [Elwing et al., 1988]. However these latter gradients also contain a counter-gradient of charges which significantly influences the hydrophilicitylhydrophobicity balance in the adsorption of proteins so that such surfaces do not constitute true hydrophobicity gradients unless the charge gradient is effectively quenched. Sigmoidal curves cannot only be obtained for proteins on hydrophobic binding-site lattices as discussed above but also for the binding of cells to immobilized residues on surfaces. The adhesion of hepatocytes to immobilized sugars is a sigmoidal function of the surface concentration of the immobilized carbohydrate molecules [Weigel et al., 1979]. Thus the adhesion of erythrocytes to
HERBERT P. JENNISSEN
138
w oa::
o
II)
A.
§3-NH-CH 3
~
B.
§j-NH-CH 2-CH 3
:I:
~ o
w ~ u
12000 C. §3-NH-CH 2-CH 2-CH 3 D. §3-NH-CH 2-CH 2-CH 2-CH 3
f
~ ~
•
8000
;:)
w
~
z
2 w II) ~ 4000
>-
a:: o
:I:
a..
10
II)
o
100
:I:
a..
10
20
30
40
50
,0 MOLES ALKYLAMINE Iml PACKED SEPHAROSE
FIGURE 1 Dependence of the Adsorption of phosphorylase kinase on the chain-length and surface concentration parameters of a homologous series of alkyl-Sepharoses at low ionic strength. The amount of adsorbed enzyme activity per ml packed Sepharose was calculated from the difference between the total amount of applied units and the amount excluded from the gel. The crude extract was applied to columns containing ca. 10 ml packed gel. Insert: Double logarithmic plots of adsorbed phosphorylase kinase as a function of the degree of substitution. Experiments with purified phosphorylase kinase are included. The gels were prepared by the CNBr procedure. The left upper comer depicts a homologous series (Seph-C 1 to Seph-C4) of hydrocarbon coated agaroses. Only three members of this series (A, B, D) were employed in Fig. 1. For conversion of units to katal see [Jennissen and Botzet, 1993]. For further details see the text. From [Jennissen and Heilmeyer, 1975]. Seph-C]: (e) crude extract; (0) purified phosphorylase kinase Seph-C 2("') crude extract; (,0,.) purified phosphorylase kinase Seph-C 4 (D) crude extract
hydrocarbon-coated agaroses [Halperin and Shaltiel, 1976] can be expected to show a comparable behavior (see also [Jennissen, 198Ib]).
3.3.2. Multivalence: Cooperative interaction of multiple immobilized residues with the protein
A straightforward interpretation of the sigmoidal curves (Figs. 1-2) became feasible when the concepts of multivalence and cooperativity of protein adsorption were developed a year later [Jennissen, 1976b]. It became clear that the sigmoidicity and the "critical hydrophobicity" were due to the multivalence of the
139
HYDROPHOBIC INTERACTION
A
B
1200 800 400
0.1
0.2
0.3
c
0.4
FIGURE 2 Dependence of the adsorption of phosphorylase b on the suiface concentration parameter of Seph-C4 at 5°C and 34°C at high ionic strength The adsorbed amount of phosphorylase in the presence of 1.1 M ammonium sulfate was calculated from adsorption isotherms measured at each point at an apparent equilibrium concentration of free bulk protein of 0.07 mg/m!. The adsorbed amount of enzyme (v) is expressed in relation to the anhydrodisaccharide content of agarose in mol/mol anhydrodisaccharide. Similarly C indicates the immobilized butyl residue concentration in relation to the anhydrodisaccharide content of agarose in mol alkylresidue/mol anhydrodisaccharide. A monomer molecular mass of 105 was employed for phosphorylase b. The alkyl agaroses were synthesized by the CNBr method. A. Adsorption isotherms ("lattice site binding function" [Jennissen, 1981, 1988; Jennissen and Botzet, 1993]) of phosphorylase b in Cartesian coordinates. Insert: Scatchard plots of the sigmoidal binding curves with extrapolation of fractional saturation at 610 (5°) and 1220 (34°) /lmol enzyme/mol anhydrodisaccharide (corresponding to 6.2 and 13.4 mg/ml packed gel respectively). The broken lines indicate the mode of extrapolation. (e) 5°C; (0) 34°C. B. Hill-type plots of the sigmoidal binding curves. e the fractional saturation was calculated from the extrapolated saturation values of the Scatchard plot (A). The Hill-type coefficients nH (i.e. lattice interaction coefficients) are given in the graph. The apparent dissociation constants of half-maximal saturation (K'D.0.5) are 0.137 and 0.167 mol butylresidue/mol anhydrodisaccharide at 5° and 34° respectively (which corresponds to 14.0 and 17.0 /lmol butylresidues/ml packed gel respectively). (e) 5°C; (0) 34°C. For further details see the text [Jennissen, 1978a; Jennissen, 1981; Jennissen, 1988; Jennissen and Botzet, 1993]. From [Jennissen, 1978a]
interaction (i.e. the necessity for a simultaneous interaction of more than one alkyl residue or more than one separate local surface site with the protein moiety) [Jennissen, 1976b, 1976c; Jennissen, 1978a, Jennissen and Botzet, 1979]. The conclusion of multivalence and of protein binding on a binding-site lattice was not biased by putatively interspersed charges but could be generalized to include
140
HERBERT P. JENNISSEN
homologous (alkyl residues alone) and heterologous (alkyl residues + charges) lattices which should principally not behave differently in generating sigmoidal protein binding curves in a concentration series. In addition, it could be expected that the heterologous lattice would become "functionally" homologous by a quenching of charges in the presence of 1.1 M ammonium sulfate [Jennissen, 1978a]. Thus, the sigmoidal binding curves were not eliminated by high salt concentrations and on the Seph-C4 concentration series. Protein binding now displayed a positive temperature coefficient as is characteristic for hydrophobic interactions (see Fig. 2) [Jennissen, 1976b; Jennissen, 1978a]. Later a mathematical model was developed [Jennissen, 1976c; Jennissen, 1978a; Jennissen, 1981a] allowing an estimation of the minimum number of alkyl residues (see Fig. 2B) interacting with the protein as derived from the lattice interaction coefficient (see Jennissen and Botzet, 1993). The model of multivalence was also confirmed from a different perspective by equilibrium binding studies of phosphorylase b (1 mM) with 14C-hexylamine (> 1 mM) in solution at a salt concentration of 1.1 M ammonium sulfate (see Fig. 3). In this microdialysis experiment [Jennissen et ai., 1982] the addition of 1 mM phosphorylase b did not alter the free bulk concentration of 14C-hexylamine indicating that phosphorylase b did not bind hexylamine under these conditions. This allowed the estimation of a maximal binding affinity for hexy1 amine protein binding to Ko.5 -10-100 M- 1[Jennissen et ai., 1982]. Since in binding studies with immobilized butyl residues (Seph-C 4) binding constants of 9-16 x 104 M- 1 for phosphorylase b had been obtained [Jennissen, 1976b; Jennissen and Botzet, 1979] this microdialysis experiment (Fig. 3) clearly demonstrated that not even the long C6 residue, when immobilized, would be capable of adsorbing a molecule of phosphorylase in a monovalent manner (i.e. on a single immobilized residue). On a surface only the cooperative interaction of a critical number of such immobilized residues leads to adsorption. Since the multivalent mechanism of protein adsorption appeared to be the common denominator of most types of adsorption chromatography (hydrophobic and ion exchange) the term "multivalent interaction chromatography" was suggested [Jennissen, 1978a] but did not come to general use. Later the concept of multivalence was also applied to certain forms of reversed phase chromatography [Pearson et at., 1982] consisting possibly of a mixed form of hydrophobic interaction and reversed phase chromatography. As indicated above the model of multivalence has also been extended to the interaction of cells with immobilized molecules [Jennissen, 1981b]. The possibility of a "multi-point" or "multiple contact" type of binding in hydrophobic protein adsorption was also envisaged by Hofstee and Hjerten [Hofstee, 1973b; Hjerten et ai., 1974]. However the concept and term of multivalence are at variance to terms such as "multiple contacts" since the latter term does not differentiate between the binding of
HYDROPHOBIC INTERACTION
141
a protein to separate alkyl residues (separate local surface sites, multivalence) or just to different segments of one and the same alkyl residue, which also comprise multiple contacts.
~
>
6000
1
2
~
•
o 0 o • •
C E 4000
0
0
••••
-l
u.
2000
U. W
0 0
5
10
15
20
TIME, min FIGURE 3 Binding analysis 0!n-1 J4 C-hexylamine to phosphorylase b at high ionic strength in a flow dialysis cell according to [Feldmann, 1978]. The arrows indicate the addition of 10 ~ of 0.05 M n-I- 14C-hexylamine (5 x 107 cpmlml) to the upper chamber containing 0.5 ml buffer (10 mM Tris/maleate, 5 mM DTE, 1.1 M ammonium sulfate, 20% sucrose, pH 7.0). Samples were taken at the indicated times from the effluent fractions of the lower chamber (1 fraction/min). (e) buffer alone; (0) buffer + enzyme (85 mg/ml) For further details see the text and [Jennissen et ai., 1982]. From [Jennissen et al., 1982]
3.3.3. Adsorption hysteresis
An important consequence of cooperative multivalent protein binding on alkyl substituted surfaces was protein adsorption hysteresis [Jennissen, 1978b; Jennissen, 1985; Jennissen and Botzet, 1979]. Protein adsorption hysteresis implies that the· adsorption isotherm is not retraced by the desorption isotherm, due to an increase in binding affinity after the protein is adsorbed. The binding affinity increase can be attributed to an increase in multivalence [Jennissen and Botzet, 1979] which is either due to a reorientation of the protein on the surface or a to conformational change in which buried hydrophobic contact sites (valences) are
142
HERBERT P. JENNISSEN
exposed by the binding strain on the adsorbed protein [Jennissen, 1985; Jennissen and Botzet, 1979]. In the case of the adsorption of phosphorylase b to butyl agarose there was no evidence that a very large conformational change was taking place [Jennissen and Botzet, 1979], since the fully active enzyme could be desorbed from such gels by "deforming buffer" [Er-el et af., 1972] (see above). The danger of a protein undergoing an irreversible conformational change and even a denaturation after adsorption has however been shown for other proteins binding to hydrophobic matrices [Ingraham et aZ., 1985; Wu et aZ., 1986] as well as to non-hydrophobic supports [Jost et ai., 1974]. Adsorption hysteresis provided the first strong evidence for the concept that protein adsorption to multivalent surfaces is thermodynamically irreversible (~iS >0) and that a true equilibrium has not been reached [Jennissen and Botzet, 1979; Jennissen, 1985]. This conclusion of non-equilibrium adsorption applies to all types of surfaces whether they be flat non-porous such as quartz glass or metal surfaces or porous alkyl agarose gels as long as the protein is adsorbed multivalendy i.e. on several separate local surface sites in overlapping binding units [Jennissen, 1978a, Jennissen and Botzet, 1979]. Another conclusion from this concept is that protein adsorption in hysteretic systems may not be thermodynamically but moreover kinetically controlled [Jennissen, 1988]. Multivalence and adsorption hysteresis have decisive effects on the chromatographic behavior of proteins during hydrophobic interaction chromatography leading to non-linearity and skewed elution peaks in zonal chromatography and to false "irreversibility", i.e. extremely high affinity, in adsorption chromatography [Jennissen, 1981a, 1986]. In general two modes of chromatography can be distinguished on hydrophobic gels (see [Jennissen, 1981a]): (a) "irreversible" adsorption chromatography followed by lyotropic eluent displacement (e.g. H 20 [Geng et ai., 1990] or salt gradients [Arakawa and Timasheff, 1984; Szepesy and Horvath, 1988]) (b) linear zonal chromatography. Both types converge to ideal systems when multivalence and hysteresis are reduced by a decrease in the surface concentration of immobilized residues [Jennissen, 1981a]. In agreement with these observations it was reported that the resolution of hydrophobic interaction chromatography columns increased if the matrix ligand density was reduced [Fausnaugh et af., 1984]. Kargers group [Lin et aZ., 1991] has also described an interesting form of adsorption hysteresis of ~-lactoglobulin A on a weakly hydrophobic silica methyl polyether column. In this case hysteresis and a possible concomitant increase in "valence" was due to the shifting of a dimer-tetramer equilibrium from the dimer in solution to the tetramer species on the gel surface. Since the protein binding behavior of a hysteretic chromatographic system depends on the history of the procedures i.e. the time sequence of events (e.g.
143
HYDROPHOBIC INTERACTION
,
KSCN
.098
2
,
CaCl2
E ...... t')
•0
-
2
)(
en ....
Z
:;,
4
.195
NH4 Cl
.200
NaCI
2
w
en oct
z
~
w
en oct
...J
> ex: 0
:z:
Q.
~
:z: Q.
Na 2S04
.450
.1
.2
.3 IONIC STRENGTH,
.,
.5
.6
~
FIGURE 4 Influence of the salt parameter on the desorption of purified phosphorylase kinase from Seph-C2 (25 llmollml packed gel) with salt gradients of different ionic composition. Each column with 5 ml of the above gel (CNBr method) was loaded with ca. 11 mg of the enzyme. The gradients were produced from 100 mllow ionic strength adsorption buffer and 100 ml salt containing buffer. The number at the maximum of the elution profiles indicates the ionic strength of the peak fraction. For conversion of units to katal see [Jennissen and Botzet, 1993]. For further details see the text and legend to Fig. 1 and [Jennissen and Heilmeyer, 1975]. From [Jennissen and Heilmeyer, 1975]
144
HERBERT P. JENNISSEN
adsorption step, washing step, elution step as a function of time ) reproducibility depends on a strict adherence to the respective procedure. In addition sequential [Jennissen, 1976b, 1986, 1988] and competitive adsorption [Brash and ten Hove, 1984, Jennissen, 1986, 1988] together with displacement effects [Jennissen, 1986, 1988, Majetshak et al. 1998] in protein mixtures reflect this history dependence on a molecular scale and make such chromatographic systems very complex when applied to the purification of proteins.
3.4. The salt-parameter 3.4.1. Salting-out and salting-in on hydrophobical/y substituted hydrophilic gels The enhancement of hydrophobic interactions by high concentrations of salts [von Hippel and Schleich, 1969] on a hydrophobic matrix was first shown by Porath [Porath et al., 1973] on uncharged benzyl ether agarose which displayed hydrophobic and possibly charge-transfer i.e. electron-donor-acceptor (EDA) interactions [Porath, 1990]. Trypsin inhibitor could be purified 25-fold after being adsorbed at 3 M NaCl followed by elution in buffer without salt. In the same year this principle was applied to the fractionation of serum proteins on a homologous series of alkyl agaroses of the Shaltiel-type by Hjerten [Hjerten, 1973]. A similar chromatographic technique was described by Rimerman and Hatfield [Rimerman and Hatfield, 1973] who salted out E. coli proteins on an alanine-Sepharose with 1 M potassium phosphate and eluted with a decreasing salt concentration gradient. Application of a cell free E. coli extract to the column and development with a negative concentration gradient led to the elution of E. coli proteins in relation to their solubility in concentrated ammonium sulfate [Rimerman and Hatfield, 1973]. Similarly alkaline phosphatase could be adsorbed on phenylalanine-Sepharose in the presence of 1.25 M ammonium sulfate and eluted by a negative salt gradient [Doellgast and Fishmann, 1974]. Proof as to the mechanism and principle underlying these salt effects came in the simultaneous, independent reports that the effect of salts on the adsorption and elution of proteins on alkyl agaroses indeed followed the Hofmeister series of salts [Jennissen and Heilmeyer, 1975; Raibaud et al., 1975]. We could show that phosphorylase kinase is eluted ("salted-in") from a Seph-C 2-column by increasing salt gradients [Jennissen and Heitmeyer-Jr, 1975]. In a series of experiments the ionic strength of the eluted peak fractions was inversely related to the salting-in power of the anions ions in the gradient in agreement with the Hofmeister series of salts (see Fig. 4). Raibaud et al. [Raibaud et al., 1975] reported that the concentration of salts necessary for the inhibition of ~-galactosidase
HYDROPHOBIC INTERACTION
145
20
-
15
A 2
1
E
rn
IZ ~
~ :5
I'MO'O 10 5
rn
10
5
>a:
20
:c a..
15 1
15
20
25
0
rn
8
0
:c a..
10
2
3
r
,"""S'
SALT
1.1 M (NH 4 )2 504
I'MNaa
5
5
10
15
20
25
FRACTION NUMBER FIGURE 5 Application of the salt parameter by utilizing the inverse salt dependence for the chromatography of purified phosphorylase b on Seph-C1 (30 J.lmollml packed gel). In this form of salting-out chromatography the equilibration buffer contained 10 mM sodium ~-glyc erophosphate, 20 mM mercaptoethanol, 2 mM EDTA, 20% sucrose, 0.5 JiM PMSF, pH 7.0 (buffer B) to which either 1.1 M ammonium sulfate or NaCI was added. 6 mg/3ml phosphorylase b was added to 20 ml Seph-Cl in a 2 cm i.d. x 17 cm column. Fractions of 6.5 ml were collected. The gel was prepared by the CNBr procedure. A. Application of enzyme to a column equilibrated with buffer without (NH4hS04 or NaCI (1) Application of phosphorylase b in buffer B (2) Elution with buffer B + 1 M NaCl B. Application of enzyme to a column equilibrated with buffer with (NH4hS04 (1) Application of phosphorylase b in buffer B + 1.1 M ammonium sulfate (2) Elution with buffer B (3) Elution with buffer B + NaCI 1 Unit corresponds to 16.7 nkatal. For further details see the text and [Jennissen, 1976a]. From [Jennissen, 1976a]
146
HERBERT P. JENNISSEN
adsorption on a Seph C3-column shows a similar inverse relationship to the salting-in power of the anions employed. These experiments clearly indicated that the action of the ions was not due to a pure electrostatic but to a lyotropic effect. It could be argued that this effect is again due to the interspersed charges in the NCF-gels. However, proteins could also be eluted (salted-in) from uncharged CF-gels (octyl-Sepharose) by an increasing salt gradient as shown by [Raymond et aI., 1981]: after adsorption of proteins at high ionic strength a first major fraction of proteins was eluted by a decreasing salt gradient followed by the elution of a second fraction by an increasing MgCl 2 gradient [Raymond et al., 1981]. In agreement with these findings it was shown that hemoglobin, serum albumin [Memoli and Doellgast, 1975] and phosphorylase b [Jennissen, 1976a; Jennissen, 1976b] could be salted-out by ammonium sulfate on NCF-alkyl Sepahroses, in column [Jennissen, 1976a] (see Fig. 5) and in batch experiments [Jennissen, 1976b] under conditions, where unsubstituted Sepharose showed no adsorption. Finally Pahlman et al. [Pahlman et al., 1977] showed that the salting-out power of anions, for the adsorption of HSA on uncharged Seph-Cs, followed the order of the Hofmeister series of salts. In sum, these experiments again demonstrate the similar properties of the hydrophobic matrices containing some residual charges (NCF-gels) and the uncharged hydrophobic gels (CF-gels) in respect to the lyotropic action of salts on protein adsorption. In the case of the NCF-gels the mechanism of protein binding was further analyzed at different temperatures under equilibrium binding conditions [Jennissen, 1976b; Jennissen and Botzet, 1979]. At very low ionic strength (10 mM TrisIHCI, pH 7.0) the adsorption of phosphorylase b exhibited a negative temperature coefficient on Seph-Cl as well as on Seph-C4 [Jennissen, 1976b]. At high ionic strength (l.IM ammonium sulfate) Seph-C 1 retained its negative coefficient but Seph-C4 now exhibited a positive temperature (see Fig. 2) coefficient [Jennissen, 1976b; Jennissen and Botzet, 1979] demonstrating that thermodynamically two forms of salting-out adsorption on hydrophobic gels can be distinguished as exothermic and endothermic salting-out chromatography respectively (see above) [Jennissen, 1978a].1t may be argued that one may not be observing the temperature dependence of adsorption but that of a temperature dependent conformational change on the gel surface as reported for a-lactalbumin on a C2-ether silica column [Wu et al., 1986]. However, besides having to explain two thermodynamically different conformational changes for phosphorylase on Seph-C 1 and Seph-C4 under otherwise identical conditions, the putative temperature-dependent conformational changes of the protein have not been observed in solution. Experiments in aggreement with these results have been shown in the hydrophobic interaction chromatography of myoglobin and cytochrome c. In this
HYDROPHOBIC INTERACTION
147
case it was documented that a temperature increase enhanced binding without a major change in protein conformation [Ingraham et ai., 1985]. Since the salting-in action of salts can be counteracted by salting-out salts in the same system [Hanstein, 1979] various binary and ternary salt systems [el-Rassi et al., 1990] and combinations with detergents [Buckley and Wetlaufer, 1990] have been devised. Another variant of salt mediated chromatography is so-called "mild" hydrophobic chromatography [Hubert et al., 1991; Ling and Mattiasson, 1983]. In this method e.g. poly(ethylene glycol) or poly(vinyl alcohol) are immobilized on carbohydrate gels. Proteins can be retained on the column at high salt (e.g. phosphate) concentrations (1-3 M) and are eluted by negative step gradients. The inclusion of salt in this system generates an immobilized poly(ethylene glycol)-salt aqueous two-phase system. Since the immobilized poly(ethylene glycol) can be expected to retain its liquid phase properties (i.e. 3D system) this procedure is more likely a form of hydrophobic liquid-liquid partition chromatography (see also [Huddleston et al., 1996; Mueller, 1986; Mueller, 1990]) than true hydrophobic interaction chromatography (2 D system).
3.4.2. Theories of Salt effects All of the theories concerned with the action of salts on proteins and their hydrophobic adsorption to surfaces are based on the current theory of the nature of hydrophobic interactions [Israelachvili, 1985; Israelachvili and Pashley, 1984;' Kauzmann, 1959; Lewin, 1974; Tanford, 1973]. In general, there can be no doubt that water plays the decisive role in mediating these interactions thermodynamically and possibly also kinetically. Commencing from these theories there have been several approaches to understanding the salt effects on protein binding to hydrophobic surfaces. One of the earliest approaches is based on the action of chaotropic ions [Hamaguchi and Geiduschek, 1962] on the solubility of proteins [Hatefi and Hanstein, 1969]. The action of the chaotropic anions (SCN-, CL04-, r, N0 3-, Br-), which can be arranged in a series similar to the Hofmeister (lyotropic) series [Hofmeister, 1887; Hofmeister, 1888], is regarded to involve a structure-breaking effect on water since the ionic entropies of hydration are positive. The original chaotropic series was later supplemented by the chaotropic haloacetates [Han stein et al., 1971]. Thus one can observe a decreasing increment of the salt series on the surface tension of water and an increase in surface potential [Hatefi and Hanstein, 1969]. The authors conclude that "by making the water more disordered and lipophilic in the presence of the appropriate ions, it should be possible to weaken the hydrophobic bonds of membranes and multi component enzymes and increase the water solubility (salting-in) of particulated proteins and nonelectrolytes" [Hatefi and Hanstein, 1969]. This interpretation is
148
HERBERT P. JENNISSEN
supported by the solvent isotope effects observed in D 20 versus H 20 [Hanstein et al., 1974]. The opposite effect can be observed with so-called antichaotropic salts (SOi-, HPoi-, F) which reverse the resolution of hydrophobic interactions by stabilizing the water structure [Davis and Hatefi, 1972] and which generally have negative values of the ionic entropy of hydration [Hanstein et al., 1971]. For a quantitative description of the effect of salt concentrations on the solubility of the model solute 2-methylnaphthoquinone an extended Setschenow equation has been utilized [Hanstein, 1979; Hanstein et al., 1971]. The general conclusion of this work, with consideration of electrostatic and dispersion forces, is that chaotropes interact indirectly with solutes (e.g. proteins) mainly through their effect on water structure [Hanstein, 1979]. In the solvophobic approach of Melander and Horvath the surface tension of water is again of central importance. The energy necessary for bringing a solute into solution is equated with the energy needed for forming a corresponding cavity in water against the surface tension with a reduction in free volume. The net free energy involved in the association of two molecules is thus related to a reduction of the cavity size. The solvophobic theory allows the description of reversed-phase retention for a wide variety of conditions including organic solvents and neutral salts. Salting-in and salting-out is explained on the basis of the respective surface-tension-decreasing/-increasing effect of the salt which is applicable to reversed-phase chromatography as well as to hydrophobic interaction chromatography [Melander and Horvath, 1977; Melander et al., 1984]. Specifically in respect to retention in hydrophobic interaction chromatography it has been suggested that water also plays a central role as a displacing agent [Geng et al., 1990]. In addition to the dependence of retention on the non-polar molecular area of the protein and the interfacial tension in the aqueous salt solution there also appear to be specific salt effects (divalent cations) resulting from a direct binding of the salt to the protein e.g. of MgCl 2 [Arakawa and Timasheff, 1984; Szepesy and Horvath, 1988]. In the paper of Arakawa and Narhi [Arakawa and Narhi, 1991] the approach, that salts do not act by influencing the surface tension of water but by directly binding to the protein, is extended. The action of thiocyanates is therefore interpreted as due to a preferential binding to proteins followed by a destabilization of the hydrophobic interactions. All in all, in hydrophobic interaction chromatography the above theories concerning the water structure apply specifically to the individual hydrophobic interaction e.g. between an alkyl residue and a hydrophobic patch on the protein in the multivalent hydrophobic binding process and it appears that the different mechanistic approaches probably complement each other. In addition a general effect of the salt on the protein itself is possible which modulates the interaction. Irrespective of the mechanism, the applicability of the Hofmeister (lyotropic) series of salts, expanded by the chaotropic series, to hydrophobic interaction
HYDROPHOBIC INTERACTION
149
chromatography has been verified by many groups and these salts are important tools in controlling the adsorption and elution of proteins on these resins. The individuality of each protein in its quantitative interactions in such a system especially when the native state should be conserved, however, remains a very important factor in determining the behavior of each individual chromatographic system.
4. PRACTICAL ASPECTS: OPTIMIZATION OF HYDROPHOBIC CHROMATOGRAPHIC SYSTEMS
4.1 The homologous series procedure
As stated above there are two methods for the synthesis of controlled hydrophobicity gels: (a) via the homologous series of hydrocarbon-coated Sepharoses (variation of alkyl chain length [Er-el et al., 1972; Halperin et al., 1981; Shaltiel, 1984]) and (b) via the concentration series (variation of the alkyl surface concentration [Jennissen, 1976b; Jennissen, 1978a; Jennissen and Demirolgou, 1992]). From another view point both gel series essentially correspond to members of "hydrophobicity gradients". 4.1. 1. The Shaltiel Strategy
Shaltiel's homologous series method of synthesizing hydrocarbon coated agaroses was supplemented by a so-called exploratory kit for choosing the most appropriate column and for optimizing resolution for the purification of enzymes [Shaltiel, 1974]. This analytical kit, which was commercially available for some years, contained a homologous series of small columns from Seph-C 1 to Seph-ClO with two control columns [Shaltiel 1974]. The principle was to apply ca. 1 mg protein Iml packed gel and determine the lowest member of the homologous series capable of retaining the desired enzyme at a low salt concentration (ca. 10-100 mM). Simultaneously the elution procedure which often involved a reversible denaturation ("deforming buffer") or high neutral salt concentrations (salting-in) was optimized to yield the highest specific activity of the desorbed enzyme. The corresponding Seph-C n column was then selected for the purification of the desired protein. 4.1.2. The Hjerten Strategy
As discussed above proteins can also be adsorbed to hydrophobic alkyl agaroses at high salt concentrations only to be eluted by an inverse salt gradient. This was shown for homologous alkyl agarose series of the uncharged [Hjerten et al.,
150
HERBERT P. JENNISSEN
1974; Rosengren et aI., 1975] and the charge-containing type [Hjerten, 1973), The optimization strakgy which was also utilized by a number of commercial firms involved the application of the protein sample (ca. 5-40 mg/ml packed gel) at high salt concentrations (500-1500 mM), selection of the column from a homologous series which efficiently bound the protein and elution with an inverse salt gradient plus organic solvents (e.g. ethylene glycol, glycerol, ethanol, propanol, chaotropic neutral salts, urea, guanidine hydrochloride, detergents such as SDS or Triton XI00). As can be concluded form this list of additives the main problem of the procedure was again the inability to gently elute the protein by an inverse salt gradient. This procedure or variants thereof are still the method of choice for many groups today. However as illustrated by the paper of [Oscarsson et al. 1995] the number of failures is probably very high.
4.2. The critical-hydrophobicity procedure
The importance of the second of the two procedures, i.e. optimization via the surface concentration series, has been underestimated. Although the decisive importance of the immobilized alkyl residue concentration for the hydrophobic adsorption of proteins (critical hydrophobicity) has been stressed for many years [Jennissen and Heilmeyer,1975; Jennissen, 1976b; Jennissen, 1978a; Jennissen, 1979; Jennissen and Demirolgou, 1992] no hydrophobicity gradient gel series has ever been produced commercially. Against the above background of pressing problems in hydrophobic interaction chromatography a novel rational basis for the design of a hydrophobic chromatography system for the purification of a selected protein will be presented. The procedure has already been successfully employed for the purification of fibrinogen from plasma [Jennissen and Demiroglou, 1994] and seems promising for the purification of other proteins as well (unpublished). 4.2.2 Application of critical hydrophobicity supports
This concept is based on a very simple idea. High yields in hydrophobic interaction chromatography can only be obtained if the protein to be purified is fully excluded from the gel under as physiological elution conditions as possible i.e. at low ionic strength. This means that the gel should be fully non-adsorbing under these conditions. On the other hand, since a purification is only possible if the protein is adsorbed to the gel, the matrix should be constructed in a way, that adsorption can be easily induced by other means without denaturing the protein. According to this concept, working at or near (juxtacritical) the critical hydrophobicity point should solve both problems. At critical or juxtacritical hydrophobicity of the matrix no protein is adsorbed at low ionic strength but adsorption
HYDROPHOBIC INTERACTION
151
can be induced by a finite high salt concentration with the possibility of utilizing either (a) elution chromatography by a negative salt gradient ("irreversible" adsorption conditions) or (b) a linear zonal chromatography (reversible adsorption conditions) [Jennissen, 1981a). If the selected gel is significantly below the critical hydrophobicity protein adsorption may not be inducible by salt and if the gel is significantly above the critical hydrophobicity a complete elution of the protein will not be possible due to "irreversible" adsorption. The aim of this procedure is, therefore, to synthesize uncharged alkyl-Sepharose supports very close to the "critical hydrophobicity" for adsorbing the protein under physiological conditions. Although essentially any type of molecule displaying hydrophobic properties could be coupled to the matrix, even proteins [Jennissen and Botzet, 1993], the immobilized residues should be restricted to alkane derivatives, to insure a "purity" of hydrophobic interactions. This restriction, in the first instance, also excludes pi- or d-electron containing molecules e.g. phenyl residues or sulfur atoms which should be avoided, because they may introduce additional charge-transfer interactions [Porath, 1978; Porath, 1989, Jennissen, 1995]. Strongly salting-out ions should also be avoided, since they may induce nonspecific protein adsorption on the carbohydrate support itself (see discussion above). NaCI centrally located in the Hofmeister series appears to be an ideal salt to initially employ from the biochemical as well as the physical chemical view point. The critical hydrophobicity procedure involves the optimization of the 3 basic parameters and their combination in the final chromatography procedure under standard conditions: 1. Chain-length parameter: Selection of a gel with an appropriate alkyl chain length from a homologous series (protein sample: 05-1 mg/ml packed gel) by its capability of adsorbing ca. 50% of the applied selected protein at a low salt concentration (10-150 mM). 2. Surface concentration parameter: Determination of the critical hydrophobicity at the predefined low ionic strength on the alkyl gel of the chain length selected from the homologous series under point 1. 3. Salt parameter: Determination of the optimum high salt concentration(1-5 M N aCl) necessary for a complete adsorption of a specified amount of the selected protein (e.g. 0.5-1 mg/ml packed gel) on the critical hydrophobicity support determined under point 2. 4. Chromatography: Adsorption of protein on critical hydrophobicity support at specified high salt concentration and elution by an inverse salt gradient. The three parameters (1-3) can be determined by a simplified version of quantitative hydrophobic interaction chromatography [Jennissen and Demiroglou, 1992] aimed primarily at high affinity adsorption sites. In distinction to the previ-
152
HERBERT P. JENNISSEN
Sepharose 4B
0.3
zW
(!)
0-
A UREA
0.2
l
z..E.
-01
a: E CO
u::
0.1 0
0
5
15
10
20
25
FRACTION NUMBER Alkyl Seph. (low hydrophobicity) UREA
0.3
w
(!)
0-
0.2
z..E.01 -
a: E CO
u::
run through
0.1 0
8
l
Z
0
5
10
15
20
25
FRACTION NUMBER
Alkyl Seph. (high hydrophobicity)
0.3
z
w
(!)
0-
UREA
0.2
C
SOS
z..E.01
-
a: E
CO
u::
0.1 0
0
5
10
15
20
25
FRACTION NUMBER
FIGURE 6 Quantitative hydrophobic chromatography offibrinogen on butyl-SS-Sepharose gels. Fibrinogen (1 mg) was applied in 1 ml to a column (0.9 cm i.d. x 12 cm) containing 2 ml packed gel in 50 mM TrisIHCI, 150 mM NaCI, I mM EGTA, pH 7.4. Fractions of 1.5 ml were collected. The column was washed with 15 ml buffer followed by elution either with 7.5 M urea or, at high hydrophobicity of the gel, with 1% SDS. The butyl-SS-Sepharoses were synthesized by the tresyl chloride method. A. un substituted Sepharose 4B B. Butyl-SS-Sepharose of low hydrophobicity (7 Ilmollml packed gel) C. Butyl-SS-Sepharose of high hydrophobicity (47 Ilmollml packed gel) For further details see Methods, the text and [Jennissen and Demirolgou, 1992]. Except for C (unpublished) adapted from [Jennissen and Demirolgou, 1992]
HYDROPHOBIC INTERACTION
153
ously employed quantitative chromatographic method [Jennissen and Heilmeyer, 1975] the selected protein in this case is not applied to the column until equilibration is complete as determined from an identical concentration of protein or catalytic activity in the run-through and sample. Instead, a constant amount of enzyme activity or a protein (in this case fibrinogen) at an intemlediate bed load (e.g. 0.5 mg/ml packed gel for a purified protein) is applied, followed by a wash with protein-free buffer until no more protein (enzyme activity) is detected in the run-through (see Methods). In single protein systems the adsorbed protein can then be eluted by urea or SDS to gain additional information on the adsorption state. A typical series of model experiments illustrating the principle of the evaluation system at different hyrophobicities and affinities is shown in Fig. 6 employing immobilized butane thiol gels (butyl-SS-Sepharose) which in our hands yielded the highest affinities towards proteins [Jennissen and Demiroglou, 1992; Zumbrink et ai., 1995]. On the unsubstituted control Sepharose 4B (Fig. 6A) the applied fibrinogen (l mg) is excluded from the column and is quantitatively found in the run-through. No protein is eluted by 7.5 M urea (Fig. 6A) nor by 1% SDS (not shown). On a Sepharose of low hydrophobicity (Fig. 6B) most of the retained protein can be desorbed by 7.5 M urea. Thus the urea peak becomes a simple measure of the protein bound with average affinity at low to medium hydrophobicity. On a highly hydrophobic column (Fig. 6C) however only 27% of the adsorbed fibrinogen can be eluted by urea. An additional amount of 52% can only be desorbed by 1% SDS and may be termed "tightly" bound protein. At this high hydrophobicity there is still some SDS-resistant protein bound (termed "irreversibly bound") some of which can be eluted by 1% SDS in 0.1 M N aOH (not shown). Selection of the appropriate alkyl chain length. In the first basic step information is gained on the general protein binding properties of each type of uncharged alkyl Sepharose in a homologous series. In Fig. 7 B it can be seen that even at the highest degree of substitution available (47 Ilmollml packed gel) for Seph-C 4 only ca. 50% of the applied fibrinogen is bound. At 22 Ilmol/ml packed gel (Fig. 7A) no fibrinogen binding can be measured. In contrast, on Seph-C 5 at 221lmoIfmI packed gel (Fig. 7D) 68% of the fibrinogen is adsorbed and at a similar degree of substitution on Seph-C 6 (Fig. 7H) 100% is adsorbed. Half maximal adsorption occurs at ca. 47 Ilmol Iml packed gel on Seph-C4 (Fig. 7B), at ca. 22 Ilmollml packed gel on Seph-C5 (Fig. 7D), and at ca. 8 Ilmol/ml packed gel on Seph-C6 (Fig.7G). The reciprocal of these half-saturation values gives a crude estimation of the affinity of the respective immobilized alkyl residue
154
HERBERT P. JENNISSEN
.
towards fibrinogen (see lattice adsorption constant K~.5 in [Jennissen, 1981a; Jennissen and Botzet, 1993]). 0.4
r E
z
~
A 0.3
0.2
~
it CD
0.1
iL
'A \.
5
0.4
f-E
0.3 0.2
it CD
iL
0.1
f.
0.3
z w 8z
0.2
CD
0.1
E
1
it iL 10 15 FRACTION NUMBER
20
r-.A 5
10
~\
0.4
B
1
15 jJJT10IJml packed gel
i
8z iiiiL
15
zw
10 15 FRACTION NUMBER
1
0.2
0.1
........ 5
1
0.1
~~ 5
l
E
0.3
i
0.2
u::
0.1
15
II: CD
20
G
5
FRACTION NUMBER
FRACTION NUMBER 0.4
E
'"Z
0.3
8;;!;
0.2
E
I~
w
a: