Bifunctional Fusion Proteins of Calmodulin and Protein A as Affinity

The proteinaceous bifunctional affinity ligand can be easily removed by addition of EGTA, and fresh ProtA−CaM can be easily reloaded onto the column...
0 downloads 0 Views 185KB Size
Download PDF
Anal. Chem. 1996, 68, 3939-3944

Bifunctional Fusion Proteins of Calmodulin and Protein A as Affinity Ligands in Protein Purification and in the Study of Protein-Protein Interactions Nathaniel G. Hentz and Sylvia Daunert*

Department of Chemistry, University of Kentucky, Lexington, Kentucky 40506-0055

An affinity chromatography system is described that incorporates a genetically designed bifunctional affinity ligand. The utility of the system in protein purification and in the study of protein-protein interactions is demonstrated by using the interaction between protein A and the heat shock protein DnaK as a model system. The bifunctional affinity ligand was developed by genetically fusing calmodulin (CaM) to protein A (ProtA). The dual functionality of protein A-calmodulin (ProtA-CaM) stems from the molecular recognition properties of the two components of the fusion protein. In particular, CaM serves as the anchoring component by virtue of its binding properties toward phenothiazine. Thus, the ProtA-CaM can be immobilized on a solid support containing phenothiazine from the C-terminal domain of the fusion protein. Protein A is at the N-terminal domain of the fusion protein and serves as the affinity site for DnaK. While DnaK binds specifically to the protein A domain of the bifunctional ligand, it is released upon addition of ATP and under very mild conditions (pH 7.0). In addition to obtaining highly purified DnaK, this system is very rugged in terms of its performance. The proteinaceous bifunctional affinity ligand can be easily removed by addition of EGTA, and fresh ProtA-CaM can be easily reloaded onto the column. This allows for a facile regeneration of the affinity column because the phenothiazine-silica support matrix is stable for long periods of time under a variety of conditions. This study also demonstrates that calmodulin fusions can provide a new approach to study proteinprotein interactions. Indeed, the ProtA-CaM fusion protein identified DnaK as a cellular component that interacts with protein A from among the thousands of proteins present in Escherichia coli. Affinity chromatography is a purification technique that takes advantage of highly selective binding interactions found in biological processes, such as replication, repair, transport, immune response, communication, etc.1,2 In affinity chromatography, one of the two components of a binding pair (the ligand) is immobilized to a support matrix and is used to purify its affinity counterpart (the analyte). When the affinity ligand is proteinaceous, there may be a reduction or loss of its binding ability/ specificity if harsh conditions are used for its immobilization to the support matrix. In addition, the immobilized ligand may be susceptible to a loss of affinity upon repetitive use or storage for (1) Walters, R. R. Anal. Chem. 1985, 57, 1099A-1114A. (2) Sii, D.; Sadana, A. J. Biotechnol. 1991, 19, 83-98. S0003-2700(96)00512-4 CCC: $12.00

© 1996 American Chemical Society

long periods of time. Finally, proteolysis and heavy metal contamination may deactivate the ligand. To alleviate the problem of affinity matrix inactivation, bifunctional ligands have been explored where the bifunctional ligand acts as a noncovalent bridge between the analyte and the support matrix.3 Several systems that employ bifunctional affinity ligands have been reported.3-7 In these examples, bifunctional affinity ligands were synthesized by covalently attaching two different binding domains together. In a different approach, bifunctional ligands can be prepared by genetically engineering fusion proteins. By fusing an oligohistidine tail to a polypeptide fragment of the upstream stimulatory factor (USF), a bifunctional ligand was created for the purification of full-length USF on the basis of the ability of full-length USF to form a dimer with the USF polypeptide.8 In this case, the bifunctional ligand was reversibly immobilized from the oligohistidine domain onto a column of chelating resin charged with Ni2+, while the USF domain of the ligand was used to study protein-protein interactions or to purify the full-length USF. Similarly, other groups have used gene fusion techniques to produce recombinant bifunctional fusion proteins to study other protein-protein interactions.9-11 For example, by genetically fusing a polyhistidine tail to the upstream regulatory element binding protein (UREB1), Gu et al. were able to purify antibodies raised against UREB1 by using the polyhistidine domain to attach UREB1 to a column of chelating resin charged with Ni2+.12 In this manner, the polyhistidine-UREB1 fusion protein acted as a bifunctional ligand. Another bifunctional ligand has been described where the biotin carboxyl carrier protein subunit of Escherichia coli acetyl-CoA carboxylase was fused to the Fab fragment of an IgG1 antibody to human tumor necrosis factor R (TNF).13 In this method, the biotin domain was immobilized to streptavidin-agarose beads and the Fab domain was used to purify TNF. (3) Mattiasson, B.; Linne´, E.; Kaul, R. In Molecular Interactions in Bioseparations; Ngo, T. T., Ed.; Plenum Press: New York, 1993; pp 395-401. (4) Labrou, N.; Clonis, Y. D. J. Biotechnol. 1994, 36, 95-119. (5) Mattiason, B.; Olsson, U. J. Chromatogr. 1986, 370, 21-28. (6) Kaul, R.; Olsson, U.; Mattiasson, B. J. Chromatogr. 1988, 438, 339-346. (7) Olsson, U.; Mattiasson, B. J. Chromatogr. 1986, 370, 29-37. (8) Lu, T.; Van Dyk, M.; Sawadogo, M. Anal. Biochem. 1993, 213, 318-322. (9) Chatton, B.; Bahr, A.; Acker, J.; Kedinger, C. BioTechniques 1995, 18, 142145. (10) Coleman, R. A.; Taggart, A. K. P.; Benjamin, L. R.; Pugh, B. F. J. Biol. Chem. 1995, 270, 13842-13849. (11) Lizano, S.; Johnston, K. H. J. Microbiol. Methods 1995, 23, 261-280. (12) Gu, J.; Stephenson, C. G.; Iadarola, M. J. BioTechniques 1994, 17, 257262. (13) Weiss, E.; Chatellier, J.; Orfanoudakis, G. Protein Expression Purif. 1994, 5, 509-517.

Analytical Chemistry, Vol. 68, No. 22, November 15, 1996 3939

Figure 1. (A) ProtA-CaM fusion protein as a bifunctional ligand: The CaM domain binds to the immobilized phenothiazine in a Ca2+dependent manner, and DnaK binds to the ProtA domain. (B) DnaK is eluted upon addition of ATP. (C) The bifunctional ligand, ProtACaM, can be released by removing Ca2+ with EGTA.

In the system described in this article, calmodulin (CaM) was genetically fused to protein A (ProtA) and the resulting ProtACaM fusion protein was used as a bifunctional affinity ligand for the efficient purification of the biologically important heat shock protein, DnaK (Figure 1). Calmodulin is a protein that exhibits Ca2+-dependent binding toward a number of hydrophobic compounds,14-16 including the antipsychotic drug phenothiazine.17 In the presence of Ca2+, calmodulin assumes a conformation that allows binding to phenothiazine. By using a complexing agent such as EDTA or EGTA, Ca2+ can be stripped from CaM, inducing a conformational change that causes the complete dissociation of the phenothiazine-calmodulin complex.17 The protein A domain of the bifunctional ligand acts as a recognition site for DnaK,18,19 and as demonstrated herein, protein A can selectively interact with DnaK from among the thousands of proteins present in E. coli lysates. In the presence of ATP, DnaK changes conformation,20 which allows the dissociation of the protein A-DnaK complex, yielding pure DnaK. The protein DnaK belongs to a class of 70-kDa heat shock proteins, hsp70, and plays a variety of roles in cells19-23 by acting (14) Tanaka, T.; Ohmura, T.; Hidaka, H. Mol. Pharmacol. 1982, 22, 403-407. (15) Maulet, Y.; Cox, J. A. Biochemistry 1983, 22, 5680-5686. (16) Levin, R. M.; Weiss, B. Mol. Pharmacol. 1977, 13, 690-697. (17) Foster, W. S; Jarrett, H. W. J. Chromatogr. 1987, 403, 99-107. (18) Hellebust, H.; Uhle´n, M.; Enfors, S.-O. J. Bacteriol. 1990, 172, 5030-5034. (19) Sherman, M. Y.; Goldberg, A. L. J. Bacteriol. 1991, 173, 7249-7256. (20) Shi, L.; Katoaka, M.; Fink, A. L. Biochemistry 1996, 35, 3297-3308. (21) Burkholder, W. F.; Panagiotidis, C. A.; Silverstein, S. J.; Cegielska, A.; Gottesman, M. E.; Gaitanaris, G. A. J. Mol. Biol. 1994, 242, 364-377. (22) LaRossa, R. A.; Van Dyk, T. K. Mol. Microbiol. 1991, 5, 529-534. (23) Gross, C. A.; Strauss, D. B.; Erickson, J. W.; Yura, T. In Stress Proteins in Biology and Medicine; Morimoto, R., Tissie`res, A., Georgopoulos, C., Eds.; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, NY, 1990; pp 166-190.

3940 Analytical Chemistry, Vol. 68, No. 22, November 15, 1996

as a molecular chaperone.24,25 The chaperone effect between the DnaK and polypeptides is thought to be triggered by a conformational change induced by the binding of ATP.20 Therefore, binding of ATP plays an important role in the function of DnaK as a molecular chaperone. DnaK has been purified by many methods that include a number of chromatographic steps.20,26,27 All of these purification schemes take advantage of the ATP-dependent binding of DnaK by immobilizing ATP on agarose.20,27 A disadvantage of this method is that not only are there a variety of other ATP-dependent binding proteins that may interfere with the DnaK purification, but in addition, the ATP-agarose affinity matrix may be susceptible to breakdown by ATPases. In another purification method for DnaK, Hellebust et al. reported the use of protein A as the affinity ligand.18 In this technique, DnaK was purified by immobilizing a DnaK-protein A complex (formed in vivo) through an interaction between the protein A domain and an IgG-Sepharose column. The DnaK was then eluted upon addition of a buffered ATP solution. A potential disadvantage of this method stems from the proteinaceous nature of the affinity matrix. Proteinaceous affinity columns typically have a limited lifetime, and therefore, it should be advantageous if a purification scheme can be devised by using nonproteinaceous affinity matrices. The system described in this article has the advantage that a stable phenothiazine affinity matrix is used and that the bifunctional affinity ligand (i.e., ProtA-CaM) can be easily added or removed, allowing for facile regeneration of the column. This affinity chromatography column has been stable for extended periods of time under continuous use in the laboratory. In addition, the described system can be considered as a model system for protein affinity chromatography that employs genetically engineered bifunctional ligands. Indeed, by using gene fusion techniques, this system can be adapted to attach proteins other than protein A to calmodulin, to yield bifunctional ligands that can be used in bioseparations and to study protein-protein interactions. EXPERIMENTAL SECTION Reagents. Ethylenebis(oxyethylenenitrilo)tetraacetic acid (EGTA) and 2-(trifluoromethyl)phenothiazine were obtained from Aldrich (Milwaukee, WI). The bicinchoninic acid protein microassay kit was purchased from Pierce (Rockford, IL). Ethylenediaminetetraacetic acid (EDTA) was purchased from Fisher Scientific (Fair Lawn, NJ). Adenosine 5′-triphosphate (ATP), kanamycin, phenylmethanesulfonyl fluoride, leupeptin, magnesium chloride, and low molecular weight standards were obtained from Sigma (St. Louis, MO). N-(2-Hydroxyethyl)piperazine-N′-2-ethanesulfonic acid (HEPES) and tris(hydroxymethyl)aminomethane (Tris) were purchased from Research Organics (Cleveland, OH). DnaK was obtained from Epicentre Technologies (Madison, WI) and Tween 20 was obtained from Promega (Madison, WI). LuriaBertani (LB) broth was obtained from Difco Laboratories (Detroit, MI). Deionized (Milli-Q water purification system, Millipore, (24) Hendrick, J. P.; Hartl, F. U. Annu. Rev. Biochem. 1993, 62, 349-384. (25) Morimoto, R.; Tissie`res, A.; Georgopoulos, C. In Biology of the Heat Shock Proteins and Molecular Chaperones; Morimoto, R., Tissie`res, A., Georgopoulos, C., Eds.; Cold Spring Harbor Laboratory Press: Plainview, NY, 1994; pp 1-30. (26) Zylicz, M.; Georgopoulos, C. J. Biol. Chem. 1984, 259, 8820-8825. (27) Zylicz, M.; Ang, D.; Georgopoulos, C. J. Biol. Chem. 1987, 262, 1743717442.

Bedford, MA) distilled water was used to prepare all solutions and mobile phases. Preparation of Affinity Chromatography Support. 2-(Trifluoromethyl)-10H-(3′-aminopropyl)phenothiazine hydrochloride (TAPP) prepared from commercially available 2-(trifluoromethyl)phenothiazine,28 was coupled to Protein-Pak epoxy-activated silica (40-µm diameter, 300-Å pore size) (Millipore) as described earlier.29 The final ligand density was determined to be 12 ( 3 µg of TAPP/mg of silica (three different preparations). The TAPPsilica affinity medium was stored in a solution of 10 mM TrisHCl, pH 7.5, containing 1 mM sodium azide and kept in the dark at 4 °C. Apparatus. The silica-TAPP affinity medium was packed into an HR 5/10 glass column (5 mm × 100 mm) (Pharmacia Biotech, Uppsala, Sweden). The buffers were pumped with a BioCAD Sprint liquid chromatography system (PerSeptive Biosystems, Framingham, MA) at a flow rate of 1.50 mL/min. The BioCAD system was equipped with a 100- or 500-µL sample injection loop, and detection was carried out by UV absorption at 280 nm. The area under the chromatographic peaks was reported in units of microvolts per second as provided from the software of the BioCAD system. For the expression of the protein A-calmodulin fusion proteins, an IEC Centra-7R refrigerated bench-top centrifuge (International Equipment Co., Needham Heights, MA) and a Micro Speed Fuge SFR11K (Savant Instruments, Farmingdale, NY) were employed. A sonic dismembrator Model 550 (Fisher Scientific) was used to lyse all bacterial cells before isolating the ProtA-CaM fusion proteins. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) of the purified DnaK fractions was carried out with a PhastSystem electrophoresis setup (Pharmacia Biotech, Uppsala, Sweden). Preparation of Protein A-Calmodulin Bifunctional Affinity Ligand. Transformed E. coli containing the plasmids pALP1CaM or pDS100 (plasmids kindly provided by Stirling et al.)30 was incubated at 37 °C overnight in 40 mL of LB medium containing kanamycin. After centrifugation at 4000 rpm for 15 min, the pellets were washed in TS buffer (50 mM Tris-HCl, pH 7.6, containing 150 mM NaCl), centrifuged again, and resuspended in 3 mL of TS buffer. A volume of 30 µL of a solution of protease inhibitors (20 mM phenylmethanesulfonyl fluoride, 10 µg/mL leupeptin, and 20 mM EDTA) was added to the 3-mL cell mixture. The cells were then lysed with the sonic dismembrator for 5 min. Cellular debris was removed by centrifugation at 11 000 rpm for 10 min, and Tween 20 was added to the supernatant to give a final concentration of 0.05% (v/v) Tween 20. Purification of DnaK Protein. After the bacterial cells were lysed and centrifuged, 100 or 500 µL of the crude lysate was injected onto the TAPP affinity column that had been preequilibrated with a wash buffer containing 10 mM HEPES-HCl, 0.50 mM CaCl2, and 1.0 M NaCl, pH 7.0. A DnaK wash buffer containing 10 mM HEPES-HCl, 0.50 mM CaCl2, 1.0 µM ATP, 1.0 M NaCl, and 1.0 µM MgCl2, pH 7.0, was used to elute DnaK. An EGTA-containing buffer, 10 mM Tris-HCl, and 5.0 mM EGTA, pH 8.0, was used to remove the ProtA-CaM from the column, when (28) Hart, R. C.; Bates, M. D.; Cormier, M. J.; Rosen, G. M.; Conn, P. M. Methods Enzymol. 1983, 102, 195-199. (29) Charbonneau, H.; Hice, R; Hart, R. C.; Cormier, M. J. Methods Enzymol. 1983, 102, 17-39. (30) Stirling, D. A.; Petrie, A.; Pulford, D. J.; Paterson, D. T. W.; Stark, M. J. R. Mol. Microbiol. 1992, 6, 703-713.

the latter needed to be regenerated. Elution of proteins was monitored by UV absorbance at 280 nm. It should be noted that the only sample pretreatment employed was preinjection filtration carried out by a syringe filter (0.45-µm pore-size nylon filter; Alltech, Deerfield, IL). The purity of the expressed DnaK protein was verified by SDS-PAGE on 12.5% polyacrylamide PhastGels (Pharmacia Biotech), which were developed by silver staining (Development Method 210, Pharmacia Biotech). Amino acid sequencing of DnaK was performed at the Macromolecular Center of the University of Kentucky. RESULTS AND DISCUSSION Chromatographic separations based on bifunctional affinity ligands can be operated in two different modes. The first mode allows for the bifunctional ligand to become immobilized to the solid support, before the introduction of the sample. In the second mode, the bifunctional ligand is allowed to associate with the analyte in solution, and the complex is then applied to the solid support. The bifunctional ligand affinity chromatography system described in this article can be operated in both modes and takes advantage of the ATP-dependent binding of DnaK toward protein A. In the system described herein, two bifunctional affinity ligands were expressed in E. coli from plasmids pALP1-CaM and pDS100. In these ligands, calmodulin was genetically fused to protein A. The two ligands differ in that the ProtA-CaM expressed from pALP1-CaM lacks the 12 N-terminal amino acids normally encoded by the native calmodulin gene, CMD1. In contrast, the fusion protein expressed from pDS100 contained all of the amino acids of calmodulin plus a spacer of 10 amino acids located between calmodulin and protein A. Both bifunctional affinity ligands have two different domains that allow for two binding interactions, namely, the phenothiazine-calmodulin interaction and the protein A-DnaK interaction (Figure 1). As the ProtA-CaM is expressed in the cells, the DnaK binds to the protein A domain of the ProtA-CaM fusion protein in vivo, leaving the calmodulin domain free. The cells are then lysed, and the whole lysate is applied to the phenothiazine column after preinjection filtration. It has been shown previously that, in the presence of Ca2+, calmodulin binds to phenothiazine with a Kd ≈ 10-6 M.16 Because of this interaction, the complex of ProtACaM with DnaK can be attached to the phenothiazine support through the calmodulin domain in the presence of Ca2+. Unretained cellular components can be washed away under mild conditions and without disturbing the binding of the immobilized complex of ProtA-CaM with DnaK. This concept is illustrated in the chromatogram shown in Figure 2, where a 100-µL sample of crude cellular lysate containing the ProtA-CaM fusion protein expressed from pDS100 was injected. Peak A in the chromatogram corresponds to unretained cellular components that were washed off the column with the wash buffer (contains Ca2+ and 1.0 M NaCl), while the complex of ProtA-CaM with DnaK was retained. The immobilized DnaK was then eluted (peak B, Figure 2) with the ATP-containing buffer followed by the subsequent release of the ProtA-CaM fusion protein (peak C, Figure 2) by elution with the EGTA buffer. The sequential release of the DnaK and ProtA-CaM fusion protein from the column indicates that the interaction between ProtACaM and the phenothiazine on the column, and ProtA-CaM with DnaK, is not disturbed under the separation conditions used in this experiment. Analytical Chemistry, Vol. 68, No. 22, November 15, 1996

3941

Figure 2. Chromatogram of a 100-µL sample of cellular lysate of E. coli containing plasmid pDS100. Initial elution with the wash buffer removed unretained components (peak A). The buffer was changed to the ATP-containing buffer (indicated by the solid arrow) allowing the elution of DnaK (peak B). The ProtA-CaM fusion protein (peak C) was eluted upon addition of an EGTA-containing buffer (indicated by the broken arrow).

Figure 3. Chromatogram of a series of 100-µL injections of E. coli lysate. The first injection (A) represents loading of the fusion protein from E. coli cellular lysate that expressed ProtA-CaM, while DnaK (labeled as such) was eluted leaving the ProtA-CaM/phenothiazine interaction intact. Subsequent injections (B-D) represent lysate from E. coli that did not express the ProtA-CaM fusion protein. In each of these cases, DnaK (denoted as DnaK*) was selectively released upon addition of the ATP-containing buffer (indicated by the arrow).

To demonstrate that this affinity chromatography system can be used repeatedly, an injection of E. coli lysate containing the ProtA-CaM fusion protein was made (Figure 3) to immobilize the bifunctional fusion protein onto the phenothiazine column. As mentioned above, the expressed ProtA-CaM complexes DnaK in vivo through the protein A domain, and as a result of this injection, DnaK becomes immobilized on the column through the bridging ProtA-CaM bifunctional ligand. The DnaK was then released from the column by elution with the ATP-containing buffer, leaving the fusion protein immobilized. Subsequent injections of E. coli lysate from a different bacterial population that did not contain the pDS100 plasmid needed to express the fusion protein demonstrate the ability of this system to purify 3942 Analytical Chemistry, Vol. 68, No. 22, November 15, 1996

endogenous DnaK. Figure 3 shows the first three injectionelution cycles, after immobilizing the fusion protein affinity ligand onto the column. In each of these cases, the DnaK was eluted with the ATP-containing buffer (indicated as DnaK*, Figure 3) after the unwanted cellular components were eluted (peaks B-D, Figure 3). The peak areas that correspond to DnaK* are almost identical for several injection-elution cycles, which demonstrates the reproducibility of the described system. The larger DnaK peak (denoted DnaK in Figure 3), from the injection that immobilized the fusion protein-DnaK complex onto the column, may be explained by the fact that the DnaK purified from E. coli that did not express the fusion protein (denoted DnaK* in Figure 3) comes from a different E. coli variant. Figure 3 also illustrates that while the addition of ATP alters the conformation of DnaK (allowing dissociation from the protein A domain), the binding of the bifunctional affinity ligand, ProtA-CaM, to the phenothiazine is not altered, and thus ProtA-CaM remains immobilized on the column. To examine the effect of column loading, the affinity column was subjected to several 100-µL injections of E. coli lysate containing the ProtA-CaM fusion protein. After each injectionelution series, only the DnaK was eluted; no EGTA-containing buffer was added, and thus, the fusion protein remained immobilized to the column. Each time, more ProtA-CaM became bound to the column (in addition to the amount from the previous injections) giving more DnaK-binding sites. This demonstrates that the phenothiazine sites become saturated with increasing amounts of ProtA-CaM, which further indicates the selective nature of the interaction between ProtA-CaM and immobilized phenothiazine. During these studies, it was also found that a single 500-µL injection of E. coli lysate containing the fusion protein was sufficient to occupy all of the phenothiazine sites with ProtA-CaM. After numerous injections of the cellular lysate, the peak areas started to vary, which was attributed to biofouling. To remedy this problem, the ProtA-CaM was removed from the phenothiazine column by eluting with an EGTA-containing buffer (to remove the Ca2+). This changes the conformation of CaM in a way that releases it from the phenothiazine column. After the bifunctional ligand was removed, the phenothiazine column was further cleaned by using high salt (1.0 M NaCl) and/or methanol washes. The amount of ProtA-CaM that occupied all of the TAPP sites was determined to be 175 µg, which corresponds to a bifunctional ligand density of approximately 300 µg/g of resin. It should be noted that it is possible to further increase the TAPP ligand density. However, previous studies have determined that the biospecificity of the phenothiazine-calmodulin interaction is compromised at densities of TAPP higher than 16 µg/mg of resin, which is higher than the TAPP density of our support.31 The ProtA-CaM was then reloaded to regenerate the column by adding fresh E. coli cell lysate containing the ProtA-CaM fusion protein. By following this procedure, the column was restored to its original performance. Under these conditions, the column was stable and gave reproducible results (relative standard deviation of less than 2%) for 4 months (Figure 4). It is worth noting that even after 6 months the column retained about 70% of its binding capacity. To further demonstrate the high reproducibility of the affinity chromatography system, conditions different from those used in the data shown in Figure 3 were employed. For that, the column (31) Jarrett, H. W. J. Chromatogr. 1986, 363, 456-461.

Figure 4. Lifetime of the silica-TAPP affinity column determined as a function of both peak area of the ProtA-CaM eluted and reproducibility over a period of several months.

Figure 5. Chromatogram of a series of injections of E. coli lysate from a strain containing the pDS100 plasmid that expressed the ProtA-CaM. The first injection (injection volume 500 µL) loads the phenothiazine column to capacity with the bifunctional ProtA-CaM. After injection A, DnaK was released upon addition of the ATPcontaining buffer (indicated by the arrow). Subsequent injections (BD) of 100 µL of lysate from the same E. coli that expressed the bifunctional fusion protein allowed the interaction of only the free DnaK (denoted by DnaK*).

was first stripped of the ProtA-CaM bifunctional ligand, washed with 1.0 M NaCl and methanol, and then reloaded by using 500 µL of cellular lysate to occupy all of the phenothiazine sites on the column with freshly expressed ProtA-CaM bifunctional ligand. DnaK was then purified from lysed E. coli containing plasmid pDS100, in contrast to the data shown in Figure 3 that involve cellular lysate of E. coli that did not contain the plasmid that encodes for the ProtA-CaM bifunctional ligand (corresponding to peaks denoted as DnaK* in Figure 3). In the chromatogram shown in Figure 5, with the first injection all of the TAPP binding sites were bound to capacity with the fusion protein-DnaK complex formed in vivo. This accounts for the larger DnaK peak (denoted DnaK in Figure 5) as compared to the other DnaK peaks (denoted DnaK* in Figure 5). The reason for the smaller DnaK* peaks is that part of the DnaK contents of the cell is complexed to the ProtA-CaM expressed by the cells. However, since the column has already been saturated with ProtA-CaM from the first injection, the additional DnaK/ProtA-CaM complex is

Figure 6. SDS-PAGE gel (silver stained) demonstrating the purity of the DnaK obtained by the bifunctional ligand affinity system, where lane 1 shows the commercially available DnaK (sold as >95% pure at 4 µg/µL) at ∼69 kDa. Lanes 2 and 3 show the DnaK released from the bifunctional ligand ProtA-CaM (expressed from the plasmids pALP1-CaM and pDS100, respectively). Lane 4 contains the molecular weight markers.

unretained by the column, and only free DnaK in the cellular lysate becomes immobilized. The fractions from the affinity column that correspond to the peaks for DnaK and DnaK* were collected and subjected to a bicinchoninic acid protein microassay. It was found that the amount of DnaK and DnaK* in each of the eluted peaks corresponds to 15.7 and 3.6 µg of DnaK/500 µL of lysate, respectively. As explained above, peaks A-D correspond to unretained cellular components, while peaks containing DnaK eluted with a buffer containing 1 µM ATP are labeled as such in Figure 5. The injection-elution sequences shown in Figures 3 and 5 demonstrate the reproducibility of the system described in this article under different conditions. To better evaluate the reproducibility of this system, the peak areas were calculated for DnaK* in Figures 3 and 5; the relative standard deviation of the peak areas for DnaK* in Figure 3 was 1.7%, while that for DnaK* in Figure 5 was 2.3%. Similar behavior was observed with ProtA-CaM expressed from plasmid pALP1-CaM in E. coli. This demonstrates that the deletion of the 12 N-terminal amino acids in the CaM molecule expressed from pALP1-CaM, or the addition of a 10-amino acid spacer between the CaM and the ProtA in the fusion protein expressed from pDS100, had no deleterious effect on the binding of the DnaK to protein A. Likewise, the similarity of the behavior of the two bifunctional ligands also shows that the calmodulin domain is functional in both ligands and can interact with the immobilized phenothiazine on the column. Once the reproducibility of the bifunctional affinity ligand system was determined, the purity of the isolated DnaK was verified. The fractions that correspond to the peaks labeled as DnaK and DnaK* (Figure 5) were collected, pooled, and subjected to SDS-PAGE analysis along with the corresponding fractions obtained when using ProtA-CaM expressed from pALP1-CaM (Figure 6). For molecular size comparisons, commercially available molecular weight markers (lane 4, Figure 6) were used. A commercial DnaK sample is shown in lane 1. Lanes 2 and 3 contain purified DnaK using the bifunctional ProtA-CaM fusion proteins expressed from plasmids pALP1-CaM and pDS100, respectively. The single bands in lanes 2 and 3, which correspond to a molecular mass of ∼69 kDa, demonstrate the high selectivity of the ProtA-CaM bifunctional affinity ligand. Analytical Chemistry, Vol. 68, No. 22, November 15, 1996

3943

To verify that the bands at ∼69 kDa (lanes 2 and 3, Figure 6) were indeed DnaK, the fraction that contained the protein was sequenced to confirm that the first 10 N-terminal amino acids were consistent with those of DnaK. According to literature,32 the 10 N-terminal amino acids in DnaK are Gly-Lys-Ile-Ile-Gly-Ile-AspLeu-Gly-Thr, which are in complete agreement with the results of the sequencing of the DnaK protein isolated by using the bifunctional affinity ligand system described in this article. The SDS-PAGE gel in Figure 6 also demonstrates the utility of calmodulin-based bifunctional ligands in the study of proteinprotein interactions. Indeed, the ProtA-CaM bifunctional ligand identified DnaK as a protein that interacts with protein A. It should be noted that, by using a gene fusion approach to assemble the bifunctional ligand, it is possible to produce this ligand in a controlled fashion with the protein of interest bound to the N-terminus of calmodulin without disturbing the interaction of calmodulin with immobilized phenothiazine. Our studies demonstrate that this system can also be employed in the identification of compounds that disrupt protein-protein interactions, as specifically demonstrated by the use of ATP to release DnaK from the DnaK/ProtA-CaM immobilized complex. It should also be feasible to use this system to identify compounds that enhance a given protein-protein interaction. Consequently, genetically designed CaM-based bifunctional ligands may find applications in drug discovery/screening where the desired drug needs to promote or inhibit the interaction between two proteins in vivo. In conclusion, the affinity chromatography system described in this article demonstrates the feasibility of using a gene fusion approach in the design of proteinaceous bifunctional affinity ligands. In particular, this affinity chromatography system provides an efficient method for the purification of the heat shock protein DnaK by taking advantage of the ATP-dependent binding (32) Schmid, D.; Jaussi, R.; Christen, P. Eur. J. Biochem. 1992, 208, 699-704.

3944

Analytical Chemistry, Vol. 68, No. 22, November 15, 1996

of DnaK toward protein A. The described bifunctional ligand approach can be used for affinity bioseparations of other proteins, as well as for the study of protein-protein interactions. For that, recombinant DNA techniques can be employed to fuse a protein of interest to calmodulin in a manner similar to the one described in this article. The resulting fusion protein can be immobilized to the stable phenothiazine column and used to purify (or identify) other biomolecules that interact with the fused protein. A unique advantage of this approach is that since the calmodulin-phenothiazine complex can be dissociated under controlled conditions, the phenothiazine column need only be packed once. Indeed, this column can be loaded with the bifunctional ligand of choice, and after completion of the needed experiments, this bifunctional ligand can be removed by addition of EGTA. Then, a structurally different bifunctional ligand can be loaded onto the column through the interaction between the CaM domain and the immobilized phenothiazine, which generates a column with totally different affinity properties. ACKNOWLEDGMENT This work was supported in part by the National Aeronautics and Space Administration (Grant NCCW-60) and the National Institutes of Health (Grant GM 47915). The plasmids pALP1-CaM and pDS100 were a gift from D. A. Stirling,30 and E. coli strain K12 × 289 (F-, strs, T6s, T3r, l-) was kindly provided by S. C. Straley.

Received for review May 22, 1996. Accepted September 1, 1996.X AC960512P X

Abstract published in Advance ACS Abstracts, October 1, 1996.