Construction and Characterization Of

4314 Vol. 6, 4314–4322, November 2000 Clinical Cancer Research

Construction and Characterization of Bispecific Costimulatory Molecules Containing a Minimized CD86 (B7-2) Domain and Single-Chain Antibody Fragments for Tumor Targeting

Florian Rohrbach, Bernhard Gerstmayer, INTRODUCTION Markus Biburger, and Winfried Wels1 Human tumors of epithelial origin often overexpress mem- 2 Chemotherapeutisches Forschungsinstitut Georg-Speyer-Haus, D- bers of the ErbB/EGFR related family of receptor tyrosine 60596 Frankfurt am Main [F. R., M. B., W. W.], and MEMOREC kinases. This receptor family comprises ErbB/EGFR, ErbB2/ Stoffel GmbH, D-50829 Ko¨ln [B. G.], Germany HER-2/neu, ErbB3, and ErbB4 (1). In particular, overexpression of EGFR and ErbB2 has been shown to directly contribute to malignancy (2, 3). Because of their aberrant expression on ABSTRACT tumor cells and their accessibility from the extracellular space, Efficient T-cell activation requires two signals. The these receptors are suitable targets for directed tumor therapy. first signal, which confers specificity, is provided by in- Mabs specific for the extracellular domains of such receptors teraction of the T-cell receptor with peptides presented by have been shown to interfere with signal transduction and are MHC molecules. One of the second costimulatory signals under evaluation for their therapeutic application in clinical is induced by binding of B7 proteins on the surface of trials (4). To enhance their growth inhibitory potential, recom- antigen-presenting cells to CD28 on the T-cell surface. binant single-chain derivatives of such antibodies have been Expression of B7 molecules on tumor cells can result in constructed that target a potent cytotoxin to tumor cells express- the activation of tumor specific T lymphocytes and induce ing the appropriate receptor (5). In experimental models, this protective antitumor immunity. However, at present such strategy has been applied successfully to reduce tumor load and gene-therapeutic approaches are limited by the inability eliminate metastatic cells (6). However, treatment with such antibodies and cytotoxic antibody fusion proteins does generally to selectively target B7 gene expression to cancer cells. As not result in the induction of specific antitumor immunity and an alternative approach we exploited recombinant anti- cannot prevent possible tumor recurrence if disseminated tumor body fragments to localize a costimulatory B7 molecule to cells escape cytotoxic therapy. To achieve long-lasting antitu- the surface of tumor cells. We constructed chimeric pro- moral effects, immunotherapeutic strategies have been devel- teins that contain in a single polypeptide chain a portion oped that combine antibodies for specific targeting to tumor of human B7-2 (CD86) genetically fused to single-chain cells with cytokines or other immunomodulatory activities for (sc) Fv antibody domains specific for the tumor-associ- the induction of tumor-specific T lymphocytes (7–9). ated antigens epidermal growth factor receptor and the For optimal induction and clonal expansion, T cells require closely related ErbB2 receptor tyrosine kinase. A small at least two activating signals (10). The first signal is antigen- recombinant fragment of human CD86 was characterized specific and is generated by interaction of the T-cell receptor that corresponds to amino acid residues 1–111 (CD86111) with peptide-bearing MHC molecules. A second, costimulatory

of the mature protein. CD86111 produced in the yeast signal is provided by binding of CD28, the major costimulatory ϩ ϩ Pichia pastoris and CD86111 expressed in bacteria was signal receptor on CD4 and CD8 T cells, to its cognate functionally active and displayed specific binding to B7 ligands B7-1 (CD80) and B7-2 (CD86), which are expressed on

counter receptors. Bacterially expressed CD86111-scFv fu- the surface of antigen-presenting cells (11, 12). T-cell receptor sion proteins specifically localized to the respective target stimulation without costimulation can result in T-cell anergy or antigens on the surface of tumor cells and markedly apoptosis. Although many tumor cells express MHC class I enhanced the proliferation of primary T cells when bound molecules and are able to present antigens, most do not provide to immobilized tumor antigen. costimulation, a possible mechanism for tumor cells to evade immune surveillance. Strategies to provide tumor cells with members of the B7 family of costimulatory molecules have led to promising results. In experimental models, rejection of antigenic tumor cells transfected with B7-1 or B7-2 genes has been demonstrated, which also resulted in protection of mice Received 2/29/00; revised 8/11/00; accepted 8/14/00. 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. 2 The abbreviations used are: EGFR, epidermal growth factor receptor; 1 To whom requests for reprints should be addressed, at Chemothera- Mab, monoclonal antibody; sc, single chain; FBS, fetal bovine serum; peutisches Forschungsinstitut, Georg-Speyer-Haus, Paul-Ehrlich- AOX1, alcohol oxidase 1; IPTG, isopropyl ␤-D-thiogalactopyranoside; Strasse 42-44, D-60596 Frankfurt am Main, Germany. Phone: 49-69- FACS, fluorescence-activated cell sorting; GST, glutathione S-transfer- 63395-188; Fax: 49-69-63395-189; E-mail: [email protected]. ase; PBMC, peripheral blood mononuclear cell.

aaaagtcgacgctagcGCTGCTCCTCTG-3Ј) and CD86-IgV-anti- sense (5Ј-aaatctagacgatcgatAAGCACTGACAG-3Ј; CD86 sequence in upper case), which introduce SalI and NheI restriction sites at the 5Ј end and ClaI and XbaI restriction sites at the 3Ј

end of the PCR product. The amplified CD86111 cDNA fragment was digested with SalI and XbaI and ligated in frame 5Ј to a synthetic sequence that encodes the Myc tag and a cluster of six histidine residues (His tag) in a modified pBluescript KSϩ vector (Stratagene, Heidelberg, Germany; Ref. 17). For expres-

sion in the yeast P. pastoris,anNheI/NotI CD86111 fragment including Myc and His tags was isolated from the cloning vector and inserted into the AvrII/NotI-digested yeast expression vector Fig. 1 Localization of the IgV-like domain of CD86 by an attached scFv antibody fragment to a target antigen mimics the activity of pPIC9 (Invitrogen), resulting in the plasmid pPIC9-CD86111 wild-type CD86 expressed as a transmembrane protein. (Fig. 2A). CD86111-scFv fusion genes were derived by isolating the ErbB2-specific scFv(FRP5) and the EGFR-specific scFv(14E1) single-chain antibody domains as ClaI/XbaI fragments from plasmids pWW152-5 (5) and pWW152-14E1 (19), against subsequent challenge with B7-negative wild-type and inserting them between CD86 and Myc/His tag sequences in tumors (13–16). the pBluescript cloning vectors. For bacterial expression of We recently developed an alternative approach to specifi- recombinant CD86 proteins, plasmid pSW5 was constructed by cally localize costimulatory activity to the surface of tumor cells replacing the E. coli ompA signal peptide and synthetic FLAG that does not require transduction or transfection of the cells sequences of the vector pFLAG-1 (IBI Biochemicals, New with B7 gene constructs (Fig. 1). A soluble fusion protein was Haven, CT) with a modified multiple cloning site. Vector back- constructed that consisted of the extracellular domain of the bone, Myc/His tag, and CD86111 and CD86111-scFv fragments human CD86 molecule (CD86225), for CD28 activation, genet- were assembled stepwise, resulting in plasmids pSW5-CD86111 ically fused to an ErbB2-specific scFv antibody fragment for (Fig. 2A), pSW5-CD86111-5 and pSW5-CD86111-14E1 (Fig. tumor targeting (9). Here we have refined this strategy and have 3A). In the yeast and E. coli expression plasmids, start codons characterized a minimized recombinant CD86 domain for translation of the gene products are provided by the vector (CD86111) that, when expressed in yeast and bacteria, retains backbones. binding to B7 counter receptors. Fusion proteins containing this Expression of Truncated CD86111 in the Yeast P. CD86 domain and scFv antibody fragments specific for the pastoris. The pPIC9-derived expression plasmid encodes, un- human EGFR and ErbB2 molecules specifically localize to the der the control of the methanol-inducible AOX1 promoter, the surface of tumor cells expressing the respective target receptors ␣ CD86111 fragment fused to an N-terminal -factor secretion and, upon binding to antigen, stimulate proliferation of primary signal. In addition, the plasmid contains a functional histidinol T lymphocytes. dehydrogenase (HIS4) gene for positive selection in the P.

pastoris HIS4 mutant strain GS115. pPIC9-CD86111 DNA was MATERIALS AND METHODS linearized by SalI digestion and used for transformation of P. Cell Lines and Cell Culture Conditions. CHO and pastoris GS115 cells by electroporation. His4ϩ/methanol-utili- CHO-CTLA-4 cells expressing glycosylphosphatidyl inosi- zationϩ (mutϩ) yeast colonies were isolated on selection media tol-anchored human CTLA-4 on the cell surface (17) were following established protocols (20), and the phenotype of the maintained in MEM␣ with deoxyribonucleosides (Life Tech- resulting clones was confirmed by PCR using AOX1 5Ј and 3Ј nologies, Inc., Karlsruhe, Germany), containing 2 mM gluta- primers (Invitrogen). ␮ ␤ mine, 50 M -mercaptoethanol, 10% heat-inactivated FBS, For expression of recombinant CD86111, a single yeast and 1 mg/ml G418 (CHO-CTLA-4). Murine renal carcinoma colony was grown to an absorbance at 600 nm of 3 in buffered (Renca) cells stably expressing Escherichia coli ␤-galacto- glycerol-complex medium (pH 8). The medium was then ex- sidase and either human ErbB2 (18) or EGFR (6) were changed with methanol-containing buffered methanol-complex cultured in RPMI 1640 supplemented with 10% FBS, 2 mM medium (pH 8), and protein expression was induced for 72 h at glutamine, 100 units/ml penicillin, 100 ␮g/ml streptomycin, 30°C. Yeast cells were removed by centrifugation at 20,000 ϫ

0.25 mg/ml Zeocin, and 0.48 mg/ml G418. Pichia pastoris g. Supernatant containing soluble CD86111 protein was passed GS 115 yeast cells (Invitrogen, Groningen, the Netherlands) through a 45 ␮m filter and applied onto a Ni2ϩ-saturated Che- were propagated in buffered glycerol-complex medium and lating Sepharose column (Amersham Pharmacia Biotech, expression of recombinant proteins was induced in buffered Freiburg, Germany); recombinant protein specifically bound to methanol-complex medium according to the distributor’s rec- the column via the COOH-terminal polyhistidine tag was then ommendations. eluted with PBS containing 250 mM imidazole. Fractions con-

Construction of CD86111 and CD86111-scFv Expression taining the fusion protein were identified by SDS-PAGE and Plasmids. A cDNA fragment encoding the IgV domain of immunoblotting with Mab 9E10 specific for the Myc tag (21), human CD86 (amino acid residues 1–111 of the mature protein; pooled, concentrated, and dialyzed against PBS. A CD86 pro- referred to as CD86111) was derived by PCR using CD86 cDNA tein fragment comprising the complete CD86 extracellular do- Ј as a template (17) and the oligonucleotides CD86-sense (5 - main (CD86225) was expressed in the yeast P. pastoris as

Fig. 2 Construction and expression of recombinant CD86111 proteins. A, schematic representation of the expression cassettes in derivatives of yeast pPIC9 and E. coli pSW5 vectors encoding amino acids 1–111 of human CD86. Expression in the yeast P. pastoris is regulated by AOX1 and is ␣ ␣ directed to the extracellular space via the yeast -factor secretion signal ( ). In E. coli, cytoplasmic expression of CD86111 is controlled by the IPTG-inducible tac promoter (tac). M, Myc tag; H, polyhistidine tag. B and C, SDS-PAGE and immunoblot analysis of recombinant CD86 proteins. B, immunoblot analysis of P. pastoris culture supernatants containing recombinant CD86 proteins comprising the complete extracellular domain

(CD86225; Lane 1) or a CD86 fragment corresponding to the IgV-like domain (CD86111; Lane 2). Lane 3, Coomassie-stained CD86111 protein purified 2ϩ from P. pastoris culture supernatants by Ni affinity chromatography. C, immunoblot analysis of purified CD86111 before (Lane 1) or after (Lane 2) treatment with protein N-glycosidase F. The positions of differently glycosylated forms of CD86111 and of molecular mass standards (kDa) are indicated. Also shown is immunoblot analysis of CD86111 purified from P. pastoris culture supernatants (Lane 3)orE. coli lysates (Lane 4). Immunoblot analysis was performed with the Myc-tag-specific Mab 9E10 and horseradish peroxidase-conjugated antimouse antibody followed by

chemiluminescent detection with the ECL kit. D, binding of recombinant CD86 proteins to CTLA-4. The binding of CD86225 or CD86111 purified from P. pastoris culture supernatants (left and center, respectively) and of CD86111 purified from E. coli lysates (right) to CHO cells (open peak) and CHO-CTLA-4 cells stably transfected with a human CTLA-4 cDNA (filled peak) was detected by FACS analysis with Mab 9E10 and FITC-labeled goat antimouse IgG.

described previously (17). N-Linked glycosylation of purified Bacterial Expression of CD86111 and CD86111-scFv CD86111 protein was analyzed in a deglycosylation reaction. In Proteins. Plasmids pSW5-CD86111, pSW5-CD86111-5 and ␮ brief, 0.2 g of the protein were heated to 100°C for 10 min in pSW5-CD86111-14E1 were transformed into E. coli BL21 PBS containing 0.1% SDS. Triton X-100 at a final concentration (␭DE3) trxBϪ (22). Single colonies were grown at 37°C to an of 1% and 1 unit of N-Gycosidase F (Boehringer Mannheim, absorbance at 550 nm of 0.8–1.0 in Luria-Bertani medium Mannheim, Germany) were added. After incubation for 16 h at containing 0.6% glucose and 100 ␮g/ml ampicillin. Expression 37°C in a total reaction volume of 100 ␮l, samples were ana- of recombinant proteins was induced with 0.2 mM IPTG for 1 h lyzed by SDS-PAGE and immunoblotting with Mab 9E10. at room temperature. Cells were harvested by centrifugation at

dialyzed against 400 mML-arginine in PBS at 4°C. Finally, L-arginine was removed from the samples by stepwise dialysis against PBS.

Binding Assays. The binding of purified CD86111, CD86111-scFv(FRP5), and CD86111-scFv(14E1) proteins to the B7 counter receptor CTLA-4 was determined by FACS analysis using CHO-CTLA-4 cells. A single cell suspension was pre- pared by treatment with trypsin, and 5 ϫ 105 CHO-CTLA-4 cells were incubated for 45 min at 4°C with 0.2–1 ␮g of each of the recombinant CD86 proteins, followed by incubation with 2 ␮g of Mab 9E10 and FITC-labeled goat antimouse IgG (Phar- Mingen, Heidelberg, Germany) for 30 min. Binding of CD86 proteins was detected using a FACScan (Becton Dickinson, Heidelberg, Germany). CHO cells served as a control. Similarly,

the binding of CD86111-scFv(FRP5) and CD86111-scFv(14E1) to their target receptors ErbB2 and EGFR was determined by FACS analysis using ErbB2- or EGFR-expressing Renca-lacZ/ ErbB2 and Renca-lacZ/EGFR cells. For visualization of receptor binding by immunofluorescence, Renca-lacZ/EGFR cells were grown on glass coverslips, washed with PBS, and fixed in a 1:1 (v/v) mixture of acetone and methanol for 10 min at Ϫ20°C. Cells were incubated with ␮ 0.5 g of CD86111-scFv(14E1) in TTBS [10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.05% Tween 20] supplemented with nor- mal goat serum (1:100) for1hat37°C. Bound CD86 fusion protein was detected by incubation with 0.5 ␮g of Mab 9E10 followed by 0.5 ␮g of FITC-conjugated goat antimouse IgG (PharMingen). Each incubation was carried out in TTBS for 45 min at 37°C and followed by washing steps. Cell nuclei were stained with 50 ␮g/ml propidium iodide (Roth, Karlsruhe, Ger- many). Coverslips were mounted on glass slides and analyzed under a fluorescence microscope (Zeiss Axiophot; Carl Zeiss, Jena, Germany). Fig. 3 Construction and bacterial expression of chimeric CD86 fusion T-Cell Proliferation Assays. Spleen cells from BALB/c proteins. A, schematic of CD86-antibody fusion genes in derivatives of mice were suspended in 900 ␮l/spleen of PBS containing 0.5% the E. coli expression plasmid pSW5 encoding amino acids 1–111 of FBS, and incubated with 100 ␮l of MACS rat antimouse CD90 human CD86 fused to the scFv(FRP5) and scFv(14E1) single-chain antibody domains specific for ErbB2 and EGFR, respectively. Cytoplas- (Thy 1.2) antibody-conjugated colloidal super-paramagnetic beads (Miltenyi Biotech, Bergisch Gladbach, Germany) for 15 mic expression of CD86111-scFv proteins is controlled by the IPTG- inducible tac promoter (tac). M, Myc tag; H, polyhistidine tag. B, min at 6°C. Cells were washed, resuspended in 1 ml of PBS ϩ Coomassie-stained CD86111-scFv(14E1) (Lanes 1 and 2) and CD86111- containing 0.5% FBS, and applied on a LS -column (Miltenyi scFv(FRP5) (Lanes 3 and 4) proteins purified from E. coli lysates by Ϫ ϩ Biotech) placed in a magnetic field. CD90 cells were elimi- Ni2 affinity chromatography. C, immunoblot analysis of purified nated from the column by rinsing with PBS containing 0.5% CD86111-scFv(14E1) (Lane 1) and CD86111-scFv(FRP5) (Lane 2) proteins performed as described in the legend of Fig. 1. The positions of FBS. Subsequently, the column was removed from the magnetic ϩ CD86111-scFv proteins and of molecular mass standards (kDa) are field, and CD90 T cells were eluted in 5 ml of the same buffer. indicated. The isolated T cells were kept in DMEM medium supplemented with 10% FBS, 2 mM glutamine, 100 units/ml penicillin, 100 ␮g/ml streptomycin, 50 ␮M ␤-mercaptoethanol, 10 mM HEPES. 5000 ϫ g for 10 min at 4°C, and a cell pellet from 1 liter of Flat-bottomed 96-well plates were coated overnight with the culture was resuspended in 35 ml of PBS containing 8 M urea. mitogenic antimouse CD3 Mab 2C11 (23) alone or in combi- Cells were disrupted in a French press, and the lysates were nation with purified recombinant GST-ErbB2 fusion protein (9). incubated for 45 min at room temperature and clarified by The plates were washed with PBS, and 1 ϫ 105 splenic T cells centrifugation at 16,000 ϫ g for 30 min at 4°C. Recombinant were added to each well with or without the addition of varying

CD86111, CD86111-scFv(FRP5) and CD86111-scFv(14E1) pro- concentrations of soluble CD86111-scFv(FRP5). After incuba- teins were purified from the supernatants via binding of the tion at 37°C for 60 h, the cells were pulsed with 0.25 ␮Ci/well polyhistidine tags included in the molecules to Ni2ϩ-saturated [3H]thymidine (NEN DuPont, Zaventem, Belgium) for 12 h. Chelating Sepharose equilibrated with PBS, 8 M urea and elution The cultures were then harvested, and the incorporation of 3 with PBS containing 8 M urea and 250 mM imidazole. Frac- [ H]thymidine was measured with a liquid scintillation counter tions containing recombinant proteins were identified by (Beckman Instruments, Munich, Germany). SDS-PAGE and immunoblotting with Mab 9E10, pooled, and Human PBMCs were isolated by Ficoll gradient centrifu-

gation and depleted of monocytic cells by adherence to a plastic Functional Characterization of Recombinant CD86111 surface, and 2 ϫ 105 of the remaining cells in 200 ␮l of RPMI Proteins. Full-length CD86 is expressed as a transmembrane medium supplemented with 10% FBS, 2 mM glutamine, 100 protein and binds to the B7 counter receptors CD28 and units/ml penicillin, 100 ␮g/ml streptomycin were added to each CTLA-4 on the surface of T cells (24, 25, 28). To examine the well of flat-bottomed 96-well plates precoated with 1.2 ␮g/well functionality of truncated, soluble CD86111 protein, the interac- of a recombinant protein comprising the extracellular domain of tion of recombinant protein with B7 counter receptors was human EGFR (kindly provided by D. W. Schneider, Berlex investigated by FACS analysis using CHO-CTLA-4 cells. These Biosciences, Richmond, CA). After incubation for 3, 6, or 8 cells stably express the extracellular domain of CTLA-4 attached to the cell membrane via a glycosylphosphatidyl inositol days in the presence or absence of CD86111-scFv(14E1) fusion 3 anchor (17). CHO-CTLA-4 cells were incubated with CD86 protein, the cells were pulsed with 0.5 ␮Ci/well [ H]thymidine 111 proteins purified from yeast culture supernatants or bacterial (Amersham Pharmacia Biotech) overnight. The cultures were lysates. Yeast-expressed CD86 protein, which represents a then harvested, and the incorporation of [3H]thymidine was 225 soluble fragment of the complete extracellular domain of CD86 measured as described above. (17), was included as a control. Specifically bound protein was detected with Mab 9E10 and FITC-labeled goat antimouse IgG. RESULTS Binding of CD86225 as well as truncated CD86111 expressed in yeast and E. coli to CHO-CTLA-4 cells but not to CHO control Construction and Expression of Recombinant CD86 111 cells could be detected (Fig. 2D). These results demonstrated Proteins. We constructed yeast and bacterial expression vec- that the IgV-like domain of CD86 is sufficient to facilitate tors for the production of a soluble, recombinant CD86 fragment binding to B7 counter receptors similar to a CD86 protein comprising 111 N-terminal amino acids of human CD86 (re- containing IgV-like and IgC-like domains. Thus, glycosylation ferred to as CD86111). This fragment includes the IgV-like but of the protein is not required because the bacterially expressed not the IgC-like domain of the protein (24–27). A cDNA CD86111 protein bound to CTLA-4 with comparable efficiency. fragment encoding CD86111 was derived by PCR and inserted Construction and Bacterial Expression of Chimeric into the yeast vector pPIC9 for expression of CD86 as a 111 CD86111-Antibody Fusion Proteins. To localize the CD86 secreted protein. The resulting plasmid pPIC9-CD86111 is domain selectively to the surface of tumor cells, chimeric fusion

shown schematically in Fig. 2A and encodes under the control of proteins were constructed that contained an N-terminal CD86111 the methanol-inducible AOX1 promoter the CD86 fragment fragment fused to recombinant single-chain antibody (scFv) fused to an N-terminal ␣-factor secretion signal from yeast, a fragments with specificity for EGFR and ErbB2 (Fig. 1). cDNA COOH-terminal Myc epitope recognized by Mab 9E10, and a fragments encoding the EGFR-specific scFv(14E1) and the polyhistidine cluster, which facilitates protein purification (17). ErbB2-specific scFv(FRP5) antibody domains were derived P. pastoris GS115 cells were transformed with the expres- from previously described constructs (5, 19) and assembled with sion construct, recombinant clones were isolated on selection CD86111, Myc, and polyhistidine sequences in the bacterial media, and the phenotype of the clones was confirmed by PCR expression vector pSW5. The resulting constructs pSW5- CD86 -5 and pSW5-CD86 -14E1 are shown schematically following established protocols (17). Expression of CD86111 111 111 was induced by the addition of methanol, and recombinant in Fig. 3A. Cytoplasmic expression of CD86111-scFv proteins protein was purified from culture supernatants via Ni2ϩ affinity was induced in E. coli cells harboring the plasmids by the chromatography. In SDS-PAGE analysis, the purified material addition of IPTG; the cells were then lysed under denaturing conditions, and recombinant proteins were purified by Ni2ϩ migrates as a smear of bands ranging from ϳ30 to 50 kDa (Fig. affinity chromatography followed by refolding. After a single 2B, Lane 3). Mab 9E10 immunoblot analysis confirmed that the round of purification, the typical yield of proteins was 1 mg for smear of bands represented recombinant CD86 (Fig. 2B, 111 CD86 -scFv(FRP5) and 3 mg for CD86 -scFv(14E1) per Lane 2; Fig. 2C, Lanes 1 and 3). Bands of higher apparent 111 111 liter of original bacterial culture with a purity of ϳ70–90%, as molecular mass represented differentially glycosylated forms of determined by SDS-PAGE and Coomassie brilliant blue stain- yeast-expressed CD86111 as indicated by the increased electro- ing (Fig. 3B). The identity of chimeric CD86111-scFv proteins phoretic mobility of the protein upon N-glycosidase F treatment was confirmed by Mab 9E10 immunoblot analysis (Fig. 3C). (Fig. 2C, Lane 2). Characterization of CD86111-scFv Binding Specificity. For the generation of unglycosylated CD86 protein, a 111 In the chimeric CD86111-scFv(FRP5) and CD86111- bacterial expression system was used. The CD86111 cDNA scFv(14E1) proteins, antibody domains were included as a fragment and Myc and polyhistidine sequences were assembled means for anchoring and presentation of the CD86 domain on in a pFLAG-1-derived E. coli expression vector, resulting in the the surface of cells via binding to the respective target antigen. plasmid pSW5-CD86111 (shown schematically in Fig. 2A). Af- The functionality of the scFv domains was examined by FACS ter induction of cytoplasmic expression in E. coli cells harboring analysis using murine Renca-lacZ/ErbB2 and Renca-lacZ/ the plasmid, the cells were lysed under denaturing conditions, EGFR cells stably expressing human ErbB2 and EGFR, respec- 2ϩ and recombinant CD86111 protein was purified by Ni affinity tively (6, 18). The cells were incubated with purified CD86111- chromatography and refolded by dialysis. In SDS-PAGE and scFv(FRP5) and CD86111-scFv(14E1) proteins. Cells treated immunoblot analysis, the bacterially expressed, purified with CD86111 protein lacking a scFv domain served as a control. CD86111 protein was detected as a single band of the expected Specifically bound proteins were detected with Mab 9E10 and molecular mass (Fig. 2C, Lane 4). FITC-labeled goat antimouse IgG. The results are shown in Fig.

Fig. 5 Binding of CD86111-scFv fusion proteins to CTLA-4. The binding of purified CD86111-scFv(FRP5) (A, filled peak) and CD86111- scFv(14E1) (B, filled peak) to CHO-CTLA-4 cells was detected by FACS analysis with Mab 9E10 and FITC-labeled goat antimouse IgG. Control cells were incubated with secondary antibodies in the absence of

CD86111-scFv proteins (A and B, open peaks).

with purified CD86111-scFv(FRP5) and CD86111-scFv(14E1) proteins. Specifically bound protein was detected with Mab 9E10 and FITC-labeled goat antimouse IgG. As shown in Fig. 5,

binding of both CD86111-scFv proteins to CTLA-4-expressing CHO cells could be demonstrated. These results illustrate that Fig. 4 Binding of CD86111-scFv fusion proteins to the surface of tumor cells. The binding of purified CD86111-scFv(FRP5) to ErbB2-express- both the B7 domain and the tumor cell recognition domain of ing Renca-lacZ/ErbB2 cells (A, filled peak) and of purified CD86111- the fusion proteins are functional. scFv(14E1) to EGFR-expressing Renca-lacZ/EGFR cells (B, filled Costimulatory Activity of Chimeric CD86 -scFv peak) was detected by FACS analysis with Mab 9E10 and FITC-labeled 111 Protein. We have previously shown that recombinant human goat antimouse IgG. As a control, the binding of CD86111 protein to the cells was also analyzed (A and B, open peaks). C, immunofluorescence CD86 comprising the complete extracellular domain of the microscopy of Renca-lacZ/EGFR cells incubated with CD86111- protein (CD86225) provides a costimulatory signal to murine T scFv(14E1) followed by Mab 9E10 and FITC-labeled goat antimouse cells when immobilized on a plastic surface but not in solution IgG. D, Renca-lacZ/ErbB2 control cells incubated with EGFR-specific (9, 17). Similarly, truncated yeast or bacterially expressed CD86111-scFv(14E1) and secondary antibodies. CD86111 proteins characterized in this study enhanced the proliferation of murine T cells when coated on a plastic surface together with an anti-CD3 antibody (data not shown). In the case

4. Specific binding of CD86111-scFv(FRP5) to ErbB2-express- of chimeric CD86111-scFv proteins, binding to their target an- ing tumor cells (Fig. 4A) and CD86111-scFv(14E1) binding to tigens is intended to immobilize the B7 domain in high density EGFR-expressing tumor cells (Fig. 4B) was observed. The on the surface of tumor cells, thereby facilitating multiple con-

CD86111 control protein did not bind to either tumor cell line. tacts with neighboring T cells and enabling the CD28 clustering Upon treatment of adherent Renca-lacZ/EGFR cells with required for efficient costimulation (30).

CD86111-scFv(14E1) and subsequent immunofluorescence mi- To test the dependence of CD86111-scFv-mediated co- croscopy, strong membrane staining was found (Fig. 4C), stimulation on the binding to the scFv target antigen, we de- whereas Renca-lacZ/ErbB2 control cells incubated with the signed a simplified in vitro assay. Mitogenic 2C11 anti-CD3

EGFR-specific CD86111-scFv(14E1) and secondary antibodies antibody (23) and bacterially expressed GST-ErbB2 fusion pro- displayed only weak nuclear staining (Fig. 4D), most likely tein representing the epitope recognized by scFv(FRP5) (9) because of cross-reactivity of Mab 9E10 with endogenous mu- were immobilized on the surface of tissue culture plates. Splenic rine c-Myc (29). Binding of the CD86111-scFv proteins to their T cells from BALB/c mice were added and incubated for 60 h target receptors was also confirmed by ELISA (19). The appar- in the presence of purified CD86111-scFv(FRP5) protein. Con- ent binding affinities of CD86111-scFv(14E1) to EGFR and trol cells were grown in the absence of GST-ErbB2, anti-CD3, CD86111-scFv(FRP5) to ErbB2, calculated as the half-maximal and/or CD86111-scFv(FRP5). T-cell proliferation was then 3 saturation value, were 32 and 150 nM, respectively (data not measured by [ H]thymidine incorporation. The results are shown). shown in Fig. 6A. CD86111-scFv(FRP5) alone had no effect on

To confirm that the chimeric CD86111-scFv proteins are T-cell growth, whereas stimulation of the cells with anti-CD3 bispecific and in addition to tumor cell recognition facilitate antibody alone resulted in a basic level of T-cell proliferation, interaction with B7 counter receptors, the binding of CD86111- similar to previous observations (17). In the presence of immo- scFv(FRP5) and CD86111-scFv(14E1) to CTLA-4 was investi- bilized anti-CD3 antibody, addition of CD86111-scFv(FRP5) gated by FACS analysis. CHO-CTLA-4 cells were incubated resulted in markedly enhanced T-cell proliferation. For this

Fig. 6 Costimulation of primary T cells by CD86111-scFv fusion proteins. A, costimulatory activity of CD86111-scFv(FRP5). Mitogenic 2C11 anti-CD3 antibody and recombinant GST-ErbB2 fusion protein were immobilized on 96-well plates as indicated. Splenic T cells from BALB/c mice

were added and grown in the presence of purified CD86111-scFv(FRP5) protein. Control cells were grown in the absence of GST-ErbB2, anti-CD3, and/or CD86111-scFv(FRP5) as indicated. B, costimulatory activity of CD86111-scFv(14E1). Recombinant EGFR protein was immobilized on 96-well ␮ plates (1.2 g/well). Human PBMCs were added and grown in the presence or absence of purified CD86111-scFv(14E1) protein for the indicated time periods. Proliferation of murine and human cells was measured by [3H]thymidine incorporation as described in “Materials and Methods.” Each value was determined in triplicates. Bars, SD.

effect, the presence of immobilized GST-ErbB2 was required, by the inability of available vectors to selectively target B7 gene

indicating that CD86111-scFv(FRP5) clustered by binding to the expression to the tumor. target antigen, but not monovalent CD86111-scFv(FRP5) in so- The possibility to generate functional recombinant B7 pro- lution, elicits costimulatory activity and mimics the effect of teins might provide a basis for novel therapeutic concepts. The natural transmembrane B7 on antigen-presenting cells. transmembrane and intracellular domains of CD80 and CD86 The costimulatory activity of the EGFR-specific CD86111- are not required for their costimulatory activity. When expressed scFv(14E1) protein was tested in a similar assay using human in mammalian cells, recombinant proteins comprising the extra- PBMCs. The cells were grown for 3, 6, or 8 days in the presence cellular domain of B7 molecules retain the ability to bind to B7 of purified CD86111-scFv(14E1) on tissue culture plates pre- counter receptors on the T-cell surface (32, 33). We have coated with a recombinant protein comprising the extracellular previously demonstrated that functional soluble forms of human domain of human EGFR (19). Control cells were grown in the CD80 (CD80208) and CD86 (CD86225) can also be produced in absence of the CD86 fusion protein. T-cell proliferation was recombinant form in the yeast P. pastoris. When coated to a 3 then measured by [ H]thymidine incorporation. As shown in plastic surface, such proteins transmit a costimulatory signal and Fig. 6B, the addition of CD86111-scFv(14E1) resulted in mark- in the presence of immobilized anti-CD3 antibody facilitate the edly enhanced cell proliferation if the cells were exposed to the proliferation of primary T cells (17). Here we have characterized fusion protein for 6 or more days. a smaller recombinant fragment of human CD86 that corre-

sponds to amino acid residues 1–111 (CD86111) of the mature DISCUSSION protein. The extracellular domains of CD80 and CD86 contain Efficient activation of T cells requires interaction of the an IgV-like and an IgC-like domain (26). Whereas for CD80 T-cell receptor with peptides presented by MHC molecules and sequences, both domains appear to be critical for CD28 binding induction of a costimulatory signal by binding of B7 proteins on (34), in the case of CD86, the IgV-like domain contained within the surface of antigen-presenting cells to CD28 on the surface of CD86111 is sufficient for functional activity (this report and Ref. the T cells (11, 12). Direct expression of B7 molecules on tumor 27). Thus, for CD86225 and CD86111 from yeast or CD86111 cells can result in the activation of tumor-specific T lympho- produced in bacteria, no obvious differences in the binding to cytes and induce protective antitumor immunity. Strategies for CTLA-4-expressing CHO cells were noted. Likewise, the pro- the presentation of B7 proteins on cancer cells include trans- teins displayed similar costimulatory activity when immobilized duction or infection of the cells with viral vectors (13–15) and on a plastic surface (Ref. 17 and data not shown). This indicates direct injection of B7 plasmid constructs (31). However, the that previous difficulties in isolating functional recombinant broad application of such gene therapeutic approaches is limited CD86 from E. coli were not attributable to the absence of

glycosylation in bacterially expressed protein but most likely toxins differ in their cell-killing activity on tumor cells expressing both were attributable to folding problems, which appear to be more receptor proteins. Int. J. Cancer, 60: 137–144, 1995. 6. Schmidt, M., Maurer-Gebhard, M., Groner, B., Ko¨hler, G., severe for bacterially expressed CD86225 (17) than for the shorter CD86 molecule analyzed here. Brochmann-Santos, G., and Wels, W. Suppression of metastasis forma- 111 tion by a recombinant single chain antibody-toxin targeted to full-length Combining the ability of antibodies to specifically localize and oncogenic variant EGF receptors. Oncogene, 18: 1711–1721, 1999. to tumor-associated antigens on the surface of cancer cells with 7. Gillies, S. D., Reilly, E. B., Lo, K. M., and Reisfeld, R. A. Antibody- the potent costimulatory activity of recombinant B7 molecules targeted interleukin 2 stimulates T-cell killing of autologous tumor cells. might overcome present limitations of B7 gene therapeutic Proc. Natl. Acad. Sci. USA, 89: 1428–1432, 1992. strategies and antibody-based therapies. Here we report the 8. Bohlen, H., Manzke, O., Patel, B., Moldenhauer, G., Do¨rken, B., von construction of fusion proteins that contain in a single polypep- Fliedner, V., Diehl, V., and Tesch, H. Cytolysis of leukemic B-cells by tide chain the costimulatory IgV-like domain of human CD86 T-cells activated via two bispecific antibodies. Cancer Res., 53: 4310– 4314, 1993. and scFv antibody domains for specific tumor targeting. EGFR 9. Gerstmayer, B., Altenschmidt, U., Hoffmann, M., and Wels, W. and the closely related ErbB2 receptor tyrosine kinases were Costimulation of T cell proliferation by a chimeric B7-2 antibody fusion chosen as suitable targets for such molecules on the tumor cell protein specifically targeted to cells expressing the erbB2 proto-onco- surface because of their enhanced expression on a wide variety gene. J. Immunol., 158: 4584–4590, 1997. of human tumors and their involvement in malignant transfor- 10. Schwartz, R. H. Costimulation of T lymphocytes: the role of CD28, mation (2, 3). We demonstrated that the bacterially expressed CTLA-4, and B7/BB1 in interleukin-2 production and immunotherapy. CD86-scFv fusion proteins are bifunctional. They localize spe- Cell, 71: 1065–1068, 1992. cifically to the respective target antigens on the surface of tumor 11. Boussiotis, V. A., Freeman, G. J., Gribben, J. G., and Nadler, L. M. The role of B7-1/B7-2:CD28/CLTA-4 pathways in the prevention of cells and interact with natural B7 counter receptors. Importantly, anergy, induction of productive immunity and down-regulation of the as shown for the ErbB2-specific CD86-scFv(FRP5) protein, the immune response. Immunol. Rev., 153: 5–26, 1996. molecule was able to markedly enhance the proliferation of 12. Lenschow, D. J., Walunas, T. L., and Bluestone, J. A. CD28/B7 primary T cells when bound to immobilized tumor antigen, but system of T cell costimulation. Annu. Rev. Immunol., 14: 233–258, not in solution. Likewise, the EGFR-specific CD86-scFv(14E1) 1996. protein strongly enhanced the proliferation of human PBMCs 13. Chen, L., Ashe, S., Brady, W. A., Hellstro¨m, I., Hellstro¨m,K.E., Ledbetter, J. A., McGowan, P., and Linsley, P. S. Costimulation of grown in the presence of immobilized EGFR. This suggests that antitumor immunity by the B7 counterreceptor for the T lymphocyte in an in vivo situation, T-cell activation would be restricted to molecules CD28 and CTLA-4. Cell, 71: 1093–1102, 1992. the tumor site because the multiple contacts between B7 mole- 14. Townsend, S. E., and Allison, J. P. Tumor rejection after direct cules and CD28 required for CD28 clustering and activation costimulation of CD8ϩ T cells by B7-transfected melanoma cells. (30) could be provided by cell-bound, but not circulating, mono- Science (Washington DC), 259: 368–370, 1993. valent fusion protein. 15. Hodge, J. W., Abrams, S., Schlom, J., and Kantor, J. A. Induction Because of their smaller size, recombinant scFv antibody of antitumor immunity by recombinant vaccinia viruses expressing B7-1 or B7-2 costimulatory molecules. Cancer Res., 54: 5552–5555, 1994. fragments penetrate tumors more easily than whole antibodies 16. Yang, G., Hellstro¨m, K. E., Hellstro¨m, I., and Chen, L. Antitumor (35). This will also be of importance for the in vivo application immunity elicited by tumor cells transfected with B7-2, a second ligand of reagents that link antibodies with costimulatory activity (9, for CD28/CTLA-4 costimulatory molecules. J. Immunol., 154: 2794– 36, 37). Here we showed that the CD86111 fragment comprising 2800, 1995. the IgV-like domain of human CD86 is sufficient to provide 17. Gerstmayer, B., Pessara, U., and Wels, W. Construction and ex- tumor-specific scFv antibody fragments with costimulatory ef- pression in the yeast Pichia pastoris of functionally active soluble forms fector functions, thereby restricting these chimeric molecules to of the human costimulatory molecules B7-1 and B7-2 and the B7 the minimal functional domains required for activity. Such counter-receptor CTLA-4. FEBS Lett., 407: 63–68, 1997. B7-scFv fusion proteins can be expressed in yeast or E. coli,a 18. Maurer-Gebhard, M., Schmidt, M., Azemar, M., Altenschmidt, U., Sto¨cklin, E., Wels, W., and Groner, B. Systemic treatment with a potential advantage for their production at higher yields and for recombinant erbB-2 receptor-specific tumor toxin efficiently reduces further evaluation of their potential for specific cancer immu- pulmonary metastases in mice injected with genetically modified carci- notherapy in suitable in vivo models. noma cells. Cancer Res., 58: 2661–2666, 1998. 19. Schmidt, M., Vakalopoulou, E., Schneider, D. W., and Wels, W. Construction and functional characterization of scFv(14E1)-ETA—a REFERENCES novel, highly potent antibody-toxin specific for the EGF receptor. Br. J. 1. Klapper, L. N., Kirschbaum, M. H., Sela, M., and Yarden, Y. Cancer, 75: 1575–1584, 1997. Biochemical and clinical implications of the ErbB/HER signaling net- 20. Barr, K. A., Hopkins, S. A., and Sreekrishna, K. Protocol for work of growth factor receptors. Adv. Cancer Res., 77: 25–79, 2000. efficient secretion of HSA developed from Pichia pastoris. Pharmacol. 2. Hynes, N. E., and Stern, D. F. The biology of erbB-2/neu/HER-2 and Eng., 12: 48–51, 1992. its role in cancer. Biochim. Biophys. Acta, 1198: 165–184, 1994. 21. Evan, G. I., Lewis, G. K., Ramsay, G., and Bishop, J. M. Isolation 3. Gullick, W. J., and Srinivasan, R. The type 1 growth factor receptor of monoclonal antibodies specific for human c-myc proto-oncogene family: new ligands and receptors and their role in breast cancer. Breast product. Mol. Cell. Biol., 5: 3610–3616, 1985. Cancer Res. Treat., 52: 43–53, 1998. 22. Proba, K., Ge, L., and Plu¨ckthun, A. Functional antibody single- 4. Fan, Z., and Mendelsohn, J. Therapeutic application of anti-growth chain fragments from the cytoplasm of Escherichia coli: influence of factor receptor antibodies. Curr. Opin. Oncol., 10: 67–73, 1998. thioredoxin reductase (TrxB). Gene, 159: 203–207, 1995. 5. Wels, W., Beerli, R., Hellmann, P., Schmidt, M., Marte, B. M., 23. Leo, O., Foo, M., Sachs, D. H., Samelson, L. E., and Bluestone, Kornilova, E. S., Hekele, A., Mendelsohn, J., Groner, B., and Hynes, J. A. Identification of a monoclonal antibody specific for a murine T3 N. E. EGF receptor and p185erbB-2-specific single-chain antibody polypeptide. Proc. Natl. Acad. Sci. USA, 84: 1374–1378, 1987.

24. Azuma, M., Ito, D., Yagita, H., Okumura, K., Phillips, J. H., Lanier, immunization via codelivery of costimulatory molecule genes. Nat. L. L., and Somoza, C. B70 antigen is a second ligand for CTLA-4 and Biotechnol., 15: 641–646, 1997. CD28. Nature (Lond.), 366: 76–79, 1993. 32. Linsley, P. S., Brady, W., Grosmaire, L., Aruffo, A., Damle, N. K., 25. Freeman, G. J., Gribben, J. G., Boussiotis, V. A., Ng, J. W., Restivo, and Ledbetter, J. A. Binding of the B cell activation antigen B7 to CD28 V. A., Jr, Lombard, L. A., Gray, G. S., and Nadler, L. M. Cloning of costimulates T cell proliferation and interleukin 2 mRNA accumulation. B7-2: a CTLA-4 counter-receptor that costimulates human T cell pro- J. Exp. Med., 173: 721–730, 1991. liferation. Science (Washington DC), 262: 909–911, 1993. 33. McHugh, R. S., Ahmed, S. N., Wang, Y. C., Sell, K. W., and 26. Jellis, C. L., Wang, S. S., Rennert, P., Borriello, F., Sharpe, A. H., Selvaraj, P. Construction, purification, and functional incorporation on Green, N. R., and Gray, G. S. Genomic organization of the gene coding tumor cells of glycolipid-anchored human B7-1 (CD80). Proc. Natl. for the costimulatory human B-lymphocyte antigen B7-2 (CD86). Im- Acad. Sci. USA, 92: 8059–8063, 1995. munogenetics, 42: 85–89, 1995. 34. Peach, R. J., Bajorath, J., Naemura, J., Leytze, G., Greene, J., 27. Rennert, P., Furlong, K., Jellis, C., Greenfield, E., Freeman, G. J., Aruffo, A., and Linsley, P. S. Both extracellular immunoglobulin-like Ueda, Y., Levine, B., June, C. H., and Gray, G. S. The IgV domain of domains of CD80 contain residues critical for binding T cell surface human B7-2 (CD86) is sufficient to co-stimulate T lymphocytes and receptors CTLA-4 and CD28. J. Biol. Chem., 270: 21181–21187, 1995. induce cytokine secretion. Int. Immunol., 9: 805–813, 1997. 35. Milenic, D. E., Yokota, T., Filpula, D. R., Finkelman, M. A., Dodd, 28. Hathcock, K. S., Laszlo, G., Dickler, H. B., Bradshaw, J., Linsley, S. W., Wood, J. F., Whitlow, M., Snoy, P., and Schlom, J. Construction, P., and Hodes, R. J. Identification of an alternative CTLA-4 ligand binding properties, metabolism, and tumor targeting of a single-chain Fv costimulatory for T cell activation. Science (Washington DC), 262: derived from the pancarcinoma monoclonal antibody CC49. Cancer 905–907, 1993. Res., 51: 6363–6371, 1991. 29. Siegel, J., Brandner, G., and Hess, R. D. Cross-reactivity of the 36. Challita-Eid, P. M., Penichet, M. L., Shin, S. U., Poles, T., Mosam- monoclonal antibody 9E10 with murine c-MYC. Int. J. Oncol., 13: maparast, N., Mahmood, K., Slamon, D. J., Morrison, S. L., and Rosen- 1259–1262, 1998. blatt, J. D. A B7.1-antibody fusion protein retains antibody specificity 30. Ledbetter, J. A., Imboden, J. B., Schieven, G. L., Grosmaire, L. S., and ability to activate via the T cell costimulatory pathway. J. Immunol., Rabinovitch, P. S., Lindsten, T., Thompson, C. B., and June, C. H. 160: 3419–3426, 1998. CD28 ligation in T-cell activation: evidence for two signal transduction 37. Holliger, P., Manzke, O., Span, M., Hawkins, R., Fleischmann, B., pathways. Blood, 75: 1531–1539, 1990. Qinghua, L., Wolf, J., Diehl, V., Cochet, O., Winter, G., and Bohlen, H. 31. Kim, J. J., Bagarazzi, M. L., Trivedi, N., Hu, Y., Kazahaya, K., Carcinoembryonic antigen (CEA)-specific T-cell activation in colon Wilson, D. M., Ciccarelli, R., Chattergoon, M. A., Dang, K., Ma- carcinoma induced by anti-CD3 ϫ anti-CEA bispecific diabodies and halingam, S., Chalian, A. A., Agadjanyan, M. G., Boyer, J. D., Wang, B7 ϫ anti-CEA bispecific fusion proteins. Cancer Res., 59: 2909–2916, B., and Weiner, D. B. Engineering of in vivo immune responses to DNA 1999.

Downloaded from clincancerres.aacrjournals.org on September 25, 2021. © 2000 American Association for Cancer Research. Construction and Characterization of Bispecific Costimulatory Molecules Containing a Minimized CD86 (B7-2) Domain and Single-Chain Antibody Fragments for Tumor Targeting

Florian Rohrbach, Bernhard Gerstmayer, Markus Biburger, et al.

Clin Cancer Res 2000;6:4314-4322.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/6/11/4314

Cited articles This article cites 36 articles, 18 of which you can access for free at: http://clincancerres.aacrjournals.org/content/6/11/4314.full#ref-list-1

Citing articles This article has been cited by 2 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/6/11/4314.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/6/11/4314. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.