Bkconlugete Chem. 1991, 2, 275-280
275
Specific Interchain Cross-Linking of Antibodies Using Bismaleimides. Repression of Ligand Leakage in Immunoaffinity Chromatography+ Michel Goldberg,’ Kaja L. Knudsen, David Platt, Fortune Kohen,* Edward A. Bayer, and Meir Wilchek Department of Biophysics and Department of Hormone Research, The Weizmann Institute of Science, Rehovot, Israel. Received May 2, 1991 The extensive use of antibody-containing affinity columns in the purification of biologically active compounds (e.g., genetically engineered proteins) is severely hampered by the leaching of antibody (or portions thereof) from the immunoaffinity resin during elution of the target antigen. One of the major problems in this context is the combined use of reducing (i.e., thiols) and chaotropic (e.g., detergents and denaturants) agents in the elution step, which causes the disassociation of heavy and/or light chains from the immobilized antibody, thereby contaminating the resultant product. In order to overcome this problem, we have cross-linked the four antibody chains at their sites of disulfide interlinkage, thus producing a single antibody chain. To accomplish this, interchain disulfide bonds were reduced, and the resultant thiol groups were cross-linked by using bifunctional SH-specific reagents (particularly bismaleimides). Cross-linking of up to 95% of the available SH groups produced was achieved with concomitant retention of antigen-binding activity. The cross-linked antibody was immobilized onto CNBr-activated Sepharose, and the resultant column was found to be substantially more stable to harsh elution conditions than similar columns which contain the un-cross-linked antibody.
INTRODUCTION In affinity chromatography, antibodies are commonly used as immobilized ligands to purify specifically, and in one step, many kinds of antigens (Wilchek et al., 1984; Mohr & Pommerening, 1985). The recoveryof the purified compound occurs during the elution step, which is performed under conditions which cause a decrease in the affinity between the antibody and the antigen. The purification process thus reflects the biological recognition which takes place between the antibody and antigen and, when successfully applied, is much more efficient than conventional procedures for the purification of biologically active materials. An important drawback of immunoaffinity chromatographic purification is the partial chemical or enzymatic hydrolysis of the antibody which can result in a leakage of the antibody or portion thereof (e.g., light and/or heavy chains) which then contaminates the affinity-purified product (Dean et al., 1985). One source of such leakage involves the chemistry of immobilization of an antibody to the resin and has been dealt with extensively in previous studies (Wilchek et al., 1975;Wilchek & Miron, 1987). Another source of leakage involves the presence of reducing agents commonly used in many of the purification steps (Linn, 1990;Deutscher, 1990). These may sever the interchain disulfide bonds which connect the four chains of the immunoglobulin molecule; those of which not directly coupled chemically to the resin may then be released. Many methods have been suggested to avoid the problem of antibody leakage (Ernst-Cabrera & Wilchek, 1988), including the intramolecular cross-linking of the antibody and the selection of conjugation methods that yield a more stable chemical linkagebetween the antibody and the resin. For example, glutaraldehyde has been used both as a t Abbreviations used BMH,N,”-bismaleimidohexane; MPHPD, N,”-bis(3-maleimidopropionyl)-2-hydroxy-l,3-propanediamine; PDM, NJV-o-phenylenebismaleimide;ELISA, enzymelinked immunosorbent assay; PBS, phosphate-buffered saline, pH 7.4; Fv,the variable region of the light and heavy antibody chains. Department of Hormone Research.
*
general reagent to cross-link between the chains of the immobilized antibody (Kowal & Parsons, 1980)and, in a more specific case, to cross-link (and thus stabilize chemically) antibody molecules which have been bound to protein A columns (Gyka et al., 1983). Cross-linking with glutaraldehyde usually involves lysine groups of the antibody, some of which may occupy or indirectly affect the combining site and thus the binding activity of the resultant cross-linked immunoaffinity column would have been reduced. The antibody combining site is located in the Fv portion of the antibody molecule. The Fv, which includes the variable parts of both the heavy and light chains, contains lysine residues, but the interchain disulfides are located elsewhere in the antibody molecule (Le., in the constant regions). The two chains of the Fv have recently been prepared by genetic engineering, either as a single chain linked together by a designed peptide (Bird et al., 1988) or separately as double chains which associated spontaneously after their production (Ward et al., 1989). The engineered bacterial proteins expressed high levels of biological (antigen-binding) activity. This shows clearly that disulfide groups are not important for recognition and that the chains can reassemble themselves in the correct orientation. In this context, the purpose of this study was to prepare a stable single-chain equivalent of a native antibody by cross-linking the interchain thiol groups via noncleavable bonds. The results have demonstrated the feasibility of using bismaleimides (Figure 1)for this purpose. To this end, the disulfide bonds of antibiotin antibodies were reduced, cross-linked with the desired reagent, and immobilized onto CNBr-activated Sepharose, and the resultant immunoaffinity column was used to isolate biotinlinked proteins. This approach served as a model system to demonstrate enhanced stability which can be achieved for an immobilized, cross-linked antibody. EXPERIMENTAL PROCEDURES Materials. NJV-Bismaleimidohexane (BMH) was a product of Pierce, Rockford, IL. NJV-Bis(3-maleimidopropionyl)-2-hydroxy-l,3-propanediamine and NJV-
1043- 18021B 112BQ2-Q275(bQ2.5Q/Q0 1991 American Cbmicai Society
276 Bloconlugete Chem., Vol. 2, No, 4, 1991
Goldberg et al. 0
Br CH,-C
II
-CH,Br
Dibromoacetone
OH
BMH
I
Br CH,-CH-CH,Br
- (CH,
),-
0 OH 0 C-NHCHZCHCH,NH-C-(CH,)~-N I1 I II
Dibromopropanol 0
Br CH,-C
MPHPD
II
- NH (CH,),
0
II
NH -C -CH,Br
N,N'-Dlbromoacetylethyiene diamide
C1 CI$-CH,O/CH, Epichlorohydrin
PDY
Figure 1. Chemical structures of bifunctional reagents used in this study.
o-phenylenebismaleimide (MPHPD and PDM, respectively) were obtained from Sigma ChemicalCo. (St.Louis, MO). Dibromoacetone (DBA) was purchased from K & K Laboratories (Cleveland, OH). Ferritin (2X crystallized) was from Pentex Biochemicals (Kankakee, IL). 2,2'-Azinobis(3-ethylbenzthiazolinesulfonic acid) diammonium salt (ABTS), and all other reagents, proteins, chemicals, and biochemicals were obtained from Sigma. N P - D i bromoacetylethylenediamide was prepared by reacting diaminoethane with a 3-fold excess of bromoacetic anhydride in dioxane. The product was precipitated with water, filtered, and dried in vacuo. Purification of Antibodies. Antibiotin antibodies were elicited in mice against biotinylated bovine serum albumin mainly according to Dakshinamurti and Spectro (1990) following a procedure described by Kohen et al. (1982). The antibodies were purified from ascites fluids on a protein A-Sepharose CL-4B column (Pharmacia, Uppsala, Sweden) according to the manufacturer's instructions. Briefly, the ascitic fluids were diluted with an equal volume of glycine buffer (1.45 M glycine, 3 M NaC1, pH 8.9) and passed through the protein A column (1mL of ascites per 3 mL of gel) preequilibrated with the same buffer. After a 15-minincubation period, the column was washed with glycine buffer until the absorbance (280 nm) of the eluate approached a minimum value. The elution of the monoclonal antibiotin antibodies was performed with sodium citrate buffer (40 mM citric acid, 20 mM NaC1, pH 3.2). The antibody preparation was dialyzed overnight a t 4 "C against phosphate-buffered saline, pH 7.4 (PBS), resuspended a t a concentration of 1.5-2 mg/ mL and stored at -20 OC. Reduction of Antibodies. An antibody sample (1.52.0mg/mL) was reduced for 1h a t 37 OC with 2 mM dithiothreitol (Cleland, 1964)accordingto Gunewardena (1966). The reduced antibody was dialyzed against 5 mM EDTA in PBS (presaturated with Nz)in a stirred ultrafiltration cell (Amicon, Danvers, MA) using an XM 50 membrane. The ultrafiltration step (three washings, 50 mL of buffer per wash) was performed until no trace of dithiothreitol could be detected in the ultrafiltrate solution as measured spectrophotometrically with 5,5'-dithiobis( 2-nitrobenzoic acid) (Ellman, 1959). Cross-Linkingwith Bismaleimides. Antibodieswere cross-linked by mixing the reduced antibody solution with a preparation of the desired cross-linker, diluted (as
indicated later in the text) in a dimethylformamide-water (1:l)solution. The volume of the antibody solution was 3-5-fold that of the reagent solution. For example, a solution (300 pL) of antibody (10.8 rM, 1.6 mg/mL of PBS containing 5 mM EDTA) was mixed with a solution (100 pL) of 20 mM MPHPD (8 mg/mL in PBS-DMF, 1:l). In this particular case, the ratio of reagent to interchain disulfide bond was 120. The reaction was carried out overnight at room temperature. Measurement of Binding Activity. An ELISA test, based on the method described by Eshhar (19851, was used to compare the binding of the native and bismaleimide-cross-linked antibody. Briefly, 100-pL aliquota of biotinyl-N-(c-aminocaproy1)-BSA(50 rg/mL in PBS) were incubated in a 96-well poly(viny1 chloride) microplate (Flow Laboratory, Irvine, Scotland). After washing (Wash Concentrate, Delfia, Turku, Finland), 100 pL of antibody preparation at the desired dilution (in PBS as indicated later in the text) was introduced to the wells and incubated for 1 h at room temperature. After washing, a sample (100 pL) of protein A conjugated horseradish peroxidase (3800 units/mL; BioMakor, Rehovot, Israel) was diluted 400 times in PBS (pH 8) and incubated a t 37 "C for 30 min. Following another washing step, 100 pL of freshly prepared substrate solution was added [ l mg/mL ABTS and 0.003% H202 in citrate-phosphate buffer (28 mM citric acid, 44 mM disodium phosphate, pH 4.7)]. The results were read at 600 nm with a Microplate Autoreader (BioTek Instruments, Winooski, VT). Miscellaneous Methods. The concentration of protein in solution was measured by the Bradford method (Bradford, 1976) using ovalbumin as a standard. Iodination with 1261 was performed according to the Chloramine T procedure (Bolton & Hunter, 1986). Amino acid analysis of native or cross-linked antibodies was performed by using a Dionex amino acid system (Dionex Corp., Sunnyvale, CA). Proteins (e.g., antibodies or avidin) were immobilized on CNBr-activated Sepharose following the previously described method (Kohn & Wilchek, 1984). Biotinylation was performed either with the N-succinimide ester of biotin or the long-chained N-c-aminocaproyl derivative as previously described (Bayer & Wilchek, 1990). The extent of biotinylation was assessed on an avidin column by measuring spectrophotometrically the percentage of the biotinyl-protein preparation
Interchain Crosslinking of Antibodies 1 2 3 4
5 6
1 2 3 4
5 6
Mr(KDa)
-
H2L2 H L H: HL
2, NO. 4, 1991 277
BiOCOnlUgat8 Ch8m., VOl.
-
50 125 00 75
-
50
I; -1
-
BMH MPHPD PDM Figure 2. SDS-PAGE of bismaleimide-cross-linked antibodies. Anti-biotin antibody samples were reduced with dithiothreitol and cross-linked with various concentrations of the designated bismaleimide reagent. Samples 1-6 represent 0.2-, 1-,5-, 24-, 120-, and 600-fold molar ratios, respectively, of the given reagent to the calculated amount of interchain disulfide bonds in the antibody. Following ultrafiltration, the samples were subjected to electrophoresis under reducing conditions on 10% polyacrylamide gels and stained with Coomassie Brilliant Blue. H2Lz marks the position of the cross-linked intact antibody (two heavy and two light chains); H2L, two heavy chains coupled to one light chain; H2, two crosslinked heavy chains; HL, a light chain coupled to a heavy chain; H, an uncoupled heavy chain; and L, an uncoupled light chain.
which bound to the column. For biotinyl-ferritin, a solution containing 1mg/mL (2.1 nmol; A280 = 13.9) was added to a 0.4-mL avidin column (containing 2 mg of avidin/mL of resin; Le., 12.5 nmol total). The extent of binding was determined by assessingthe amount of protein which appeared in the effluent fractions. Polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate (SDS-PAGE) was performed on slab gels according to Laemmli (1970). Samples were electrophoresed on 10 or 12% gels under either reducing (in the presence of P-mercaptoethanol) or nonreducing conditions (as indicated later in the text). The proteins were stained with Coomassie Brilliant Blue R250. Densitometry measurements were conducted on photographic transparencies of the desired lane of an SDSPAGE gel. The measurements were performed on a computing densitometer apparatus (Model 300A, Molecular Dynamics, Sunnyvale, CA), using Image Quant Software, Program 3.0. RESULTS AND DISCUSSION
Cross-Linking with Bifunctional Thiol-Specific Reagents. In order to cross-link antibodies via sulfhydryl groups, the interchain disulfides of the antibody have to be reduced. We found that a concentration of 2 mM dithiothreitol appears to reduce the interchain disulfides, whereas intrachain cystine linkages appeared to be unaffected by such treatment. Antibodies, reduced in this manner, were then subjected to ultrafiltration under reducing conditions(nitrogengas in the presence of EDTA) and immediately cross-linked with the desired reagent (Figure 1). After the reaction (employing different concentrations of the respective reagent), the antibody preparation was examined by SDS-PAGE under reducing conditions, i.e., conditions under which un-cross-linked heavy and light chains would be completely dissociated and separated on the gel (Figure 2). The gels showed six different bands which corresponded to the expected species of the cross-linked and un-crosslinked antibody chains, i.e. (from the top to the bottom of the gel), (a) the intact antibody which comprised the two heavy and two light chains (designated H2L2 in the figure), (b) two heavy chains coupled to one light chain (H,L), (c) two cross-linked heavy chains (H2),(d) one light chain coupled to a heavy chain (HL), (e) an uncoupled heavy chain (H), and (f) an uncoupled light chain (L). Intermolecular cross-linking (i.e., species larger than the
1
.1
1
I
10 100 Reagent per interchaindisulfide bond (molar ratio) 1
I lo00
Figure 3. Extent of interchain antibody cross-linking as a function of reagent concentration. Samples of anti-biotin antibody were reduced and cross-linked with various concentrations of either BMH (m),MPHPD (O),or PDM ( 0 )and treated further as described in the legend to Figure 2. The percentage of cross-linking was measured by densitometry tracings of the SDS-PAGE gel as described in the text.
intact cross-linked molecule) was not observed. A t intermediate concentrations of the cross-linkers, two additional distinct bands were also apparent at Mfi slightly less than those of the H and H2 bands. These new bands may be accounted for by interchain cross-linking between two light chains (i.e., the designated bands could represent La and HL2, respectively); alternatively, these bands may be attributed to “incorrect” cross-linking of two H chains, (i.e., the interchain or intrachain cross-linking between cysteine and lysine groups). The different cross-linked species could not be facilely separated by conventional means (such as gel filtration), since, in the absence of denaturing agents, all species migrated as a singleantibody entity, irrespectiveof internal cross-linking. Therefore, the final yields of cross-linking (defined as the molar percent of the total cross-linked species) for the various reagents were determined densitometrically from the bands on the SDS-PAGE gels. Interestingly, the results (Figure 3) showed that, for the bismaleimide derivatives, a 2-fold molar excess of the respective cross-linking reagent was sufficient to provide a majority of the final cross-linking yield (i.e., between 35 and 55%);in order to achieve significantly higher yields, a great (several-hundred-fold) excess of reagent was required. The probable reason for the high levels of reagent necessary for extensive cross-linking is that the antibody molecule was subjected to reduction but not denaturation; therefore the molecule would be expected to retain its
278
Goldberg et al.
Bioconjugte Chem., Vol. 2, No. 4, 1991 0.8
BMH 0.6 - HL
0.4
-a-H 0.2
Figure 4. SDS-PAGE of antibodies cross-linked with dibromoacetyl reagents. Antibody samples were reduced with dithiothreitol and cross-linked by using a 40-fold molar ratio (reagent to interchain disulfide bonds) of either dibromoacetone (lane 1) or dibromoacetylethylenediamine (lane 2). The gel was electrophoresed under reducing conditions. See legend to Figure 2 for additional details and key to symbols. Table I. Modification in Amino Acid Composition of Bismaleimide-Cross-Linked Antibodies antibody preparation Cys/Ala Leu/Ala Lys/Ala native 0.286 1.707 1.612 dithiothreitol-reduced 0.285 1.800 1.623 after ultrafiltration 0.273 1.726 1.522 BMH-cross-linked 0.229 1.884 1.520 MPHPD-cross-linked 0.217 1.865 1.402 PDM-cross-linked 0.214 1.809 1.538
native structure, wherein certain cysteines are likely unexposed. Only in the presence of high concentrations of reagent would some of the bismaleimidemoleculesreach and react with the newly formed thiols. Sincethe disulfidederived SH groups are in spatial juxtaposition, they would be expected to react simultaneously with the bismaleimide reagent. This would also explain why PDM, which is the shortest bismaleimide reagent, gave the highest yield (over 90 5% cross-linking). Unfortunately (perhaps due to its hydrophobicity), about 50% of the antibody precipitated. Due to the fact that complete cross-linking could not be achieved with bismaleimide reagents, we decided to try another family of reagents which contain dibromoacetyl groups (see Figure 1). In this case, we took into account the possibility that the bromoacetyl moiety can also react with other nucleophiles on the protein and that the crosslinking could therefore occur at a later stage and at other positions (not only with cysteines). With these reagents, however, a similar specificity of cross-linking was obtained, but at lower overall yields (Figure 4). Attempts to crosslink the reduced cysteines by using dibromopropanol or epichlorohydrin were unsuccessful. An interesting observation was the appearance of a series of new bands, which resulted from the reaction of the longer (but not the shorter) dibromoacetylreagent. This may reflect the type of "incorrect" cross-linking mentioned above. In view of these results, we decided to continue our studies with the bismaleimide reagents, since the goal of this work was to achieve stabilization of the antibody in the highest possible yield. Properties of t h e Cross-Linked Antibody. To examine the specificity of the cross-linking reaction, amino acid analysis was performed which showed a reduction of about 25 % of the content of cysteines in the cross-linked antibody (Table I). Since there are five interchain disulfides out of a total of 17, this result suggests that the major sites of bismaleimide-induced cross-linking indeed occur between the interchain disulfides. The data also indicate that other types of amino acids (lysines in particular) are not extensivelyinvolved in the cross-linking. To examinethe biologicalactivity of the respectivecross-
I I
0.0
I I
MPHPD 0.6
0.4
0.2
0.0
PDM
I
0.6
0.4
0.2
0.c
Figure 5. Binding activity of bismaleimide-cross-linked antibodies. Native or cross-linkedpreparations (usingthe designated reagent) of antibiotin antibody were examined by ELISA. The sampleswere incubated with a biotinylated target protein coupled to wells of a microtiter plate, washed, and treated further with protein A conjugated peroxidase, prior to incubation with substrate. The curves represent native antibody (w) and crosslinked antibody preparations using 5- (O),120- ( O ) ,and 600-fold ( 0 )molar concentrations of the given reagent over interchain disulfide groups, respectively.
linked antibody preparation, we employedan ELISA assay. For these studies, a monoclonal anti-biotin antibody was employed and its binding to a biotinylated target protein was assessed (Figure 5). It is emphasized here that since the various species were not separated, the antibody preparation consists of a mixture of un-cross-linked and partially and fully cross-linkedmolecules, the latter bearing between one and four bismaleimide adducts. Consequently, the binding measurements represent mean values which correspond to the nonhomogeneous cross-linked antibody preparations. Nevertheless, in all cases, high levels of binding activity were observed. Increasing the concentration of bismaleimide did not drastically affect the binding capacity of the resultant antibody. In the case of MPHPD, the pattern of the binding curves were quite similar to that of the native antibody, indicating once again that the cross-linked interchain disulfides are not essential for activity. In the case of PDM (and perhaps the highest concentration of BMH), the observed decrease in binding activity may be related to the insolubility of the cross-linked product. In order to determine whether such cross-linked anti-
Interchain Cr&lnklng
of Antibodies
Table 11. Release of Biotinyl-Ferritin from Native or Crorr-Linked Anti-Biotin Antibody-seDharose Columna anti-biotinantibody column eluent native cross-linked 8 46 distilled water 37 45 PBS (pH 7.1) sodium phosphate buffer (0.1 M, 23 22 DH 7.0, + 1 M NaCl)