PROTEIN

EXPRESSION

AND PURIFICATION

3,263-281

(19%)

REVIEW Immobilized Metal Ion Affinity Chromatography Jerker

Porath

Division of Biotechnology, Life Sciences South 225, University of Arizona, Tucson, Arizona Separation Center, Biomedical Center, BMC, Box 577, S-75123 Uppsala, Sweden

The introduction of immobilized metal ion affinity chromatography, directed toward specific protein side chains, has opened a new dimension in protein purification. This review covers the principles and practice of IMAC that can be performed under very mild, nondenaturing conditions. IMAC is particularly suitable for preparative group fractionation of complex extracts and biofluids, but can also be used in high-performance mode: “HP-1MAC.” Single-step purifications of lOOOfold or more may allow isolation of a particular protein from crude extracts on a milligram or gram scale. With respect to separation efficiency, IMAC compares well with biospecific affinity chromatography, and the immobilized metal ion ligand complexes are more likely to withstand wear and tear than are antibodies or enzymes. The enormous potential of IMAC and related metal affinity techniques is only in the initial stages of being explored and exploited. Synthesis of IMA adsorbents, and various modes of performing IMAC are discussed and exemplified with selected applications. Advantages and disadvantages are listed. Effective means of counteracting the few undesirable effects that can OCCUr are suggested. 0 1992 Academic Press. Inc.

We have witnessed a remarkable expansion of knowledge in the molecular aspects of biological sciences within the last half of this century. This would hardly have been possible without the continued elaboration of new techniques for isolation of various natural compounds, including the biopolymers. Since the late 196Os, biospecific affinity separation methods have come into focus and have been continuously improved. Among these important developments is the use of more robust adsorbents that are not prone to leaking immunogenic degradation products. The immobilized metal ion affinity methods, which exploit chemical affinity between immobilized metal ions 1046-5928/92

Copyright All rights

$5.00 0 1992 by Academic

of reproduction

85721;

and Biochemical

(IMA)’ and the surrounding solute molecules, belong in this category. Other IMA methods are forthcoming (l), but in this article we consider only IMA chromatography (IMAC). IMAC of proteins was introduced in 1975 under the name of metal chelate chromatography (2); it has taken a decade or so for IMAC to be generally recognized. To date, a number of review articles cover rather well the various aspects of IMAC (3-39). I therefore confine the topic of this article to general principles, some of which have not been sufficiently emphasized before, and I select a few examples for illustration. Complexation chemistry at the surface or within the network of cross-linked hydrophilic polymers is largely unexplored. One can anticipate it to be different, in some aspects, from the corresponding chemistry in free solution, and we should bear this in mind when consulting the literature for guidance in better understanding the structures and kinetics involved in IMAC. IMA

CONCEPTS

A metal ion, with affinity for analytes or ligates’ in the sample to be fractionated, is fixed to an insoluble matrix and serves as the characteristic and most essential part of an adsorption center. The interactive metal ions must be localized in exposed positions in the reticular network of the polymer to allow complexation. They i Abbreviations used: IMA, immobilized metal ion affinity; IMAC, immobilized metal ion affinity chromatography; IDA, iminodiacetate; TED, tris(carboxymethyl)ethylenediamine; CM-Asp, carboxymethylated aspartic acid, NTA, nitrilotriacetic acid, TREN, tris(2-aminoethyl)amine; DPA, dipicolylamine; C,-S gel, hexylsulfido-liganded agarose; T gel, 2(hydroxyethylsulfido)ethylsulfone-liganded agarose; EDTA, ethylenediaminetetraacetate; MIT, metal ion transfer; CASMAC, cascade-mode multiaffinity chromatography; SAP, serum amyloid P; LDH, lactate dehydrogenase; Mes, 4-morpholineethanesulfonic acid; Mops, 3[morpholino]propane sulfonic acid, POPSO, piperazine-NJ’-bis[2 hydroxypropanesulfonic acid. 2 In this article ligate refers to a ligand in the mobile phase. 263

Press,

in any form

Inc.

reserved.

JERKER

264

may be bound to carboxylic-sulfonic-, or sulfate estertype cation exchangers. With few exceptions, however, far more stable immobilization can be achieved using the chelation or macrocyclic effects of a matrix-fixed metal ion acceptor. The chain of structural units taking part in the formation of the matrix-adsorbate complex may be symbolized then by the simple formula

PORATH

in the presence of metal ions in the mobile phase. Conceivably, adsorption of metalloproteins may also occur according to this concept. Metal ion transfer. If the protein has a very strong affinity for the metal ion, demetallation of the adsorbent may occur: Qr-M+P+&+MP.

Pmx-Sp-Ch-M-*-P or even a simpler formula Q-M...

P

where Pmx Sp Ch Q

= = = =

polymer matrix spacer arm chelator the immobilized chelator, that is, the chelating adsorbent M = metal ion P = protein or peptide - = a stronger bonding than . . *.

Such a process is called metal ion transfer (MIT) (21). Desorption. Either the chelator-metal bond or the metal-protein bond may be cleaved. We may therefore distinguish the following alternative mechanisms of desorption: (a) A competing ligate, L (ligand in the mobile phase), serves as an electron donor or Lewis base (for example, imidazole or glycine): &h-M...P+L

&+M+CJr-M with a large “apparent” association constant K,. The metal ion association should not, however, be “complete” (vide infra) in the sense that all coordination sites are occupied by the immobilized chelator. The formation of the adsorbate may then be symbolized as

Since Q and P are interchangeable reversed IMAC may occur: &+MP+Ch

-0.

adsorption

partners

MP.

Reversed IMAC. This was anticipated and Belew (8). It may explain the adsorption

by Porath of proteins

\&h+LMP

(b) A metal ion, M*, can act as an electron acceptor or Lewis acid. It can desorb protein according to two alternatives:

Adsorption Formation of the adsorption compound. Formation of the adsorption compound (also called adsorbate) takes place in two steps: (1) immobilization of the metal ion by chelate formation (to form the IMA adsorbent) followed by (2) protein adsorption. In both steps the metal ion acts as an electron acceptor or a Lewis acid. Neglecting the presence of simple weakly bound ligands such as water molecules and buffer ion solutes that have no significant influence on the adsorption and desorption events, we may write the first reaction as

rCJiML+P

a--M.

..P+M*

ra-M*

+ MP

ha--M

+ M*P

(c) Protons can take the place of M*: Ch--M.. Thus desorption the pH. COLUMN

.P + H+

rQ-H

+ MP

LCJ-M

+ HP

may be accomplished

by lowering

ARRANGEMENTS

IMAC may be executed as a two-step fractionation procedure with five consecutive operations which are readily automated: column equilibration, sample introduction, removal of unbound fraction (“washing”), elution, and regeneration. These operations are treated in some detail in later sections. The sample substances are divided into two major categories: those with and those without affinity. On rare occasions, an IMA adsorbent may be uniquely specific and we may, in a single step, isolate the desired compound. Typically though, substances display different degrees of affinity and must be separated by judicious selection of elution conditions. To circumvent metal contamination of the sample or to make fractionation more efficient, we can add one or more beds to the column. For example, to remove metal ions either bound weakly to proteins (or other sample

IMMOBILIZED

METAL

ION AFFINITY

substances) or present in the column effluent, we can add beds of metal-free chelating adsorbents upstream or downstream of the IMA column (Fig. 1). Selectivity of adsorption and desorption depends to a large extent on properties of both the metal ion and the chelator ligand since they together constitute the essential part of the adsorption center. The number of combinations that can be used in tandem reflects the number of metal ions and chelators used to create various adsorbents. The following arrangements, where TED (tris(carboxymethyl)ethylene diamine) and IDA (iminodiacetate) are abbreviated names for two polymer-bound ligands described in a later section, illustrate the concept of varying metal ions and ligands in different combinations. Figure 2 shows some combinations of two metal ions, Zn2+ and Cu2+ and two adsorbents, IDA and TED gels. Note that Zn2+ forms weaker complexes than Cu2+, TED forms weaker protein adsorbents than IDA. The weaker metal-resin combination should precede the stronger, which is evidently the case for A, B, and C. Combination D is likely to be an accurate order, but experimental confirmation is lacking. However, if selectivity of the adsorbents is very different, the order of beds may be of no relevance. In other words, the same efficiency in separation is obtained independently of the tandem order (40). Since the TED-immobilized metal-protein association complex constants are much smaller than those of corresponding IDA complexes, some proteins not binding to M-TED may bind to M-IDA. After introduction of a sample and washing with the equilibration buffer, the individual columns are eluted separately. Columns can be either disconnected or, preferably for large columns, appropriately provided with three- or four-port valves, allowing the elution of individual columns in situ, that is, without physically disrupting the tandem setup. In this manner at least three fractions are obtained: proteins not binding to either column, proteins

m 0 FIG.

Bed loaded with metal ions Bed without metal ians 1.

Some IMAC column arrangements.

265

CHROMATOGRAPHY

4 Zn-TED

Zn-TED

CU-TED

Zn-IDA

Cu-IDA

Cu-IDA

CU-TED

F Zn-IDA

FIG.

2.

IMAC tandem column arrangements.

binding to and eluting from M-TED, and proteins binding to and eluting from M-IDA. To achieve more efficient group fractionation, a “cascade-mode” arrangement may be employed (16,42,43). For example, three different adsorbents may be arranged in a cascade for a two-step fractionation from each of the three individual column units (Fig. 3). Distinguishing only nonadsorbed and desorbed fractions from cascade-mode column fractionation with n separate beds will yield 2” fractions. With three elution steps per column, we obtain 4” fractions. The efficiency of the technique is based on the assumption that there is high selectively in the adsorption and desorption steps and that it is possible to elute material from each bed in such a way that part of the material is adsorbed in the next section of the component column. Cascade-mode column arrangements can serve at least two purposes: 1. to facilitate the efficient use of multiple affinity chromatography (CASMAC: cascade-mode multiaffinity chromatography (43)), and 2. to select suitable adsorbents and operation conditions for the isolation of a desired substance. Solvent (eluent) change during the chromatographic operation may be greatly facilitated in some instances

I 3. An example of cascade-mode column arrangement. 4 signifies flowthrough fraction from the bed and & signifies fraction desorbed from the bed.

FIG.

266

JERKER

by inserting columns of molecular sieves and/or other adsorbents. Concomitant fractionation can also be attained in such additional columns. An illustration of this technique is presented in a later section. SYNTHESIS

OF IMA

ADSORBENTS

All metal ions that interact specifically with proteins are potential partners in the adsorption centers for IMA methods. It is convenient to distinguish metals by Pearson’s polarizability-based classification as soft, hard, and intermediate or borderline metal ions, where soft ions are those most easily polarized in an electric field (44). Typical soft metal ions are Hg2+, Cd2+, Cu+, Ag+, Pb2+, and the platinum group of metal ions. These metal ions prefer the “soft” sulfur as a coordination partner. Coordinate bonds involving these metals are essentially covalent in nature. Immobilized soft metal ions have been only superficially studied to date as adsorbents for proteins (23). Ions of group II and III elements are typically of the hard kind, and among these Ca2+ and Mg2+ are of particular biological interest. Applications of these ions in IMAC are under intense, but not yet far-advanced, methodological study. Their coordination chemistry as relevant to IMAC is not extensive; oxygen is the preferred bonding partner, with the metal-oxygen bond essentially ionic. The heavier group III metal ions tend to have partial covalency intermixed in the complexation. This is also reflected in their IMAC behavior. For example, among the group III B3 immobilized metal ions T13+ is similar to Zn2+ and Cu2+ in being borderline, whereas A13+ is definitely hard (45). Although a more detailed discussion of the use of immobilized hard metal ion-based adsorption is premature at present, this branch of IMAC is already not only challenging but indeed promising (45-49). For example, the metal exchange routes with tripositive lanthanide ions as partners in the adsorption centers and Ca2+ as constituents in the eluent seem to offer the possibility for selective fractionation of calcium-binding proteins (vide infra). Iron forms di- and tripositive ions, Fe3+ being harder than Fe2+. Immobilized Fe3+ seems to be a potential selective adsorbent for the phosphoproteins and organophosphates in general (50-52). In a broad sense, classical protein adsorbents such as alumina and calcium phosphate (hydroxyapatite) may be considered immobile hard metal affinity adsorbents in their own class (as are insoluble zirconates and titanates). Under usual operating conditions, they function chiefly as ion exchangers. However, recent findings using polymer adsorbents indicate that it may now be worthwhile to reinvestigate these substances as potena Group III ture versions.

A or group

13 (IUPAC)

according

to other

nomencla-

PORATH

tially selective, inexpensive IMA adsorbents for use in very large-scale industrial applications. Further study of water-insoluble metal oxides and sulfides is also warranted. No sharp boundary exists in the ranking of hard and soft metal ions. Among the borderline metal ions, those belonging to the 3d-block elements, Zn”, CL?, Ni2+, and Co2+, are particularly noteworthy, and their complexation chemistry is very extensive indeed. They have been intensely studied in IMAC and when immobilized on solid supports they form very useful IMA adsorbents. Dipositive zinc and nickel ions are electrochemically stable under the experimental conditions used in protein chromatography. Cu2+ and Co2+ can be readily reduced or oxidized, respectively, by redox active solutes. Although they act as catalysts in decarboxylation, hydrolytic, or redox reactions, we have never encountered any adverse reactions with these ions while they are immobilized. On the contrary, immobilized copper ion, Cu2+, forms versatile, strong adsorbents of very high capacity for peptides and proteins. As discussed in detail elsewhere (40), reversible and rapid kinetics are of paramount importance in IMAC. Slow kinetics limit the utility of Cr3+, Co3+, and several ions of the higher transition elements. It is justified to stress again that the chemistry is different for metal complexes while they are in free solution and/or at the boundary between polymer and aqueous phase. How different cannot easily be quantitated theoretically, but often we can make educated guesses. To wit, we know from solution chemistry that Cu2+ and Ni2+ are able to form different kinds of complexes (square, tetrahedral, octahedral, etc.) that are often distorted when immobilized due to steric hindrance. Presumably, the immobilization of metal ions alters their chemistry to some extent as well. Steric hinderante caused by the “breathing” polymer network, a “dangling” spacer arm used for ligand attachment, and the differences between the structure of bulk water and water plus salt concentration gradients proximal to the polymer chains are all factors perturbing the molecular shape and reactivity of an adsorption center. Predictions or inferences regarding the chemical properties of such centers based on knowledge of similar complexes in solution are, therefore, highly tentative. Suggested “structures” are necessarily hypothetical in their details. How to Select and Attach the Chelator Ligand A polymer-bound metal ion-complexing agent serves two aims: (a) it fixes the metal ion to a solid support, and (b) it modulates the metal affinity binding and thus the strength and affinity specificity of the adsorption center, and consequently also its capacity.

IMMOBILIZED

METAL

ION

AFFINITY

Ideally, a metal ion should be strongly anchored to the insoluble matrix so that virtually no metal ions are released into the solvent phase or transferred to the proteins or other analytes, which have competing metal ion combining sites. To approach this goal, we might consider the use of a multidentate chelator for which the number of nucleophilic atoms exceeds the total number of coordination sites on the metal ion. However, such a ligand would be likely to drastically reduce the affinity between the metal ion and the protein or peptide to be isolated and may even completely abolish ligate adsorption. On the other hand if we instead use a chelator at the opposite end of the metal affinity spectrum, an unidentate ligand, as is the case with most ion exchangers, then we cannot satisfy the demand for a strong metal

RI\

267

CHROMATOGRAPHY

ion fixation. In our quest for a workable chelator, we must then combine these two partly counteracting factors in order to produce a stable yet active adsorption center. In essence, this is one of the practical problems at the heart of IMAC. By varying the number and ratio of the coordinating atoms (0, N, and S) and the overall structure of the chelator, we can produce all kinds of IMA adsorbents to meet specified demands. Many methods for covalently fixing chelators to hydrophilic polymers are available (53,54). Most ligands used in IMAC have been coupled to the polymer matrix (P) by variations of the oxirane (epoxide) method with the reactive epoxide as part of the chelator-containing group or the polymer (Al-A3). R, R,, and R, represent metal-chelating groups:

/Rl

0P -OCH,-CHOH-CH,-N

R/p

@-0-CH,-CH-CH,

‘R,

*

‘0’

@-0-CH,-CHOH-CH,-SR

A2

@-0-CH2--CH0~-C~,--0~

~3

F

If a long spacer arm is desirable, the polymer may first be reacted with a long-chain bifunctional reagent such as butanediol bis(glycidy1) ether and subsequently coupled with a chelate-forming substance in alkaline medium (53): CFTCH-CH,-0-(CH,),-0-CH,-CC\H-,,,,H, 0

0

Aliphatic chelators are conveniently attached to an oxirane polymer through amino or thiol groups as in Al and A2. Polyamines are coupled to form intermediates for preparation of many extremely useful IMA adsorbents. The simplest of these, the “IDA” gels, are commercially available and may be synthesized by carboxymethylation of cross-linked aminopolymers. The conditions for avoiding side reactions have been described by Hemdan and Porath (55). However, to avoid undesired reactions, the IDA chelator may be obtained also by coupling IDA to an epoxypolymer:

The IDA chelator is tridentate and forms a double five-membered ring chelate with tetra- and hexacoordinate metal ions. A tetradentate chelator prepared by carboxymethylation of an amine forms chelates stronger than those formed by a tridentate chelator. The following adsorbents with tetradentate chelators have been used in IMAC: carboxymethylated aspartic acid polymer (CM-Asp (47-49)) , CH,-COO-N ‘TH-COOCH,and nitrilo

triacetic

-CH , CH,COO@-0-CH,-CH-CH, ‘0’

ä

CH,COO, CH,COO-

@---0-CH,-CH~H-CH,-N

COO-

acid adsorbent

,,CH, jN\ CH,\

(NTA

56))

COOCOO-

coo-

alkali

+ NH \

Al

\

CHJOO-

“TED” agarose prepared from ethylenediamine agarose by carboxymethylation has a pentadentate chelator able to form even stronger metal ion complexes stabilized by four five-membered rings (40). The TED polymer complex for a hexacoordinated metal ion may have the structure

268

JERKER

Higher aliphatic amines are already strong chelating ligands that can, by carboxymethylation, be further strengthened in their ability to form stable metal complexes; they may be very useful for efficient removal of metal ions from aqueous solutions. The linear aliphatic amines can be anchored in the polymer at different nitrogen atoms, which gives products containing a great variety of positional chelator isomers. Tris(2-aminoethyl)amine (TREN) is symmetrical and will, therefore, give uniform adsorption centers: , CH,-CH,-NH,

0

P NNH-CH,-CH~-N ‘CH,-CH,-NH,

Theoretically, this chelator is clearly tetradentate, but if it is tri- or tetradentate in practice is an open question. Thus, the TREN ligand may give rise to three fivemembered ring metal chelates. After exhaustive treatment with bromo- or chloroacetate, we obtain the CMTREN adsorbent ,CH,CooCH,-CH,-N /

P -NH-CH,-CH,-N O I CH2COO-

’ CH,COO-

\

/ CH2COoCH,-CH,-N

\

CH,COO-

This powerful chelator can efficiently remove any metal ions that may leak from any IMA adsorbent bed and may be used to replace EDTA in scavenging metal ions that are loosely bound to proteins. It may also remove heavy metal ions in biological extracts. IMA Adsorption

Center

When Cu2+ is used as a test ion to measure the adsorption capacity of an IMA gel based on 6% agarose, a close to stoichiometric relationship between chelator and metal is usually found. The chelators are not, however, uniformly occupied by the copper ions; some ions are weakly bound while others are strongly adsorbed. In extreme cases of multidentate chelators and multicoordinate metal ions, some ions may be bound so tena-

PORATH

ciously that their release from the gel requires degradation of the matrix of the adsorbent. For good performance, the weakly bound metal ions must be removed. This is done by a proper pretreatment of the gel, for example, by washing with solvents containing low concentrations (0.001-0.01 M) of imidazole or high concentrations (0.1-l M) of glycine. In the case of agarose gels, such a pretreatment usually alleviates subsequent leakage of the metal ions. Cross-linked agarose gels (57,58) tend to leak less than the other gels tested (59). Thus, the matrix itself is of importance in determining the properties of the adsorption centers. Commercial IDA gels usually adsorb 30-50 pmol of Cu2+ per milliliter of gel, and after proper washing, they contain 30-50% fewer metal ions than they do at the maximum load It may be mentioned that studies on the strength of adsorption by Hutchens, Belew, and Yip in my laboratory (60,61) have shown that apparent equilibrium dissociation constants for binding to Cu2+-IDA agarose are on the order 10-5-10-6 M for a number of model proteins and about 10e4 M for imidazole. Steady state equilibria are attained within minutes and linear Scatchaid plots are obtained. The protein affinity for the ion sequence Co2+, Ni2+, and Cu2+ increases with the number of d electrons and decreases for Zn2+ according to the Irving-Williams order (40). Cu2+ is eluted by glycine from the gel more easily than Ni2+. Imidazole is a strong unidentate ligand/ligate displacer. It is less prone to scavenge metal ions from the gel than glycine, but more effective in desorption of proteins from the immobilized dipositive ions of Co, Ni, Cu, and Zn. Imidazole at low concentrations (0.001-0.01 M) is, therefore, the eluting agent of choice for the removal of weakly adsorbed metal ions, and it is also excellent as a protein displacer when used . m either step- or gradientwise elution (0.001-0.1 M) protocols. Other imidazole and histidine derivatives may offer improved selectivity as displacers (unpublished observations). Chelating gels are ion exchangers. Charging the gels with metal ions alters drastically their adsorption properties, although at low ionic strength they still function as ion exchangers, but with altered charge characteristics (24,36,40). An increase in salt concentration to 0.1 M or higher decreases protein adsorption by ion exchange. For Co2+, Ni2+, Cu2+, and Zn2+, the IMAC adsorption capacity passes through a minimum and then rises again with an increase in the molarity of sulfate or chloride. To promote protein adsorption and to suppress ion-ion interactions, the equilibrium buffers should contain high concentrations of such salts (20.5 M) (1,40). While metal-binding solutes may profoundly affect the properties of the IMA adsorption sites, other solutes are inert or nearly so (40). Such solutes include nonionic surfactants, urea, ethylene glycol, and dimethyl sulfox-

IMMOBILIZED

METAL

ION

AFFINITY

ide. These substances are often used to solubilize membrane proteins and other hydrophobic proteins and peptides. There are many factors contributing to the heterogeneous nature of the adsorption centers, as revealed by the gradation in the strength of metal ion binding. Some of these factors are: 1. Irregular substitution of immobilized ligands or precursors to ligands (topological heterogeneity). 2. Different microenvironments (diversity of microenvironments). 3. Reversible reactions such as cross-linking directly involving metal ions and/or the ligandlligates. Simple calculations point to an average distance between the ligand moieties of about 12 A at a degree of substitution in the range 20-40 pmol per milliliter of 6% IDA agarose gels. Thus, there should be a sufficient density of the adsorption centers to allow the multipoint attachment of proteins. However, the underlying assumption is that the metal ion distribution within the network of the gel is uniform. This most likely is not the case. In an agarose gel, the polysaccharide helical chains

CH2-C @-

are intertwined and collected to form thick bundles or dense regions. Internally located ligands are accessible only to solutes whose molecular size permits the penetration of those bundles. Thus, in contrast to amino acids and peptides, proteins will be adsorbed only at the bundle’s surface, where the ligands are likely to be irregularly distributed and probably to some extent clustered. The differences in the nature of the network structures of various solid supports undoubtedly explain some of the dissimilarities in the adsorption behavior (59). The site heterogeneity may also depend on chemical reactions that involve either the metal ions or their surrounding ligandlligates. Since a metal ion may be oxidized or reduced, both forms may be found in a mixture at an equilibrium dependent on particular conditions. Although iron is one example where this is likely to be of some importance, tripositive immobilized iron is of great potential interest in IMAC. In water, it forms hexa-aqua ions which are easily hydrolyzed to an extent that depends on pH. Although the complex is immobilized, we may infer that the following structures are present in the Fe3+ IMA gels (Fe(III)-IDA polymer):

-+ 1

// 0

4

\

0 A ,O

269

CHROMATOGRAPHY

CH2--C 0% >FecOH, 0%

CH,--C\ 0

The character of the adsorbent may thus shift from that of an anion exchanger to a cation exchanger depending on pH. T13’ exemplifies another interesting tripositive ion (45). In free solution, it may be reduced to Tl+ by twoelectron transfer. Such a reduction is prevented by the immobilization. We have found that immobilized T13+ behaves similarly to Zn’+, Cu2+, and Ni2+ with respect to serum protein adsorption. This was unexpected and indicates that immobilized T13+ is stabilized against reduction by glycine that was used in equilibration buffer.

\ 0

+H+

“; CH,-CC

0

0

,0’ 0

OH ‘FeLOH ‘OH,2

I

j

These effects are extremely interesting, but the toxicity of thallium presumably precludes its use in biotechnology. Heterogeneity is also a consequence of the variation in ligand occupancy in the coordination sphere of the metal ions. The number of occupied coordination sites around a metal ion is likely to vary with time in a statistical manner due to stereochemical nonrigidity and vibration about an equilibrium position. A coordination site in the ligand that is formally saturated with the metal ion may be released from the metal ion at a partic-

270

JERKER

ular, almost infinitesimally short, moment, thus creating on the metal ion a potential attraction center for an electron donor atom on a protein. Consequently, protein may go in and temporarily “lock” the site. This explains why a metal, which should be saturated with excess ligand atoms, nevertheless may act as a protein adsorption center.

In order to discuss this concept, we may consider the interaction of CM-Asp, as the tetradentate ligand, and Cu’+, as a presumably tetracoordinated metal ion. The hypothetical complex is formally saturated. However, it still adsorbs certain peptides and proteins by coordination to their accessible histidine residues (imidazole side chains). This may be visualized by the following scheme:

//O \

CH,-C 0’P -N-Cu< AH-LO

PORATH

O \

\CH/

7 0

c\

0

CH,COO-

2

NH -C-Protein II CH CH

CH=C-Protein CH,COOIn a densely liganded gel a “flip-flop” mechanism could also occur between adjacent immobilized ligands where the metal ions serve to form transient or permanent bridges. With an increase in “dentatity” and coordination number, the multitude of combinations soon becomes staggering. The adsorption sites are less heterogeneous in ion exchangers than in IMA adsorbents. On the other hand, IMA adsorbents interact with a greater selectivity with a variety of specific combining sites on the molecular surface of proteins. This multitude of possible interactions makes it difficult to develop isocratic elution or displacement procedures for IMAC in certain instances. However, the inherent operational flexibility of IMAC (i.e., the variety of different kinds of ligands, metal ions, and modulators) may ultimately establish IMAC as a method superior to that of ion exchange chromatography. Selection of Matrix The supporting matrix should consist of a molecular network that is permeable to proteins. It should be strongly hydrophilic and chemically inert as well as microbiologically resistent under operating conditions. Ligand immobilization should be easy and give IMA products with lo-100 mmol of metal ions per liter bed volume. For good chromatographic performance, the

matrix should be available in the form of small (5-50 pm), uniform, rigid, spherical beads. Agarose and Sephadex, the first matrices used for IMAC, are useful for many applications, but too compressible for high-pressure IMAC. This latter weakness can be largely overcome by various procedures using reagents such as divinyl sulfone, which is an extremely effective “hardener” of agarose gels. Epichlorohydrin is also useful and stabilizes the helical segments in the agarose gel matrix by formation of short cross-links within the polysaccharide chain bundles (5758). This discovery has been put to practice in the production of commercially available agarose (e.g., Sepharose CL-6B and Superose 6 and 12). Other alternatives to higher cross-linked agarose have been introduced. TSK-gel chelate 5PW particles consists of a hydrophobic resin core covered by a hydrophilic layer of unknown composition (14). IDA coupled to hydrophilized silica was first described by Vijayalakshmi (9,62) and Small et al. (10). Such silica-based IMA adsorbents have been studied by Horvath’s (l5), Regnier’s (35), and Karger’s (64) groups. All support matrices have their special merits and problems. Some may allow higher flow rates than others, but they may be inferior in other respects such as showing undesirable adsorption properties or fast deterioration. Some are too expensive for routine or large-

IMMOBILIZED

METAL

ION

AFFINITY

scale use. On a small scale, they all appear to be acceptable. THEORETICAL

ASPECTS

Forces and Interactions A variety of questions should be answered in order to better understand IMAC and facilitate its further development. The most fundamental questions concern the quantum chemistry and kinetics of protein-metal ion interaction (40). To a considerable extent, we can use the present general knowledge in complexation chemistry. However, we must consider the steric restraints imposed by the microenvironment. Furthermore, the operational chromatographic conditions deviate from metal complexation studies conducted in solution. The adsorption centers do not consist of the metal ions alone, but of the metal-ligand complex and its surrounding. Therefore, it is not surprising that a number of unexpected phenomena such as aromatic interaction are encountered. Nevertheless, progress has been made. The equilibrium dissociation constraints for lysozyme and ovalbumin on IDA gels have been reported (60,61). Protein interaction has been interpreted in terms of steric restriction on the formation of the adsorption complexes, and Scatchard plots indicated a single type of interaction. Arnold and collaborators (2) have deduced expressions for partition coefficients in IMA partitioning in aqueous two-phase systems. A similar theoretical approach may be useful for IMAC. Protein retention in IMA beds is caused by electrostatic and hydrophobic factors. El Rassi and Horvath (15) have analyzed their retention data by applying the equation log k’ = A + B log m + C, where k’ is a retention factor, m is the salt molarity, A is a system-dependent parameter, B accounts for the electrostatic interaction, and C is a hydrophobic parameter. This equation may be useful for hard ions, including Fe3+. However, it lacks a term(s) accounting for the covalent character of the bond. Covalency and its dependency on all kinds of factors operating in IMAC are not easily quantitated. Covalent bond formation may often be the chief driving force for adsorption. Hydrophobic interaction is theoretically implied. Water is coordinated to metal ion and organized to a lesser extent in “the outer hydration sphere.” In addition, organized water surrounds the immobilized ligand, the spacer arm, and the polymer matrix. A large portion of these water regions is presumably involved in the hydrophobic interactions that contribute to the adsorption. The entropic gain due to water disorganization is particularly large when high concentrations of kosmo-

271

CHROMATOGRAPHY

topic salts such as sulfates and chlorides are used. This entropy increase promotes protein adsorption. When an immobilized metal ion and a nucleophile in a protein approach each other, electron orbital overlapping occurs, and, thus, a predominately covalent bound is formed. There is also ample experimental evidence that aromatic amino acids, free or incorporated in peptides, also contribute to the strength of IMA adsorption. Contribution of Amino Acid Side Chains to IMA adsorption. and Yip in my laboratory Belew, Nakagawa, (59,67,68) and Smith and collaborators (69) have made extensive peptide IMAC studies that convincingly reveal the role played by amino acid side groups and amino- and carboxyl terminals in binding. Understanding the finer details of the absorption behavior of amino acids and peptides may shed light on the mechanisms operating in the IMAC of proteins. Amino acid and peptide retention on Cu2+ IDA and Ni2+ IDA gels is shown in Table 1. Also, the peptide adsorption studies of Belew and Porath (59) revealed His, Cys, and Trp to be by far the most important amino acids involved in the coordination of Co --, Zn series of metal ions as we originally suggested (2). The VJV,” values for Phe, Phe-Phe, and Phe-PhePhe (17, 2.6, and 7.4, respectively) and Tyr, Tyr-Tyr, and Tyr-Tyr-Tyr (19.0,4.3, and 14.5, respectively) indicate that strong enhancement effects can be obtained only if the potentially cooperating side groups are interspaced by at least one amino acid residue. Presumably, the increase in adsorption reflects the involvement of at least two adjacent coordination sites on the metal ion. Dipeptides seem generally to be less retained than their component amino acids. We have found only two exceptions to this rule: Trp-Trp and His-His were adsorbed more strongly than their parent amino acids. Anomalous behavior which supports the concept of MIT from the adsorbent to the peptide was found (28,59). This usually undesirable phenomenon occurs when the metal ion has a greater affinity for the peptide than for the adsorbent. It may suffice to exemplify MIT with data from some studies with histidine peptides. The relative elution values, VJV,, for His-Gly-Gly and Gly-Gly-His were found to be 28.0 and 1.5, respectively, and for His-Tyr and Tyr-His to be 31.3 and 1.5, respectively. The low retention values indicate that Gly-GlyHis and Tyr-His strip metal ions from the gel and that the respective peptide-metal ion complexes are “adsorption inert.” Metal Ion Transfer in Protein IMAC MIT has been discussed by Sulkowski (28). Anderson et al. (70) found that MIT took place when human 4 V, is the volume.

elution

volume

for

the

peptide

and

V, is the

total

bed

272

JERKER

PORATH TABLE

Amino Ni-IDA-Sephadex

acid

Cu-IDA-Sepharose

> CySH

Elution

with with

with

Cu*‘-

> CySH

retention coefficients Cu-IDA-TSK-Chelate imidazole gradient His > TrP > Tyr pH

adsorbents

(pH

-

7, 20°C)

> Trp

> Asn

> Trp

> Tyr

Phe Gln > Tyr > Thr

>

Ser

> Arg

> Arg

Ser > Pro Asp

Lys > Leu Ile

Ala > Val

Asp > Glu

Phe > Asn > Met Thr

for individual amino 5 PW * 20 mM sodium > Phe > Gln

> Leu

Leu > Lys

>

Ala GUY Ile > Val Glu

acids as obtained by regression analysis phosphate containing 0.5 M NaCl, pH

> CyS-S-Cy

> Val > Ile > Arg

of capacity factors 7.0 (Yip, Nakagawa,

> Gly

> Asn

> Pro

> Ser 9 Asp

cys-s-cy > Leu Ile

> Asn

4 Asp

> Glu

of 52 peptides: Porath) > Glu

gradient: His

> Trp

> Tyr

> Gln

> Phe

> Pro

> Ser > Val

serum albumin passed through a Ni2+-TED column, but it could be prevented by using CL?+--TED as adsorbent for albumin. It is worth mentioning that the IDA chelator, as anticipated, binds metal ions less strongly than does TED or CM-Asp chelator. From these experiments, we learn that a strong metal-binding chelator should be selected when there is a risk of MIT. Notably, carboxymethylated tetraethylene pentamine and CMTREN are very strong chelators. Gels containing these ligands are likely to remove any trace of metal ions that may have leached or been stripped from a protein or peptide complex. MIT with metal-free TED may be used for extremely effective demetallation of metal-dependent enzymes such as carboxypeptidase A, as shown by Muszynska et al. (71). IMAC on an Zn’+-IDA gel can be used to reactivate the apoenzyme in a controlled manner. OPERATIONAL Preparation

and Ni*‘-IDA

6B (Belew-Porath) His

Elution

interaction

(Hemdan-Porath) His

Relative

side chain

1

CHROMATOGRAPHIC

CONDITIONS

of the IMA Bed

The immobilized metal ion adsorbents may be prepared by charging the chelating gels with a slightly acid solution of the metal salt (pH 3-5). In the case of Fe3+IDA gels, the adsorbent may be prepared by washing a bed with 3-6 bed volumes of freshly prepared 50 mM FeCl, in 0.01 M HCI, followed in sequence by large volumes of water, the buffer to be used later for elution, and, finally, equilibration with the starting buffer. Washing may be done on a filter under suction to speed up the pretreatment. Charging the gel under acidic conditions is essential to avoid formation of ferric hydroxide particles in the solution. As mentioned earlier, to avoid leakage of weakly adsorbed Co2+, Ni2+, Cu2+, and Zn2+, the bed may be

> Gly

washed with glycine (0.1-l M) or imidazole (l-100 InM) before the equilibration step. (The higher concentrations may be used if the metal ions are very strongly adsorbed to the chelator.) The column can often be regenerated while metal ions are still immobilized. However, it is usually possible, and sometimes safer, to remove the metal ions by elution with 0.05-0.1 M Na-EDTA, pH 7-8. If the chelator ligand is binding the metal ion very strongly, complete stripping of the metal may not be possible. In such a case, it is necessary to devise another regeneration scheme that will restore the adsorbent to its original conditions. Brief washing of the gel with acid (pH 1-3) is likely to remove any adsorbed soluble metal, or other solutes, from IDA gels. Conditions

for Adsorption

and Desorption

The selection of conditions depends on the kind of immobilized metal ion used, whether hard (Fe3+), borderline (Co2+ --* Zn2+), or soft (Hg2’, Pd2’). Guidance can be obtained by consulting the original articles. Kagedal’s review article (30) gives detailed recommendations. There is wide latitude in the selection of conditions suitable for adsorption. For orientation, a pilot experiment may employ the following conditions: For immobilized Fe3+, use an equilibrium buffer such as acetate or Mes of low pH (pH 5.6) and low ion strength (0.05). Increasing the pH and/or ionic strength will then elute the proteins. The following buffers may be recommended (51): (i) 50 mM Mes-NaOH, pH 6.0, the equilibrium buffer followed in sequence by (ii) 50 mM Mes + 1 M NaCl, pH 6.0; (iii) 50 mM Mops-HCI, pH 7.2; (iv) 50 mM Tris-HCl, pH 8.0; or (v) 50 mM POPSO, pH 7.2-8.5. Weak phosphate buffers have also been used and are sometimes pre-

IMMOBILIZED

METAL

ION

AFFINITY

ferred to “Good’s buffers.” If some protein remains bound, it may be advisable to include 0.5 or 1 M NaCl in every buffer. The protein left on the column can always be rescued with 0.2 M Na-EDTA, pH 7.0. Some proteins may be eluted by inclusion of phosphate or magnesium chloride in the buffer. SELECTED

Immobilized

FIELDS OF APPLICATIONS Co2+, Ni”,

Cu2+, and Zn2+

Role of histidine. Whether free in solution or immobilized on a solid support, these metal ions show apreferential affinity for histidine among the amino acids present in protein hydrolysates (65). The dominance of histidine governing the adsorption is also apparent for peptides (59,67-69) and proteins (66). Sulkowski (21) was the first to accumulate evidence for the concept that histidine residues are the main contributors to the adsorption. Immobilized Ni and Cu are now frequently used for single-step purification of recombinant proteins containing tails with histidine residues. These applications were recently reviewed by Ford et al. (72). Two or more internal histidine residues, located on the molecular surface in an a-helix, may cooperate to form very strong metal ion complexes. This important chromatographic behavior was first suggested by Sulkowski (28) and experimentally verified by him and his collaborators as well as by Arnold and her collaborators (73). To indicate the importance of molecular surface-located histidine in IMAC, I select for illustration the work carried out by Hemdan et al. (66). In a methodological study, they used a sample of three cytochromes c in an artificial mixture and an IDA agarose (chelating Sepharose 6B), an equilibrium buffer of pH 7.0 consisting of 20 mM sodium phosphate, 1 M NaCl, occasionally supplemented with 0.001 M imidazole. The tuna cytochrome, devoid of external histidine residues, passes through the column unretained on a Cu’+-IDA column. The horse heart and yeast (C. krusei) cytochromes have one and two histidines residues, respectively, available for adsorption. They are consequently eluted in that order using a pH gradient (Fig. 4). The cytochrome c of horse heart, in contrast to that of C. krusei, is not adsorbed on the CL?+-IDA column when 1 mM imidazole is included in the equilibrium buffer. These two cytochromes c can be isocratically resolved on a Ni’+-IDA gel but the separation is much less efficient than that of Zn2+-IDA and Co2+-IDA, thus showing decreasing strength of adsorption in the order Cu2+ > Ni2+ > Zn2+, Co2+. IMAC in salts and denaturants. In a methodological study of human serum protein adsorption on Ni2+-TED agarose (40) we made a number of discoveries important for the development of IMAC (see original paper for details.) As seen in Fig. 5, the proteins adsorbed in

273

CHROMATOGRAPHY

g 1.5

7.0

0

it G w 1.0 Y $ 8 3

6.0 I, 5.0 f

.5

0

4.0

20

40

60 EFFLUENT

60

loo

120

(mL)

FIG. 4. Resolution of cytochromes c on sample of protein (5 ml) was applied and with 20 ml of the equilibrating buffer (pH 0.1 M sodium acetate/l M NaCl, pH 6.0, (3) gradient formed by mixing 25 ml of 0.1 M pH 4.0, into 25 ml of 0.1 M sodium acetate/l finally, irrigated with 0.1 M sodium acetate/l heart; 0, horse heart; A, C. krusei.

the IDA-Cu(I1) column. A the column was rinsed (1) 7.0) and (2) with 25 ml of then developed with a pH sodium acetate/l M NaCl, M NaCl, pH 6.0; and (4) M NaCl, pH 4.0.0, tuna

“buffer B,” 0.1 M Tris-HCl, pH 8.1 (chiefly hemopexin and a,-macroglobulin), are the same whether the experiment is performed in the absence or presence of sodium chloride up to its saturation. The behavior in the presence of Na,SO, is different. Chromatograms of high resolving power can be run in the presence of ionic surfactants, urea, guanidinium chloride, and certain hydro-organic solvents, but frequently show lowered capacity. Interestingly and importantly, neutral surfactants like Tween 80 tend to increase capacity and do not much change the selectivity, even if the concentration is raised beyond the point of micelle formation. An IMAC study of lactoferrin and transferrin made by Hutchens and Yip (74) may serve to demonstrate chromatography in urea solutions. Copper ions were loaded on an IDA-TSK chelate 5PW gel by passing 50 mM CuSO, through the bed followed by 0.1 M sodium acetate containing 0.5 M NaCl (pH 5.0) to remove the weakly adsorbed copper ions. Three columns were equilibrated with, respectively, (A) 20 mM sodium phosphate (pH 7.0) containing 3 M NaCl and 3 M urea, (B) 100 mM sodium acetate (pH 5.5-5.0) containing 0.5 M NaCl and 3 M urea, and (C) 20 mM sodium phosphate or Hepes (pH 7.0) containing l-20 mM imidazole. After the samples had been introduced, chromatograms were developed by pH or imidazole gradients with some of the eluents containing 3 M urea. Iron-free and iron-saturated lactoferrin could be chromatographically resolved by a pH gradient in 3 M urea solution, but not by an imidazole gradient. However, the iron-saturated and iron-free transferrin could be resolved only by the imidazole gradient at neutral pH. The authors concluded

274

JERKER

aI1

2

3

4

1M 5M S NaCL NaCl

b)

1

2

3

4

0.5M 1 M 0.1% S Na,SO, Tween 80

FIG. 5. Gradient gel electrophoretograms of acetate-eluted peaks from Ni2+-TED column chromatography performed with different equilibration buffers containing buffer B and various concentrations of salts. The included characteristic solute species are indicated below the diagrams. (a) Effect of NaCl concentration. (b) Effect of Na,SO, at different concentration and 0.1% Tween 80. S denotes the serum sample.

that the affinity for IDA-immobilized copper ions in urea increased in the order apolactoferrin < hololactoferrin < copper-saturated lactoferrin < diferric or holotransferrin < monoferric transferrin < apotransferrin < copper-saturated transferrin. Thus, IDA-immobilized Cu2+ can distinguish ligand-induced alterations in the protein surface structures, as had also been shown by Sulkowski in urea-free buffer systems (33). High-performance IMAC of proteins and peptides. High-performance IMAC has been reported in several articles (9,10,14,64,67,75,76). For example, in a recent study of HP-IMAC, Yip et al. (68) tested the power of the method by using a Cu(II)-TSK 5PW column (8 X 75 mm) to separate members of the angiotensin family of peptides (3-9 pg of each peptide). The sample was loaded in 20 mM sodium phosphate buffer, pH 7.0. The chromatogram was developed by a pH gradient. Figure 6 shows an excerpt from the chromatogram. Only about 1O-3 of the total adsorption capacity of the column has been utilized.

PORATH

groups play only a secondary role when the experiments are carried out near neutrality at high ionic strength. As pointed out earlier, at low ionic strength the Fe3+ gel functions either as a cation or as an anion exchanger, depending on the pH chosen. Proteins containing no phosphate groups or without excessive content of Glu or Asp residues may occasionally be adsorbed coordinatively, as illustrated in the following example, which also demonstrates the efficient upscaling IMAC. Thus, Chaga et al. (77) have shown that glycogen phosphorylase and lactic dehydrogenase (LDH) from chicken breast muscle extracts may be rapidly isolated in a single run on a Fe 3+-IDA column with activity recoveries well over 90%. As shown in Figs. 7a and 7b, essentially the same efficiency in separation was obtained on a milligram and a gram scale. Binding studies with peptides from tryptic digest of LDH suggest that the IMA adsorption is due in part to a superimposed effect of several histidine residues located in a short peptide segment (unpublished results). The Hard Metal Ions Relative metal ion affinities. These were tested on a four-step tandem column containing IDA gel charged with A13+, Ga3+, In3+, and T13+ from top to bottom (45). The chromatogram of serum proteins was developed in a neutral buffer. Significantly, the proteins were found to be adsorbed only in the thallium ion section. Thus,

A220 0.16

0.12

0.06,

0.04

I

H .6.0

0, Immobilized

Fe3’

Immobilized tripositive iron as an IMA adsorbent has been extensively studied by Muszynska, Andersson, Dobrowolska, Chaga, and myself. These adsorbents can be used for selective adsorption of phosphoproteins and phosphopeptides (50,51). They can also be used for nucleotide separations (52) and in such a study we demonstrated that the presence of a primary phosphate ester group was a requirement for adsorption. Carboxyl

-

Time (mid

FIG. 6. HP-IMAC of the angiotensin family tensin II: Asp-Arg-Val-Tyr-Ile-His-Pro-Phe; angiotensin II; B, [Sari-Iles] angiotensin II; C, II; D, [Gly-[Ilea-Vals] angiotensin II; E, [Val’] His’] angiotensin II; H, [@-Asp’] angiotensin II. al., Anal. Biochm. 183, 159-171 (1989).

of peptides. F, angioA, [Sar’-Va15-Alas] [des Asp’] angiotensin angiotensin II; G, [D*Redrawn from Yip et

IMMOBILIZED

METAL

ION

AFFINITY

b

10

50

30 Effbent

(nd)

275

CHROMATOGRAPHY

.4

60

120 The

160

(mid

FIG. 7. Demonstration of scaling up IMAC. (a) Small-scale isolation of muscle enzymes. Adsorbent, Fe3+-IDA agarose; column size, 5 X 1 cm (3.93 ml); extract volume, 15 ml; flow rate, 30 ml/h; room temperature; equilibration buffer, 0.02 M Tris-HCl, 1 M NaCl, pH 7.7. After a wash with equilibration buffer, starting at the volume indicated by arrow 1, the elution was carried out (arrow 2) by 0.02 M Mes/NaOH, pH 6.5 (no NaCl), followed (arrow 3) by 0.02 M sodium phosphate, 1 M NaCl, pH 6.4. Peak II contained a single component identified as glycogen phosphorylase; peak III contained pure or nearly pure lactic dehydrogenase. None of these enzymes were found in peak I. (b) Large-scale isolation of muscle enzymes. Adsorbent and equilibration buffer as described for (a) column size, 5 X 11.3 cm (200.8 ml); applied extract volume, 2000 ml; flow rate; 4000 ml/h; room temperature. The column was washed in succession with (1) equilibration buffer; (2) 0.02 M Mes/NaOH, pH 6.5; (3) 0.02 M sodium phosphate, pH 6.8; (4) 0.02 M sodium phosphate, 1 M NaCl, pH 6.4, and finally for regeneration of the chelating gel, 0.1 M EDTA. Two glycogen phosphorylases were discovered; one in peak II with M, - 400,000 and the other in peak III with it& 200,000. The activity yield was 97% and no contaminants were discovered. Peak V contained 96% of the original lactase dehydrogenase (in pure form), while peak IV contained only 1% of lactage dehydrogenase but no protein was detected by SDS electrophoresis.

the experiment under neutral conditions revealed the soft or borderline character of tripositive thallium, the only metal ion in the series to possess extreme toxicity. In contrast, at pH 4.5-6.0, all the column sections adsorbed the serum proteins in appreciable quantities, showing that all the immobilized metal ions in question functioned as an ion exchanger within this pH range. Immobilized Ca2+ chromatography. We prepared CM-Asp agarose) as an adsorbent for calcium ions in the hope that it would balance the requirement for sufficient residual affinity for calcium binding proteins. Mantovaara et al. (48) used CM-Asp agarose to purify factor VIII:c coagulant activity from a cell culture medium. The factor was purified to electrophoretic homogeneity in a single step (see Fig. 8 and Table 2). All activity was concentrated in one elution peak and the yield was 150% and the purification 85-fold. Thus, we demonstrated that this kind of IMAC favorably compares in efficiency with biospecific affinity chromatography. This is even more evident in the following example. To isolate serum amyloid P component (SAP) or 9.5s a-Glycoprotein (47), a suspension of CM-Asp agarose beads (6% matrix) in distilled water was degassed under vacuum and packed in a bed, 1 X 13 cm. The adsorbent was washed with 5 column volumes of 0.25 M CaCl,, pH

9.0, followed by extensive washing with 20 mM imidazole-HCl, 0.25 M in CaCl,, pH 7.0. A 40-ml sample from 10 ml of human serum diluted with CaCl, solution to 0.25 M in Ca2+ and adjusted to pH 7.0 was introduced in the column at a flow rate of 20 ml h-’ (which was maintained during further development of the chromatogram). Proteins with no affinity for the gel passed through the bed in the washing with equilibration buffer (fraction I). Adsorbed proteins were eluted by 20 mM Na-acetate, pH 5.0 (fraction II), followed by 0.1 M NaEDTA/NaOH, pH 8.0 (fraction III). The SAP protein was collected in the acetate eluted fraction, fraction II, while washing with EDTA served as a step toward regeneration of the adsorbent. A 1900-fold purification was achieved. Analysis of the protein in fraction II showed a single band on both gradient and SDS-polyacrylamide electrophoresis. The protein was identified as SAP by immunoprecipitation. This fractionation protocol deviates from those usually used in IMAC in that chromatography was carried out with the gel-immobilized metal ion included at a very high concentration in the mobile phase. This procedure made it possible to isolate the protein in its native form without dissociation of the molecule, which consists of 10 noncovalently bound subunits. A mechanism for the IMA interaction was proposed (47).

276

JERKER

PORATH

1

c in-v Za’*-DPAapos

,‘I\ ,~,~

---------

w-w-

6 II I

scpbdex c 23

III I

10 20 Fraction number

ccs g-

I

--

30

DW

FIG. 9. Schematic diagram of the setup of the tandem columns. The five columns to the left were connected in series at the start of the experiment. The columns were equilibrated with 50 mM sodium phosphate, 0.5 M in NaCl, and 0.5 M in K,SO, (pH 7.6), and 10 ml of human serum (equilibrated in the same buffer) was applied. The beds were washed until the A,, of the effluent reached the baseline. The beds were then disconnected and developed separately (see text). Solid arrows indicate liquid flow and dashed arrows indicate disconnection of a bed followed either by elution of adsorbed material directly or by elution into a new tandem column. The letters and numbers refer to the fractions collected and the buffers used, respectively.

30 ml of the C&-S gel for serum albumin. The use of these two gels sharpens the group separation of the minor proteins on the three IMA beds (6 ml total volume each). The tandem column was equilibrated with 50 mM sodium phosphate buffer containing 0.5 M NaCl and 0.5 M K,SO, (pH 7.6). The salt suppresses the ionic interactions on the IMA gels but increases adsorption capacity, presumably by promoting hydrophobic and coordinate covalent bonding of the proteins. (It should be noted that thiophilic adsorption is promoted only by the sul-

TABLE

Material Cell culture medium Negative control Peak I Peak II Peak III

Protein concentration hdml) 4.09 0 0.625 0.024 0.092

2

VIII:c

Coagulation

Activity Factor

Volume (ml) 4.2

12.5

D “II

SephdaCU

Application of CASMAC to serum fractionation. A CASMAC scheme (Fig. 9) has been devised for efficient group fractionation of human serum proteins (43) and other complex protein mixtures. For serum fractionation we used three IMA gels supplemented with a hydrophilic and a thiophilic gel. Extension of the CASMAC tree was made by Sephadex for on-line buffer change followed by HPLC on Mono Q (prepacked columns from Pharmacia, Uppsala, Sweden). The choice of the volume of each bed in the tandem column (the “stem” of the “CASMAC” tree) was based on the results of a series of preliminary experiments in which affinity and capacity for specific major proteins were roughly estimated. For a lo-ml sample of serum we used 60 ml of T gel to adsorb the immunoglobulins and

of Factor

Mom q

-------

FIG. 8. Calcium affinity purification of factor VIII:c coagulant activity. Adsorbent, CM-aspartic acid agarose pretreated by five column volumes of 25 mM CaCl, followed by the same volume of distilled water, column dimensions, 14 X 1 cm; sample, 4.2 ml centrifuged rat liver nonparenchymal cell culture medium was diluted with 5.8 ml of equilibration buffer and made 0.5 M in K,SO,; flow rate, 20 ml/h; fraction volumes, 2.5 ml, equilibration buffer, 20 mM Na-acetate/acetic acid, pH 5.5, containing 0.5 M K,SO,. Elution was performed with 20 mM Na-acetate/acetic acid (no sulfate) (first arrow) followed by 0.1 M Na-EDTA/NaOH, pH 8.0. The factor VIII:c coagulant activity was located in peak II. The activity yield was 150%. Eighty-five-fold purification was achieved and the peak II material revealed a single band in polyacrylamide electrophoresis.

Determination

:;:; 4

IU/lOO

pl

1.6 X 1O-3 0 0 7.8 x 1o-4 0

VIII:c

IU/mg

activity protein

3.9 x 1o-3 0 0 3.3 x 10-l 0

Total

IU

Yield (%)

6.5 X lo-’

9.7 x lo-*

150

IMMOBILIZED

METAL

ION

AFFINITY

CHROMATOGRAPHY

277

fate, which makes possible independent elution of hydrophilic and thiophilic proteins from a composite column by manipulating the salt composition of the eluent. This possibility, however, was not used in the experiment referred to here.) Virtually all of the proteins in the lo-ml sample were adsorbed. The breakthrough fraction contained only 0.6% of unidentified proteins or peptides (based on absorption at 280 nm). After the tandem column was dismantled, further fractionation was accomplished as indicated by the CASMAC scheme. The very high efficiency of separation in this model experiment is demonstrated in Figs. 10 and 11 and in Table 3. In this bench-scale, on-line fractionation of the serum, we obtained in a reproducible way transferrin (fraction F VI) and serum albumin (E III and E IV) at nearly 100% purity. a-Macroglobulin (B III) was isolated at more than 98% purity. Practically all ceruloplasmin was obtained in a single fraction (rear part of D VII), and all antitrypsin became trapped in Ni(II)-DPA agarose together with the major part of transferrin. All minor components of the serum were also effectively concentrated and purified, although further fractionation would be necessary for their isolation. The recova

400

Fraction

500

numbw

b A 280 2

F#-

J

1’ I

4

VII b

FIG. 10. (a) CASMAC chromatogram: A is the material passing the tandem columns, B is the material from the Zn’+-TED gel, C is the T gel, D is the Zn’+-DPA gel, E is the Cs-S gel, and F is the material from the development of the Ni’+-DPA agarose. (b) Expanded part of (a). See (43) for a detailed description of the columns and elution conditions.

FIG. 11. Electrophoresis of the materials from the chromatogram in Fig. 10. Pharmacia amide gradient gels were used.

in the peaks collected PA A4130 polyacryl-

ery was excellent (98.2%) and most likely quantitative (the 1.8% not accounted for can most probably be attributed to inaccuracy in measurement or to postcolumn losses). Some general comments regarding the CASMAC fractionation may be justified. The results confirm earlier findings (40) showing that by variation of the kind of metal ions [in this case Zn2+ and Ni2+ and the chelating ligand (in this case TED and DPA, dipicolylamine)] different protein-IMA affinities and capacities can be uti-

JERKER

278

PORATH

TABLE Amounts

and Recoveries

of Some Serum

Proteins

3

in the Different

Fractions

from the Development

of the Tandem

Columns

Serum protein (mg/liter) Subfraction

q-Macroglobulin

Albumin

AI B II

Transferrin

Ceruloplasmin

K light chain

X light chain

13 670 2.2

14

(30

Db VII Da VII

385 470

0.02 tl

114 344