Heat shock proteins such as gp96 (grp94) isolated from tumor or infected cells are able to induce specific cytotoxic T-cell responses and protective immunity. To facilitate rapid and efficient isolation,we generated gp96-specific recombinant single-chain Fv (scFv)antibodies from a semisynthetic phage display library. When immobilized on Sepharose beads, these antibodies allow a high-yield, one-step purification of native gp96 molecules from both mouse and human tumor cell lysates. gp96 molecules eluted from these affinity columns under mild conditions are still capable of generating antigen-specific CTL responses in mice. Thus, scFv-purified gp96 is still associated with peptides; however, in contrast to conventionally purified gp96,scFv-isolated gp96 is free of contaminating material such as mitogenic concanavalin A and proteolytic cathepsins. With the help of these high-yield antibody columns, it is now possible to rapidly isolate immunogenic gp96-peptide complexes from small amounts of tumor material to a purity that allows their use in cancer immunotherapy protocols.
HSPs4provide an immunogenic context to peptides that associate with them inside tumor or infected cells. These peptides are acquired by HSPs in the cytosol (e.g., for HSP70) or in the ER (for gp96). If such HSP-peptide complexes are injected into mice, potent immune responses can be induced. These responses are specific for HSP-associated peptides, as demonstrated by the induction of specific CTL responses against tumor cells, minor H and viral antigens (reviewed in Refs. 1 and 2). Importantly, HSP immunization of mice also leads to protective immunity against tumor challenge. The ER-resident HSP gp96 is most efficient in this regard(3).
What makes HSPs, especially gp96, so effective in vivo is not completely understood. Recent experiments, however, showed that APCs are equipped with a receptor specific for HSPs, resulting in the efficient uptake of HSPs (4, 5) and the presentation of the associated peptides to T cells, as postulated previously(6). On the basis of these features, HSPs can be used for cancer immunotherapy in mice (7). Furthermore, HSPs appear to be ideal candidates for patient individual immunotherapy protocols.
However, this very promising application will require the rapid purification of HSP molecules from a limited amount of sample material at a very high purity. Highly specific mAbs against the HSPs,especially against gp96 molecules, would be the ideal tool. However,the existing antibodies are of limited suitability because they react poorly with native molecules. This behavior might be a consequence of the intracellular localization of gp96. The anti-gp96 antibodies,however, may react with the native gp96 in the ER of the respective B-cell hybridoma. This interaction can be expected to interfere with normal gp96 function and the secretion of antibodies.
To circumvent this problem inherent to conserved ER proteins, we have generated recombinant antibodies specific for gp96 molecules from a semisynthetic phage display library (8). The selection of the phages was deliberately performed with native gp96 molecules; thus,the resulting recombinant antibodies were able to efficiently recognize native gp96 molecules even after immobilization on Sepharose beads. These antibodies will be a precious tool for the rapid purification of intact gp96-peptide complexes from small quantities of tumor material to be used in autologous cancer immunotherapy protocols.
MATERIALS AND METHODS
Mice and Antibodies.
C57BL/6 mice obtained from Charles River WIGA (Sulzfeld,Germany) and BALB.B mice obtained from Harlan Winkelmann (Borchen,Germany) were maintained in the animal facilities at the Institute for Cell Biology (University of Tübingen, Tübingen, Germany). mAbs to gp96 (anti-Grp94, SPA-850, and clone 9G10) and immunoglobulin heavy chain-binding protein (SPA-827 and clone 10C3) were obtained from StressGen Biotechnologies Corp. (Victoria, British Columbia, Canada).
Standard Purification of gp96.
gp96 was purified as described previously (9).
Selection of gp96-specific scFv Ab-expressing Phages.
A human semisynthetic phage display library (8) was used for selection of scFv Abs directed against mgp96. For the construction of the library, synthetic CDR3 loops comprising random sequences of 4–12 amino acids were introduced into 50 variable heavy chain germ-line genes (10). Subsequently, the heavy chain repertoire was combined in the phagemid vector pHEN1 (11)with the unmutated rearranged Vλ3 variable light chain gene IGLV3S1/DPL16 (12, 13). After preparation of scFv-expressing phage particles in Escherichia coli TG1,5 × 1012 (colony-forming units)phages suspended in MPBS were introduced for panning into Nunc Maxisorb immunotubes (Nunc, Roskilde, Denmark) coated with 80 μg mgp96. Unspecifically adsorbed phages were removed by intensive washings. Specific phages were eluted with 100 mmtriethylamine (pH 12), neutralized, expanded, and reused for further selections up to a total number of four rounds. Binding specificity was verified by phage ELISA. Briefly, microtiter plates were coated with gp96 (0.5 μg/well) and blocked with 2% MPBS, and phage particles(1010 colony-forming units) were added. Chicken anti-gp96 polyclonal serum (1:500) served as a positive control, and PBS served as a negative control. The system was developed by adding rabbit anti-M13 phage Ab (Stratagene), PO-conjugated goat antirabbit immunoglobulin and substrate (3,3′,5,5′-tetramethyl-benzidine). Absorbance at 620 nm was measured. Single colonies (n = 77) were generated after the fourth selection round, and Ab diversity was examined by digestion of plasmid DNA with the restriction enzyme BstNI. Three restriction patterns were identified, and their nucleotide sequences were determined (377 DNA sequencer; Applied Biosystems Inc., Foster City, CA). The sequences of the scFv Abs reported in this study have been deposited in the European Molecular Biology Laboratory Nucleotide Sequence Database (accession numbers AJ252274, AJ252275, and AJ2552276).
Production of Soluble gp96-specific scFv Abs.
The gene segments of three gp96-specific scFv Abs (B10C, G12D, and H11B) were obtained via NcoI-NotI digestion,ligated into the plasmid pHOG21 (14), and expressed in E. coli XL1-Blue (Stratagene). The soluble fraction of the periplasmic extract and the culture supernatant were combined,concentrated, and dialyzed against 50 mm Tris-HCl and 1 m NaCl (pH 7.0).
Chromatographic Purification of Soluble scFv Abs.
All scFv Abs contain a poly-His tag that allows their purification by immobilized metal affinity chromatography. Samples were loaded onto a Chelating Sepharose Fast Flow column (Amersham Pharmacia Biotech,Uppsala, Sweden) previously charged with Ni2+ and equilibrated with dialyzing buffer. After extensive washing the bound material was eluted with 250 mm imidazole and dialyzed against PBS. Its purity was tested by PAGE.
Western and Dot Blot Analysis of Soluble scFv Abs.
mgp96 was separated by SDS-PAGE under reducing (100 mmβ-mercaptoethanol + 2% SDS) conditions and transferred to nitrocellulose. For dot blot analysis, 500 ng of mgp96 were immobilized on nitrocellulose membrane under native (PBS), denaturing (2% SDS), or reducing (100 mm β-mercaptoethanol + 2% SDS)conditions. Western and dot blot membranes were blocked with 2% MPBS. gp96 was detected with rat IgG anti-Grp94 (SPA-850; StressGen Biotechnologies Corp.) and PO-conjugated rabbit antirat IgG (DAKO,Glostrup, Denmark) or with our scFv Abs. All scFv Abs carry a c-myc tag, allowing labeling with mouse IgG antihuman c-myc (Genosys, The Woodlands, TX) and PO-conjugated rabbit-anti-mouse Ab (DAKO). Detection of soluble scFv Abs by ELISA was performed as described for phages. The binding of scFv Abs was tested at various dilutions (1:1 to 1:8192).
Cells (107) were metabolically labeled in the presence of [35S]methionine (150 μCi) in 10 ml of methionine-free RPMI 1640 for 16 h. After lysis of the cells in PBS containing 0.5% NP40 or in hypotonic buffer [30 mm NaHCO3 (pH 7.1); both lysis buffers supplemented with iodoacetamide and phenylmethylsulfonyl fluoride], the lysates were preadsorbed overnight with protein G-Sepharose (Amersham Pharmacia Biotech). Specific mAbs or scFv anti-gp96 Abs and anti-c-myc Ab were added to the cleared lysates at 10μg/ml for 90 min before the complexes were isolated with protein G-Sepharose. In some cases, the lysates were precleared with BSA coupled to Sepharose beads, and the precipitations were performed directly with BSA-Sepharose as a control or with scFv anti-gp96 coupled to Sepharose beads. The immunocomplexes were separated by SDS-PAGE. Fixed gels were dried and exposed to a phosphorimager screen.
Purification of gp96 Using scFv Abs Coupled to Sepharose Beads.
Five mg of scFv anti-gp96 or BSA were coupled to 0.5 mg of CNBr-activated Sepharose (Pharmacia). IGELa2 cell pellets (1 ml) were homogenized as described previously (9). The homogenates were ultracentrifuged for 60 min at 100,000 × g, and the supernatants were applied to the control BSA-Sepharose column followed by the scFv anti-gp96 column. After extensive washing with PBS containing 0.5% NaCl, gp96 was eluted with 100 mm sodium acetate (pH 4.5) containing 0.15 m NaCl, 100 mmdiethanolamine (pH 10.5) containing 0.15 m NaCl,or PBS (pH 7.4) containing 1.3 m NaCl. The fractions were tested in SDS-PAGE and Western blot using mAb SPA-850(StressGen Biotechnologies Corp.). The absorbance at 280 nm was measured to determine the approximate concentration of gp96 by using an extinction coefficient of 1.0.
Immunization of Mice and Generation of CTLs.
The 8–10-week-old C57BL/6 mice were immunized with 30 μg of gp96 purified from IGELa2 cells using the standard method or the scFv anti-gp96 method in 300 μl of PBS or with 80 μl of Sepharose beads(beads volume) coupled to BSA or to scFv complexed to gp96 in 300 μl of PBS. CTLs were generated as described previously (9).
Cell Culture and CTL Assays.
The mouse IGELa2 and the human C1R cell lines were obtained from American Type Culture Collection (Manassas, VA) and cultured in RPMI 1640 containing 10% FCS supplemented with l-glutamine (0.3 mg/ml), penicillin/streptomycin (100 units/ml), and 2-ME (2μl/ml). CTL lines were generated, and CTL assays were performed as described previously (9) using Con A blasts of spleen cells as target cells.
RESULTS AND DISCUSSION
Generation of Recombinant scFv anti-gp96 Abs.
A semisynthetic phage display library generated from 50 variable human heavy chain germ-line genes with randomly varied CDR3 loops and one human Vλ3 variable light chain gene was used(8). Specific phages were obtained by four consecutive selection rounds performed in tubes coated with mgp96. A continuous increase of binding to mgp96 as tested by phage ELISA was noted after each round (binding activity after the fourth round: selected phages = 1.406 and negative control = 0.065 A620 nm). Among 77 single phage colonies, three Ab clones with different CDR3 regions (B10C, G12D, and H11B) were identified by sequence analyses (Fig. 1). All belonged to the VH1 gene family. As expected, the light chains of the Abs were identical. After recloning into the pHOG21 plasmid (14) and expression in E. coli, the Abs showed a good binding activity in ELISA (clone B10C = 1.297, G12D = 0.651,H11B = 0.557, positive control = 0.773,and negative control = 0.003 A620 nm). Dot blot analysis of purified Abs on nitrocellulose membranes coated with reduced,denatured, or native mgp96 revealed binding only to the native protein(Fig. 2).
scFv Abs Precipitate Native gp96 Molecules.
Our recombinant scFv gp96-specific Abs recognize their antigen in a native conformation if tested in an ELISA. The commercially available mAb against gp96 (SPA850) does this poorly, limiting its use for affinity purification of gp96. To examine the ability of the recombinant scFv Abs to interact with native gp96 molecules, we performed immunoprecipitation experiments after lysing the radiolabeled cells in a hypotonic buffer that allows the proteins to stay in their native conformation. As expected, only a weak band corresponding to the size of gp96 was precipitated using SPA-850 (Fig. 3, Lane SPA-850) or SPA-847, using SPA-850 or SPA-847 antibodies specific for immunoglobulin heavy chain-binding protein and cross-reacting to gp96 (Fig. 3, Lane SPA-847). In contrast,gp96 molecules were nicely precipitated using the scFv anti-gp96 Abs(Fig. 3, Lane H11B). Thus, our recombinant Abs recognize native gp96 molecules much better than the mAb does. The clones B10C and G12D showed equal results in precipitating native gp96 molecules(data not shown).
Chromatographic Purification of gp96 with scFv Anti-gp96 Sepharose Beads.
To isolate gp96 by a one-step chromatographic purification, we coupled the scFv anti-gp96 Ab H11B to Sepharose beads. BSA coupled to Sepharose beads was used as a precolumn to adsorb unspecific material. Precleared hypotonic lysates of the radiolabeled mouse cell line IGELa2 were incubated directly with the scFv-Sepharose beads. The strong radiolabeled band of the protein isolated with H11B-Sepharose beads that migrated close to the 94 kDa marker band demonstrated that the immobilized recombinant scFv H11B Ab is able to recognize gp96 molecules (Fig. 4). This band is absent when BSA-Sepharose beads are used for immunoprecipitation (Fig. 4). For use in a clinically applicable gp96 purification protocol, the recombinant scFv Abs should also recognize native human gp96 molecules. To test whether our recombinant anti-gp96 clones are able to do so, immunoprecipitation experiments were repeated using hypotonic lysates of human C1R cells. In contrast to immobilized BSA (Fig. 4), immobilized scFv H11B Ab was able to precipitate human gp96 molecules (Fig. 4). The differences in intensity between the mouse and human gp96 bands in Fig. 4 are due to higher expression levels of gp96 in IGELa2 cells than in C1R cells.5Immobilized B10C or G12D scFv Abs are able to precipitate murine and human gp96 molecules as well.
scFv-purified gp96 Activates Specific CTLs in Vivo.
Because native gp96 molecules are recognized by the immobilized scFv Abs, we tested whether they are still associated with antigenic peptides and able to induce specific immune responses. The recombinant scFv column was loaded with hypotonic lysates of IGELa2 cells (of BALB/c origin; expressing H2d), and the gp96/scFv-Sepharose beads were injected into C57BL/6 mice. BSA-Sepharose beads incubated with the cell lysate were used as a negative control. Ten days after immunization, recipient splenocytes were stimulated with irradiated BALB.B spleen cells (of the H2b haplotype). Because mice of BALB origin differ from C57BL/6 mice at multiple minor H genes and because gp96 is able to induce cross-priming (9, 15), BALB.B-reactive CTLs should have been induced in the recipients. As shown in Fig. 5,B, mice immunized with the gp96/scFv-Sepharose bead complexes generated CTLs able to recognize minor H antigens expressed on BALB.B blasts. In the control experiment, no CTL activity was detected (Fig. 5,A). These results demonstrate that affinity-purified gp96 is associated with antigenic peptides and able to induce a specific immune response. To test the usefulness of affinity-purified gp96 molecules for clinical applications, we tried to elute gp96 molecules from the Sepharose beads while retaining their immunogenicity. Only the diethanolamine elution buffer with a pH of 10.5 or a buffer containing a high salt concentration (1.3 m NaCl)successfully released the gp96 molecules, as detected by dot blot analysis (data not shown). A buffer with a pH of 4.5 was unable to elute gp96 from the H11B column. To test whether the eluted gp96 molecules are still immunogenic, we repeated cross-priming experiments by immunizing C57BL/6 mice with IGELa2 gp96 eluted from the gp96/scFv-Sepharose beads. gp96 molecules eluted from the scFv H11B-Sepharose column using the high salt buffer (1.3 m NaCl) were indeed able to generate BALB.B-specific CTLs (Fig. 5,F). The affinity-purified gp96 molecules were as efficient in generating a CTL response as the gp96 isolated from the same mouse IGELa2 cells using the standard purification protocol (Fig. 5,D). Both protocols resulted in the purification of comparable amounts of gp96 from a 1-ml cell pellet(about 200 μg; data not shown). However, no CTL response was observed in the mouse immunized with gp96 eluted from the scFv column using the pH 10.5 buffer (Fig. 5,E), suggesting a loss of peptides, and no CTL response was observed in untreated C57BL/6 mice (Fig. 5 C).
In summary, we have selected the antigen-binding domains of gp96-specific Abs from a semisynthetic immunoglobulin gene library. The resulting scFv Abs have been used for the development of a fast,one-step chromatographic purification method for the isolation of gp96 molecules. Like gp96 molecules isolated by the conventional procedure,these are still able to induce specific immune responses. The latter, however, requires ConA-Sepharose column purification, resulting in contamination of gp96 with ConA due to bleeding from the column. Because ConA is a T-cell mitogen, it might harmfully interfere with the vaccinee’s immune system. In addition, proteases such as cathepsin E have been shown to contaminate conventionally purified gp96 preparations (16). All of the potential risks due to gp96 contaminants can be avoided by using affinity purification with the recombinant scFv Abs. In addition, the anti-gp96 scFv Abs speed up the purification process and allow the processing of small tumor samples. These features render our recombinant anti-gp96 scFv Abs an ideal tool for the purification of gp96 to be used in the treatment of human cancer and infectious diseases.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Supported by Deutsche Forschungsgemeinschaft Leibnizprogram Grant Ra369/4-1 (to H-G. R.), Deutsche Forschungsgemeinschaft Sonderforschungsbereich Grant 510, C1 (to H. S.), and European Union Biomed Grant 95-1627. D. A-S. and C. K. contributed equally to this work.
The abbreviations used are: HSP, heat shock protein; minor H, minor histocompatibility; VH and VL, variable region of the heavy and the light chain,respectively; CDR, complementarity-determining region; ConA,concanavalin A; ER, endoplasmic reticulum; mAb, monoclonal antibody;Ab, antibody; MPBS, PBS with 2% skimmed milk powder; mgp96; mouse gp96; PO, peroxidase; scFv, single-chain Fv.
D. Arnold-Schild and H. Schild, unpublished observation.
We thank A. Nissim and G. Winter (MRC Center, Cambridge,United Kingdom) for the semisynthetic Ab gene library. The excellent technical assistance of Martina Finger, Claudia Schmid, and Sylvia Bayertz is gratefully acknowledged. We thank Olga Kouprianova for helpful discussions and Ulrich Christ and Joannis Mytilineos for sequencing the scFv Abs.