A B7.1-Antibody Fusion Protein Retains Antibody Specificity and Ability to Activate Via the T Cell Costimulatory Pathway¹

T cell activation and proliferation require two signals from APCs. The first signal is Ag specific and mediated by recognition of antigenic peptides presented in the context of MHC-I or II by the TCR. A second or costimulatory signal can be provided via binding of a costimulatory ligand of the B7 family on the APC to the CD28 counter-receptor present on T cells. The B7 family includes several Ig-like molecules, including B7.1 and B7.2. Provision of signal 1 without signal 2 may lead to a state of immune tolerance (1). B7.1 gene transfer into nonimmunogenic tumor cells has been shown to elicit a T cell-mediated immune response not only against transfected (B7⁺) but also against parental nontransfected tumor cells (2, 3, 4, 5, 6, 7, 8). Since T cell activation requires both B7.1 activation and TCR engagement, only cells with TCRs that recognize antigenic determinants on tumor cells should be activated (9). B7.1 gene transfer requires either ex vivo manipulation of tumor cells, which is technically difficult, or in vivo delivery via gene therapy vectors, which would not specifically target systemic tumor deposits. As an alternative approach we propose to target B7.1 to tumor cells via an Ab fusion protein with specificity for a tumor-associated surface Ag. The B7.1 costimulatory molecule localized at the tumor site by the Ab should aid in the activation of a systemic antitumor immune response.

The HER2/neu oncogene has been found to be amplified (>5⁺ copies) and/or overexpressed in as many as 30% of human breast cancers, 10 to 30% of ovarian cancers, and a subset of lung and other cancers (10, 11, 12, 13). Humanized anti-HER2/neu Ab has been demonstrated to be an effective therapeutic agent in several phase I and II clinical trials (14, 15). A phase II trial showed response in over 11% of patients and has led to ongoing phase III trials (15). These studies demonstrate the feasibility of targeting metastatic breast cancer through the HER2/neu Ag.

In this report we describe the construction and characterization of an Ab fusion protein in which the T cell costimulatory ligand B7.1 is fused to an anti-HER2/neu IgG3 Ab (B7.her2.IgG3) at the amino terminus of the heavy chain. We show that this fusion protein retains both the ability to recognize the tumor-associated Ag HER2/neu and the T cell costimulatory function of B7.1.

CHO, EL4, SKBR3, Sp2/0, and P3X63-Ag.653 cells were available in the laboratory or obtained from American Type Cell Collection (Rockville, MD). EL4, Sp2/0, and P3X63Ag8.563 cells were cultured in Iscove’s medium supplemented with 5% FBS and l-glutamine, penicillin, and streptomycin (GPS).³ SKBR3 cells were grown in RPMI 1640 medium containing 10% FBS and GPS. CHO cells were maintained in DMEM supplemented with 10% FBS and GPS. CHO/CD28 and CHO/B7 cells as well as the CD28Ig and B7Ig soluble proteins were provided by Dr. P. Linsley (Bristol-Myers-Squibb Pharmaceutical Research Institute, Seattle, WA). CHO/CD28 and CHO/B7 cells were grown in the same medium as CHO cells supplied with 0.2 mM proline and 1 μM methotrexate. CHO cells transfected with the HER2/neu cDNA were maintained under selection with 0.5 mg/ml of Geneticin (Life Technologies, Gaithersburg, MD). Soluble CTLA4Ig was purified from a hybridoma obtained from Dr. J. Allison (University of California, Berkeley, CA) using standard protein A column purification methods.

The plasmid encoding the human HER2/neu cDNA (clone 0483 generously provided by Genentech, Inc., San Francisco, CA) was digested with HindIII, filled in using Klenow polymerase, digested with XhoI, and cloned into the XhoI and filled-in BamHI sites of the retroviral vector LXSN (16). The resulting plasmid was transfected into the PA317 packaging cell line using Lipofectin reagent (Life Technologies) and cells selected in 0.5 mg/ml Geneticin. Culture supernatant from the vector-producing PA317 cells was harvested, filtered through 0.45-μm pore size filters, and used to transduce CHO cells to derive CHO/Her2 cells.

The light chain V domain of the humanized humAb4D5-8 Ab was amplified from the plasmid pAK19 (17) (provided by Dr. P. Carter, Genentech) and fused to the 3′-end of human κ leader sequence by overlapping PCR. The primers used in the first cycle of amplification are: primer a, 5′-GGGGATATCCACCATGG(A/G)ATG(C/G)AGCTG(T/G)GT(C/A)AT(G/C)CTCTT-3′ and b, 5′-GACTGGGTCATCTGGATGTCGGAGTGGACACCTGTGGAG-3′ for the leader sequence using a plasmid encoding human κ light chain sequences as template DNA; and c, 5′-CTCCACAGGTGTCCACTCCGACATCCAGATGACCCAGT-3′ and d, 5′-GCTTGTCGACTTACGTTTGATCTCCACCTTGG-3′ for the V_L sequences using pAK19 as template DNA. The resultant PCR products were mixed and used as template for the amplification with primers a and d. The final PCR product of 470 bp was digested with EcoRV and SalI and cloned into the human κ light chain expression vector previously described (18).

The strategy used to clone the heavy chain variable domain (V_H) from pAK19 is similar to the V_L cloning strategy. The primers used for amplification are: primer a, 5′-GGGGATATCCACCATGG(A/G)ATG(C/G)AGCTG(T/G)GT(C/A)AT(G/C)CTCTT-3′; b, 5′-GACTCCACCAGCTGAACCTCGGAGTGGACACCTGTGGAG-3′; c, 5′-CTCCACAGGTGTCCACTCCGAGGTTCAGCTGGTGGAGT-3′; and d, 5′-TTGGTGCTAGCCGAGGAGACGGTGACCAG-3′. The final 500-bp PCR product encoding the leader fused to the V_H sequences of anti-HER2/neu was cloned as an EcoRV/NheI fragment into the human IgG3 mammalian expression vector previously described (18).

The extracellular domain of the human B7.1 including the leader sequences was amplified using primers 5′-GGCATAAGCTTGATATCTGAACCATGGGC-3′ and 5′-GCGCGGTTAACCGTTATCAGGAAAATGC-3′ and cloned as a HindIII/HpaI fragment at the 5′ end of the (Ser-Gly⁴)³ linker sequences into a pUC19-flex plasmid. The V_H domain of the humanized humAb4D5-8 Ab was amplified by PCR from the plasmid pAK19 using primers 5′-GGCGGCGGATCCGAGGTTCAGCTGGTG-3′ and 5′- TTGGTGCTAGCCGAGGAGACGGTGACCAG-3′, digested with BamHI and HpaI, and cloned at the 3′ end of the B7.1 and flexible linker sequences. The resulting insert encoding the B7.1-linker-V_H sequences was isolated as an EcoRV/NheI fragment and cloned into the expression vector for the IgG3 heavy chain (18).

Purified recombinant anti-HER2/neu Ab alone is referred to in the manuscript as her2.IgG3, and the anti-HER2/neu Ab fused to B7.1 is referred to as B7.her2.IgG3. Transfection, expression, and purification of the recombinant Abs were performed as described previously (19). Briefly, nonsecreting Sp2/0 or P3X63-Ag.653 myeloma cells were transfected with 10 μg of each of the anti-HER2/neu light chain and heavy chain expression vectors by electroporation. Transfected cells were plated at 10,000 cells/well in 96-well U-bottom tissue culture plates. The next day, selection in 0.5 mM histidinol (Sigma Chemical Co., St. Louis, MO) was initiated and maintained for 10 to 14 days. Wells were screened for Ab secretion by human IgG-specific ELISA as previously described, and positive wells were expanded. To determine the size of the secreted recombinant Abs, supernatants from cells grown overnight in medium containing [³⁵S]methionine were immunoprecipitated with goat anti-human IgG (Zymed Laboratories, Inc., San Francisco, CA) and staphylococcal protein A (IgGSorb, The Enzyme Center, Malden, MA). Precipitated Abs were analyzed on SDS-polyacrylamide gels in either the presence or the absence of reducing agents. For purification of her2.IgG3 and B7.her2.IgG3 Abs, Ab secreting Sp2/0 clones were expanded in roller bottles in Hybridoma Serum-Free Medium (Life Technologies), and 2 to 4 l of cell-free medium was collected. Culture supernatants were passed through a GammaBind protein G column (Pharmacia Biotech, Inc., Piscataway, NJ), and the column was washed with 10 ml of PBS. The protein was successively eluted with a total of 10 ml of 0.1 M glycine at pH 4.0, 2.5, and 2.0, and the eluate was neutralized immediately with 2 M Tris-HCl, pH 8.0. The eluted fractions were dialyzed and concentrated using Centricon filters with a molecular mass cut-off of 30,000 Da (Amicon, Inc., Beverly, MA).

Cells were detached by treatment with 0.5 mM EDTA, washed, and incubated with recombinant her2.IgG3 or B7.her2.IgG3 antibodies for 1 to 2 h at 4°C, then washed and stained with FITC-conjugated anti-human IgG (Sigma) or PE-conjugated anti-human B7.1 (Becton Dickinson, San Jose, CA) and analyzed by flow cytometry.

The affinity of B7.her2.IgG3 fusion protein for the Ag was compared with that of the parental her2.IgG3 Abusing the IAsys Optical Biosensor from Fisons Applied Sensor Technology (Paramus, NJ). Soluble HER2/neu Ag (ECD, provided by Genentech) was immobilized on a sensitized microcuvette according to the manufacturer’s instructions. Her2.IgG3 or B7.her2.IgG3, at different concentrations in PBS with 0.05% Tween-20, was added to the cuvette, and association and dissociation were measured. Rate constants were calculated using the FASTfit software (supplied with the IAsys System) as previously described (20).

Human PBMC were isolated from normal donor blood using standard Ficoll-Hypaque density centrifugation. Human T cell enrichment columns (R&D Systems, Minneapolis, MN) were used for T cell purification according to the manufacturer’s instructions. Purified T cells were plated in flat-bottom 96-well tissue culture plates at 1 × 10⁵ cells/well in RPMI supplemented with 5% FBS. Irradiated (5000 rad) CHO, CHO/Her2 or CHO/B7 cells were added at 2 × 10⁴ cells/well in the presence of 0, 1, 5, or 10 μg/ml recombinant her2.IgG3 or B7.her2.IgG3 and 10 ng/ml PMA (Sigma). Plates were incubated at 37°C for 3 days, pulsed with 0.5 μCi/well of [³H]thymidine for 16 to 18 h, and harvested, and [³H]thymidine incorporation was measured.

The expression vectors for the human IgG3 heavy and κ light chains were previously described (18). The variable domains of the anti-HER2/neu Ab were amplified by PCR from the plasmid pAK19 (provided by P. Carter, Genentech) (17) and cloned into the corresponding heavy or light chain expression vectors to derive her2.IgG3. To construct a fusion Ab between her2.IgG3 and B7.1 (referred to as B7.her2.IgG3), the extracellular domain of human B7.1 was cloned at the 5′ end of the heavy chain variable region of her2.IgG3 (Fig. 1 A). A flexible (Ser-Gly⁴)³ linker was provided at the fusion site of the recombinant fusion protein to facilitate correct folding of both Ab and B7.1 domains. We elected to express B7.1 at the amino terminus of the heavy chain because B7.1 fused to the carboxyl terminus of the C_H3 domain showed decreased affinity for CD28 (data not shown). These results are consistent with a critical role for the amino terminus of B7.1 in mediating its biologic activity (21). The light chain and either the her2.IgG3 or B7.her2.IgG3 heavy chain expression vectors were cotransfected into Sp2/0 myeloma cells, and stable transfectants secreting soluble proteins were identified by ELISA.

To determine the molecular mass and assembly of the transfected proteins, cells were grown overnight in [³⁵S]methionine, and the secreted proteins were immunoprecipitated and analyzed by SDS-PAGE. In the absence of reducing agents, her2.IgG3 migrates with an apparent molecular mass of 170 kDa, while B7.her2.IgG3 is about 250 kDa (Fig. 1,B, lanes 1 and 2, respectively). Following treatment with 2-ME, light chains of 25 kDa were seen for both proteins, while her2.IgG3 had a heavy chain of approximately 60 kDa, and B7.her2.IgG3 had a heavy chain of approximately 100 kDa (Fig. 1 B, lanes 3 and 4, respectively). Therefore, proteins of the expected molecular mass were produced, and the fusion of the extracellular domain of B7.1 to the her2.IgG3 heavy chain did not appear to alter the secretion of the fully assembled H₂L₂ form of the Ab.

We tested the ability of recombinant her2.IgG3 and B7.her2.IgG3 to bind to the HER2/neu antigenic target by flow cytometry (Fig. 2). CHO cells stably expressing the HER2/neu Ag (CHO/Her2) derived by retroviral-mediated gene transfer and nontransduced CHO cells were incubated with either her2.IgG3 or B7.her2.IgG3. Binding was assayed by staining with either FITC-conjugated anti-human IgG or PE-conjugated anti-human B7.1 Abs followed by flow cytometry. Both her2.IgG3 (Fig. 2, A and B) and B7.her2.IgG3 (Fig. 2, D and E) bound specifically to CHO/Her2 and not to parental CHO cells. Therefore, fusion of the extracellular domain of B7.1 to a complete her2.IgG3 Ab resulted in a fusion Ab capable of specifically recognizing the HER2/neu Ag through the Ab domain. CHO/Her2 cells incubated with B7.her2.IgG3 also stained positively with anti-human B7.1, indicating that binding of B7.her2.IgG3 to the Ag through its Ab domain did not interfere with Ab recognition of the B7.1 fusion domain (Fig. 2 F).

The affinities of the her2.IgG3 and B7.her2.IgG3 Abs for the HER2/neu Ag were compared using the IAsys biosensor (Fig. 3). Her2.IgG3 or B7.her2.IgG3, at 1 × 10⁻⁷ M, was added to a cuvette with soluble HER2/neu Ag ECD immobilized on its surface, and the association and dissociation were measured as the samples were added and washed from the cuvette. The calculated affinity of 1.7 × 10⁻⁷ M for B7.her2.IgG3 was decreased about 2.5-fold compared with the affinity of 7 × 10⁻⁸ M obtained for the parental her2.IgG3. The modest decrease in affinity primarily reflected a reduction in the dissociation constant of B7.her2.IgG3.

The ability of the B7.1 domain in the B7.her2.IgG3 fusion protein to bind to its receptors CTLA4 and CD28 was studied by two different methods. Soluble CTLA4Ig and CD28Ig immobilized on nitrocellulose membrane were incubated with either her2.IgG3 or B7.her2.IgG3 (Fig. 4,A). We observed strong binding of B7.her2.IgG3 to CTLA4Ig, but no binding of her2.IgG3. B7.her2.IgG3 also bound CD28Ig, although with a lesser affinity than to CTLA4Ig. This was expected, since the reported affinity of B7.1 for CTLA4 is 20-fold higher than that for CD28 (22). In another experiment, we used CHO cells stably expressing CD28 to detect B7.her2.IgG3 binding (Fig. 4 B). Parental CHO or CHO/CD28 cells were incubated with either B7Ig (a gift from Dr. P. Linsley) or B7.her2.IgG3 and washed, and binding was detected by staining with FITC-conjugated anti-human IgG followed by flow cytometry. Specific binding of B7.her2.IgG3 and B7Ig to CD28 present on CHO-CD28⁺ cells, but not to control CHO cells, was observed.

Since recruitment and activation of tumor-specific T cells would depend on the presence of B7.1 on the tumor cell surface, we characterized the stability of B7.her2.IgG3 bound to the HER2/neu Ag expressed on the cell membrane. SKBR3 cells from a human breast cancer cell line known to express high levels of HER2/neu were incubated with 10 μg/ml of either her2.IgG3 or B7.her2.IgG3 at 4°C to allow maximum binding. The cells were then washed and incubated at 37°C in culture medium. At different times (0, 1, 3, or 24 h), an aliquot of cells was taken and stained with FITC-conjugated anti-human IgG and analyzed by flow cytometry. The results obtained at 0 and 24 h are shown in Figure 5,A. The mean fluorescence measured at different times was compared with that at time zero, when maximum binding was observed. The time course of Ab binding, calculated as a percentage of the maximum mean fluorescence, is illustrated in Figure 5 B. We observed a gradual decline in Ab cell surface staining intensity with time. Significant staining (42% of staining at time zero) was still detected at 24 h for both her2.IgG3 and B7.her2.IgG3.

To test for the functional ability of the B7.her2.IgG3 molecule to signal via CD28, we performed a syngeneic T cell proliferation assay using human peripheral blood T cells (Fig. 6). CHO/Her2 or control CHO cells were irradiated and incubated in the presence or the absence of either her2.IgG3 or B7.her2.IgG3 and peripheral blood-enriched T cells. PMA at 10 ng/ml was added to the cultures to provide signal 1, which was necessary for proliferation. Addition of B7.her2.IgG3 to CHO/Her2 cells resulted in a dose-dependent increase in T cell proliferation as assayed by [³H]thymidine incorporation. Results from two different donors from two experiments are presented. Levels of T cell proliferation obtained with 10 μg/ml B7.her2.IgG3 approached the levels obtained through stable expression of human B7.1 in CHO cells by gene transfer (CHO/B7). Proliferation was absent in the presence of parental CHO cells or with control her2.IgG3. The significantly lower levels of proliferation observed when B7.her2.IgG3 were incubated with CHO cells suggests that binding of B7.her2.IgG3 to the cell surface via the HER2/neu Ag is necessary for optimal T cell costimulation. Visual inspection of the coculture plates showed formation of large foci of proliferating T cells in response to control CHO/B7 cells or in response to incubation with CHO/Her2 cells in the presence of B7.her2.IgG3. Photographs of the cocultures are included in Figure 6 B. The presence of proliferating T cell colonies directly correlated with the level of proliferation detected by [³H]thymidine incorporation.

We describe the construction and characterization of a fusion Ab in which the extracellular domain of the B7.1 costimulatory molecule was fused by genetic engineering to the amino terminus of the heavy chain of an anti-HER2/neu Ab. We opted to use the IgG3 backbone for the Ab molecule, since the extended hinge region of IgG3 would be expected to provide greater flexibility in folding to accommodate the presence of B7.1 in the fusion Ab. IgG3 also exhibits Fc-mediated functions, such as complement activation and Fcγ binding (23). We chose the B7.1 costimulatory ligand in preference to B7.2, as Gajewski et al. and other investigators have suggested that B7.1-transduced tumors more successfully induce CTL activity and protect against parental tumor challenge more effectively than tumors transduced with B7.2 (24, 25, 26, 27). Although conflicting results with respect to Th1 vs Th2 differentiation have been reported using B7.1 and B7.2, results from several experimental systems suggest that B7.1 costimulation tends to favor differentiation along the Th1 pathway (1, 28, 29, 30). Therefore, we chose to link B7.1, rather than B7.2, to an antitumor Ab in an effort to preferentially stimulate a Th1-mediated immune response.

Our results indicate that B7.1 can be effectively linked to the amino terminus of the heavy chain of an anti-HER2/neu Ab, with retention of both Ab specificity and the B7.1 interaction with CD28. Binding to HER2/neu was demonstrated by flow cytometry as well as IAsys biosensor studies, albeit at a lower affinity than that observed for the control her2.IgG3. Possible reasons for the observed decrease in affinity could be steric hindrance between the anti-HER2/neu variable and the B7.1 domains or a change in the kinetics of Ag binding due to the increased size of B7.her2.IgG3. Similarly, specificity of B7.1 for both CTLA4 and CD28 was demonstrated by the ability of B7.her2.IgG3 to bind soluble CTLA4Ig and CD28Ig as well as CD28 expressed on the surface of target cells. Of note, in preliminary attempts to derive a fusion Ab, we constructed an anti-dansyl Ab fusion in which the B7.1 coding sequences were fused to the carboxyl end of the heavy chain C_H3 domain. We did not observe binding of this B7.1 fusion Ab to CD28 expressed on the surface of CHO/CD28 cells or to soluble CD28Ig, suggesting that fusion through the B7.1 amino terminus may disrupt the B7.1/CD28 interaction. This may be due to masking of the amino terminal sequences of B7.1, which are known to be in close proximity to the CD28/CTLA4 binding site (21). Whether fusion via a flexible linker will restore binding of B7.1 to CD28 is currently under investigation. At present, however, we would favor fusion of B7.1 sequences to the amino-terminal heavy chain sequences for suitable B7.1/CD28 interaction. A similar requirement for fusion at the amino terminus of the Ab to maintain activity was observed for nerve growth factor (31). Since the Fc domain remains intact in B7.her2.IgG3, binding to Fc receptors may induce Ag-dependent cellular cytotoxicity or otherwise affect B7.1 function. If this is a problem, further manipulation of the constant region could be performed to eliminate FcR binding sites.

We also sought to address whether anti-HER2/neu Abs would remain on the surface of Ag-presenting breast cancer cells. Our results indicated that approximately 40% of surface-bound B7.her2.IgG3 was detectable by flow cytometry for up to 24 h following initial incubation with human SKBR3 breast cancer cells. This suggests that stable presentation of the B7.1 costimulatory ligand on the tumor cell surface may be feasible, and that loss of presentation due to internalization or rapid antigenic shedding via HER2/neu binding may not be a significant problem (32).

The management of minimal residual disease is a central problem in breast cancer and other solid tumors. Despite the use of increased dose intensity of chemotherapy or autologous bone marrow transplantation, relapse remains a critical problem (33, 34, 35, 36). Chemotherapeutic strategies are necessarily limited by various toxicities. Additional modalities that can achieve further cytoreduction are needed. Although various clinical trials of mAbs, Ab-based conjugates, and/or radioantibodies have been performed, the results of these trials have highlighted obstacles to successful Ab-based therapy of human malignancy. Abs generally are not directly cytotoxic due to poor fixation of complement and/or inadequate activation of Ab-dependent cytotoxicity. Effective use of Abs for delivering cytotoxic agents (e.g., conjugates such as Ab-ricin, or radiolabeled Ab strategies) requires delivery to a majority of, if not all, tumor cells (37). An alternative approach is to elicit an active systemic immune response against tumor cells. Delivery of cytokines has been shown to induce an antitumor T cell response. Although gene transfer has most commonly been used to achieve increased cytokine levels at the site of the tumor, recent studies performed using an Ab-IL-2 fusion protein suggest that Abs can be used for delivering cytokines to tumors. The Ab-cytokine fusion protein retains both Ab specificity and cytokine activity and appears to be more effective than either used alone or in combination, but not covalently linked (38, 39, 40, 41, 42, 43). However, fusion of B7.1 rather than cytokine should result in activation of T cells with TCRs that specifically recognize tumor determinants rather than the nonspecific activation expected of a fused cytokine.

In assays for T cell costimulation in vitro, we show that effective stimulation of human T cells was achieved only if B7.her2.IgG3 was bound to a HER2/neu target, and limited or no stimulation was observed using target cells that did not express HER2/neu Ag. This suggests that the B7.her2.IgG3 fusion protein in soluble form may not be able to effectively provide a costimulatory signal to preactivated T cells, and that the anti-HER2/neu Ab domain in the B7.her2.IgG3 fusion protein provided specificity for the T cell costimulation. This property of B7.her2.IgG3 would allow enhanced specificity of the immune response. Similar results have been reported using fusion of the B7.2 costimulatory ligand to a single chain Ab (44). However, a single chain Ab produced in yeast cells may have considerably different glycosylation and antigenicity as well as different pharmacokinetics in vivo compared with the humanized B7.her2.IgG3 we have produced. The relative specificity and type of response achieved with B7.her2.IgG3 fusion compared with the B7.2 single chain fusion remain to be determined. It also remains to be determined whether the specificity and response achieved with B7.1 fusions will differ from those observed using bispecific antitumor/anti-CD28 Ab (45). However, genetically engineered Ab fusion proteins should present fewer problems in manufacture and purification than the described bispecific Abs, which are difficult to purify to homogeneity.

In summary, we have produced an anti-HER2/neu IgG fusion protein encoding the extracellular domain of the B7.1 costimulatory ligand. This protein retains targeting specificity via the HER2/neu Ag as well as ability to deliver a T cell costimulatory signal. The strategy offers several theoretical advantages. While expression of HER2/neu may be heterogeneous, targeting via HER2/neu may activate T cells with specificity against other unidentified tumor-associated Ags, resulting in destruction of both HER2/neu-positive and nonexpressing cells. Therefore, the Ab fusion protein may allow targeting of micrometastatic disease with relative specificity and would not itself have to bind to all tumor cells to elicit an effective response. The data presented suggest that tumor-specific Abs fused with costimulatory ligands may be a useful method for delivering a costimulatory signal for the purpose of cancer immunotherapy.

The authors thank Dr. P. Carter (Genentech, Inc.) for providing sequences of the humanized humAb4D5 Ab, and Dr. V. Planelles for carefully reviewing the manuscript.