A fully human recombinant IgG-like bispecific antibody to both the epidermal growth factor receptor and the insulin-like growth factor receptor for enhanced antitumor activity.

Both the epidermal growth factor receptor (EGFR) and the insulin-like growth factor receptor (IGFR) have been implicated in the tumorigenesis of a variety of cancers. Here we propose that simultaneous targeting of both receptors with a bispecific antibody would lead to enhanced antitumor activity. To this end, we produced a recombinant human IgG-like bispecific antibody, a Di-diabody, using the variable regions from two antagonistic antibodies: IMC-11F8 to EGFR and IMC-A12 to IGFR. The Di-diabody binds to both EGFR and IGFR and effectively blocked both EGF- and IGF-stimulated receptor activation and tumor cell proliferation. The Di-diabody also inherited the biological properties from both of its parent antibodies; it triggers rapid and significant IGFR internalization and degradation and mediates effective antibody-dependent cellular cytotoxicity in a variety of tumor cells. Finally, the Di-diabody strongly inhibited the growth of two different human tumor xenografts in vivo. Our results underscore the benefits of simultaneous targeting of two tumor targets with bispecific antibodies.

The lack of specificity of currently available chemo-and radiotherapeutic agents constitutes the major obstacle to the effective treatment of cancer. The common use of the combinatorial therapeutic regimens comprising several cytotoxic agents, e.g. various chemotherapeutics and radiations that hit cancer cells via different mechanisms, is often associated with severe toxicities to the patients. Because of their exclusive specificity and high affinity toward defined targets, monoclonal antibodies (mAb) 1 are emerging as a promising new class of effective cancer therapeutics (1,2). So far the United States Food and Drug Administration has approved eight antibody-based products, including unmodified (or naked) antibodies (Rituxan®, Herceptin®, Campath®, Erbitux®, and Avastin®), radiolabeled antibodies (Zevalin® and Bexxar®), and an antibody conjugate (Mylotarg®), for several oncology indications. Because of limited intrinsic cytotoxicity, unmodified mAb, when used alone, usually only yielded from marginal (e.g. Her-ceptin® and Avastin®) to 10 -12% (e.g. Erbitux®) objective responses in patients with solid malignancies (3,4). The therapeutic efficacy of these antibodies is significantly enhanced when combined with conventional chemotherapeutics and/or radiation, for example, Erbitux® plus a regimen of irinotecan, 5-fluorouracil, and leucovorin in third line refractory colorectal cancer patients yielded a 22.9% of objective response rate compared with that of 11% in patients treated with Erbitux® alone (5), and when combined with high dose radiation Erbitux® significantly prolonged the survival of patients with squamous cell carcinoma of head and neck (6). The dose-limiting toxicities of these combined therapies are usually associated with the cytotoxic components in the regimens. Because of their high specificity and low toxicity, it is generally believed that combination of the antitumor antibodies directed against different tumor-associated targets may yield enhanced therapeutic activity without adding significant toxicity. Clinical application of antibody combination is, however, greatly hindered by a number of factors, including limited availability of antibody products, high cost of each product, and the Food and Drug Administration-associated regulatory issues; it is most likely that each antibody has to be separately tested and approved before being tested and approved in combination. To this end, the development of bispecific or multi-specific antibodies that target two or more tumor-associated antigens simultaneously may offer a novel and promising solution.
In past years, both laboratory and early clinical studies have demonstrated that bispecific antibodies (BsAb) may have significant potential applications in cancer therapy by targeting tumor cells with cytotoxic agents including effector cells, radionuclides, drugs, and toxins (7)(8)(9). Here we explored a new concept of utilizing BsAb by constructing a novel IgG-like antibody molecule that targets two different (but relevant) tumor targets, i.e. growth factor receptors, thus blocking simultaneously two receptor activation and their downstream signaling pathways. In this "proof-of-concept" study, we chose the epidermal growth factor receptor (EGFR) and insulin-like growth factor receptor (IGFR) as our model targets. Both receptors have been implicated in the tumorigenesis of a variety of human cancers (10 -15). Targeted inhibition of EGFR with mAb or small molecular kinase inhibitors, including Erbitux® and Iressa® (ZD1839), has shown great anticancer activity in a number of animal models as well as in various clinical studies (for reviews, see Refs. 16 -20). Similarly, significant tumor inhibition has also been achieved in animal models with several IGFR targeting strategies including antisense oligonucleotides (21), dominate-negative receptor mutants (22), and neutralizing mAb (Refs. 23-25; for review, see Ref. 26). Using the variable domains of two neutralizing human antibodies as the "building blocks," one directed against EGFR and the other against IGFR, we constructed and produced an IgG-like tetravalent BsAb, a so-called "Di-diabody." The Di-diabody bound to both EGFR and IGFR, blocked the receptors from interacting with their respective ligands, and inhibited both EGF-and IGF-stimulated activation of the receptors as well as the receptor-associated downstream signaling pathways. Further, the Di-diabody was able to trigger rapid and efficient IGFR internalization and degradation and mediate effective immune effector function such as antibody-dependent cellular cytotoxicity (ADCC). Finally, the Di-diabody strongly inhibited the growth of human tumor xenografts in vivo.

EXPERIMENTAL PROCEDURES
Cell Lines and Proteins-Human tumor cell lines, DiFi and HT29 (colorectal carcinoma), A413 (cervical carcinoma), MCF7 (breast carcinoma), and BxPC3 (pancreatic carcinoma) were obtained from ATCC and maintained in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum (HyClone, Logan, UT) at 37°C in 5% CO 2 . Recombinant extracellular domain of IGFR and its ligand, IGF-I, and EGF were purchased from R & D Systems Inc. (Minneapolis, MN). Recombinant EGFR extracellular domain was produced at ImClone Systems Inc. (New York, NY). Identification from an antibody phage display library and production from mammalian cells of the anti-EGFR antibody, IMC-11F8, and the anti-IGFR antibody, IMC-A12, have been previously described (23,27). IMC-C225, a clinical grade anti-EGFR antibody, and IMC-1121, a human anti-human vascular endothelial growth factor receptor 2 antibody (28), were produced at ImClone Systems Inc.
Construction and Production of the Bispecific Di-diabody-The variable light (VL) and the variable heavy (VH) domains of IMC-11F8 and IMC-A12 were cloned and assembled into two cross-over scFv fragments, A12VL-linker-11F8 VH and 11F8VL-linker-A12VH, following a procedure previously described (29). A 5-amino acid sequence, Arg-Thr-Val-Ala-Ala, representing the first 5 residues from the N terminus of human light chain constant domain, was used as the linker (29). To construct the tetravalent bispecific Di-diabody, a gene encoding one of the "cross-over" scFv chains, 11F8VL-linker-A12VH, was further fused on its C terminus by overlapping PCR to a gene encoding the Fc fragment (CH2-CH3 domains) of an IgG, via the hinge region, to form a fusion polypeptide, 11F8VL-linker-A12VH-hinge-Fc chain. The fusion polypeptide chain was then subcloned into an expression vector along with its partner, the other cross-over scFv chain, A12VL-linker-11F8 VH (see Fig. 1 for details). The expression vector was transfected into NS0 cells, followed by expression of the soluble Di-diabody in serumfree cell culture and purification of the antibody with protein A chromatography. The purity of the Di-diabody was assayed via SDS-PAGE analysis under both reducing and nonreducing conditions.
Receptor Binding Assays-Two different assays were carried out to examine the binding specificity and efficiency of the Di-diabody. In the first assay, the cross-linking assay, the Di-diabody was tested for its capability in simultaneously binding two target antigens. Briefly, the Di-diabody or the monospecific antibodies (5 nM) were first incubated with a biotin-labeled IGFR (100 ng) in solution and then transferred to a microtiter plate coated with EGFR (100 ng/well), followed by incubation with streptoavidin-horseradish peroxidase (HRP) to measure the plate-bound biotin activity. In the second assay, the direct binding assay, various amounts of antibodies were added to triplicate wells of 96-well plates (Nunc, Roskilde, Denmark) precoated with human IGFR1 or EGFR extracellular domain (100 ng/well) and incubated at room temperature for 1 h, after which the plates were washed three times with PBS containing 0.1% Tween 20. The plates were then incubated at room temperature for 1 h with 100 l of a rabbit anti-human IgG Fc-HRP conjugate (Jackson ImmunoResearch Laboratory Inc., West Grove, PA). The plates were washed, peroxidase substrate was added, and the absorbance at 450 nm was read following a previously described procedure (27).
Cell Proliferation Assays-1 ϫ 10 4 DiFi cells in 100 l of complete medium were seeded in each well of 96-well plates and cultured overnight. Various amounts of antibodies were added in triplicate wells and allowed to culture for 4 days, after which 10 l of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (5 mg/ml; Sigma) was added to each well and incubated for additional 4 h. The plates were washed twice with PBS and incubated with 100 l of HCl/isopropanol (40 mM) at room temperature for 10 min, followed by optical density reading at 570 nm.
Receptor Phosphorylation and Signal Transduction Assay-Tumor cells were plated into 75-mm dishes and grown to 70 -80% confluence, after which the cells were washed twice in PBS and cultured overnight in serum free medium. The cells were first incubated with various antibodies at 37°C for 30 min, followed by stimulation with EGF, IGF, or both at 37°C for 20 min. The cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 0.5 mM Na 3 VO 4, 1 g/ml leupeptin, 1 g/ml pepstatin, and 1 g/ml aprotinin), followed by centrifugation of the lysate at 12, 000 rpm for 10 min at 4°C. Both EGFR and IGFR1 were immunoprecipitated from the cell lysate supernatant by using a mixture of IMC-C225 and IMC-A12, followed by the addition of 20 l of protein A/G-Sepharose beads (Santa Cruz Biotechnology, Santa Cruz, CA). The precipitated receptor proteins were resolved on a 4 -12% Nupage Bis-Tris gel (Invitrogen) and transferred to a polyvinylidene difluoride membrane. Phospho-EGFR and phospho-IGFR were detected on the blot using an anti-phosphotyrosine antibody-HRP conjugate (Santa Cruz Biotechnology). Total receptor proteins loaded on the gel were assayed with a mixture of an anti-EGFR (clone 1005) and an anti-IGFR␤ antibody (clone C20) (both from Santa Cruz Biotechnology). For phosphorylation of Akt and p44/p42 mitogen-activated protein kinase (MAPK), whole cell lysate was resolved by SDS-PAGE using a 10% acrylamide gel, and the phospho-Akt and phospho-p44/p42 were detected with an antibody mixture containing an anti-phospho-Akt and an anti-phospho-p44/p42 antibody (Cell Signaling, Beverly, MA), followed by an anti-mouse antibody-HRP conjugate. Total Akt and p44/p42 proteins were assayed with a mixture of an anti-Akt (Santa Cruz Biotechnology) and an anti-p44/p42 antibody (Cell Signaling). All of the signals were visualized with the ECL reagent (Amersham Biosciences).
Receptor Degradation Analysis-The cells were plated and incubated overnight in serum-free medium. IGF-I (50 nM), EGF (50 nM) (R & D Systems), or various antibodies were then added and incubated at 37°C for up to 4 h. The cells were washed in ice-cold PBS, lysed, and loaded onto a 4 -12% Tris-glycine gel (Invitrogen). The proteins were electrophoresed and transferred to a nitrocellulose membrane following standard protocols. IGFR and EGFR were detected by Western blotting using antibody C-20 against IGFR␤ (Santa Cruz Biotechnology) and IMC-11F8, followed by an anti-rabbit (for C-20) or anti-human (for IMC-11F8)-HRP conjugate. The signals were visualized with ECL reagent (Amersham Biosciences).
Antibody-dependent Cellular Cytotoxicity (ADCC) Assay-Blood was collected from normal volunteers and mixed with an equal volume of Hanks' balanced salt solution without Ca 2ϩ or Mg 2ϩ . The mixture was layered onto a Lymphoprep gradient (ICN Biochemicals Inc., Irvine, CA) and centrifuged at 800 relative centrifugal force ϫ g for 20 min. Peripheral blood mononuclear cells were collected at the interface and washed in HEPES-buffered saline. The cells were pelleted, resuspended in RPMI 1640 medium with 10% fetal calf serum, and stimulated by incubation with 200 units/ml interleukin-2 for 1-7 days. DiFi cells were plated at a density of 10000 cells/well in 50 l in a 96-well U-bottomed plate and incubated with IMC-11F8, IMC-A12, the Di-diabody, or a normal human IgG (Sigma) (50 l) at 37°C for 30 min. Peripheral blood mononuclear cells were added to triplicate wells in a volume of 50 l at an effector/tumor cell ratios of 100/1 and incubated at 37°C for 4 h. After centrifugation at 1500 rpm for 10 min, 100 l of supernatant was transferred to a 96-well flat bottom plate, followed by the addition of 100 l/well lactate dehydrogenase assay reagent (BioVision, Inc., Mountain View, CA) and a reading of the absorbance at 490 nm. Target maximum absorbance was determined by lysing the cells with 50 l of 4% Triton X-100, whereas target spontaneous absorbance was determined in the absence of antibody and effector cells, and effector spontaneous absorbance was determined in the absence of antibody and target cells. Total cytotoxicity was calculated as: the percentage of cell lysis ϭ (experimental OD 490 nm Ϫ effector spontaneous Ϫ target spontaneous)/(target maximum Ϫ target spontaneous) ϫ 100. The percentage of specific cell lysis mediated by the antibodies was determined as: the percentage of cell lysis in the antibody-treated group Ϫ the percentage of cell lysis in the normal human IgG group.
In Vivo Efficacy Studies-Female athymic nu/nu mice, 6 -8 weeks of age (Harlan Sprague-Dawley, Inc., Indianapolis, IN), were injected subcutaneously on the lateral dorsal surface with BxPC3 (2 ϫ 10 6 / mouse) or HT-29 (5 ϫ 10 6 /mouse) tumor cells. When tumors reached ϳ200 -300 mm 3 , mice were randomized by tumor size and divided into treatment groups. The mice were treated by IMC-11F8 (or the equivalent IMC-C225), IMC-A12, IMC-11F8 (or IMC-C225) plus IMC-A12, the Di-diabody, or the control articles (saline or normal human IgG). Each antibody was administered by intraperitoneal injections at 40 mg/kg (or 80 mg/kg for the Di-diabody in the HT29 model) twice a week. Tumor growth was evaluated twice weekly with tumor volume calculated as: /6 ϫ (length ϫ width 2 ), where length ϭ longest diameter and width ϭ diameter perpendicular to length. Statistical analysis of tumor growth inhibition was performed using a Repeated Measures analysis of variance program (JMP 5.0.1).

Construction and Production of the Di-diabody-
The variable regions of two previously characterized antagonistic antibodies, the anti-EGFR IMC-11F8 and the anti-IGFR IMC-A12, were used as the "building blocks" to construct the IgG-like tetravalent BsAb, the Di-diabody. A bispecific diabody was first constructed, followed by fusion of one of the diabody cross-over scFv chains to the Fc domain of an IgG (see Fig. 1, A and B, for details). Coexpression in mammalian cells of the Fc fusion along with the other cross-over scFv resulted in an IgG-like tetravalent molecule with two binding specificities (Fig. 1B). The Di-diabody was produced by stably transfected NS0 cells in serum-free conditions and purified from the cell culture supernatant via a protein A affinity column. No significant aggregation was observed when the purified Di-diabody preparation was assayed by an analytical size exclusion chromatography light scattering analysis. Electrophoresis analysis of the Didiabody under nonreducing conditions yielded two major protein bands with expected mobility, the Fc fusion (the top band) in dimer form (molecular mass, ϳ100 kDa) and the cross-over scFv (the lower band) in monomer form (molecular mass, ϳ25 kDa) (Fig. 1C, lane 2). IMC-11F8, an IgG, yielded a single protein band of ϳ150 kDa (Fig. 1C, lane 1). Under reducing conditions, the Di-diabody also gave rise to two bands: the top band at ϳ50 kDa (representing the Fc fusion in monomer form) and the lower band at ϳ25 kDa (the cross-over scFv) (Fig. 1D, lane 2). As controls, IMC-11F8 gave two major bands: the IgG heavy chain (50 kDa) and the IgG light chain (25 kDa) (Fig. 1D,  lane 1).
The Di-diabody Binds to Both EGFR and IGFR-Two assays were used to demonstrate that the Di-diabody was capable of binding to both EGFR and IGFR. In the first assay, the crosslinking assay, we examined whether the Di-diabody could bind to both of its targets simultaneously. Various antibodies were first incubated with a biotin-labeled IGFR in solution and then transferred to a 96-well plate coated with EGFR, followed by incubation with streptoavidin-HRP to measure the plate-bound biotin activity, i.e. the amount of IGFR that was cross-linked to the immobilized EGFR by the antibody. As shown in Fig. 2A, only the Di-diabody, but not the monospecific IMC-11F8 and IMC-A12, was able to cross-link IGFR in solution with the immobilized EGFR, as demonstrated by the plate-associated biotin activity.
In the second direct binding assay, the Di-diabody was compared with its monospecific counterparts in antigen-binding efficiency. IMC-A12 and IMC-11F8 bound only to their respective targets, whereas the Di-diabody reacted to both immobilized EGFR and IGFR, with moderately lower efficiencies compared with their monospecific counterparts (Fig. 2, B and C). The ED 50 values, i.e. the antibody concentrations that yield 50% of maximum binding, to EGFR were 0.05 nM for IMC-11F8 and 0.1-0.2 nM for the Di-diabody, and to IGFR were 0.1 nM for IMC-A12 and 0.25-0.5 nM for the Di-diabody.
Inhibition of Tumor Cell Proliferation in Vitro by the Di-diabody-Next we examined the efficacy of the Di-diabody in inhibiting tumor cell proliferation in vitro in comparison with their monospecific counterparts, using a well characterized tumor cell line, DiFi, which is known to depend on EGFR for survival and growth (30). The anti-EGFR IMC-11F8 significantly inhibited the proliferation of DiFi cells, whereas the anti-IGFR IMC-A12, as well as the control antibody (IMC-1121), showed no effect (Fig.  3). Combination of both IMC-11F8 and IMC-A12 yielded similar activity to that of IMC-11F8 alone. The Di-diabody demonstrated a much weaker anti-proliferative activity compared with the mono-specific IMC-11F8; the IC 50 values, i.e. the antibody concentrations required for 50% tumor growth inhibition, were ϳ1, 1, and 25 nM for IMC-11F8 (alone), IMC-11F8 (plus IMC-A12), and the Di-diabody, respectively.
The Di-diabody Blocks Signaling Pathways Stimulated by Both EGF and IGF-The Di-diabody was examined on it efficacy in blocking EGF-and IGF-stimulated receptor phosphorylation and downstream signal transduction. Although incubation of MCF-7 cells with individual growth factor, EGF or IGF, results in significant levels of phosphorylation of the respective receptor, combination of EGF and IGF yields activation of both EGFR and IGFR (Fig. 4, lanes 2-4). As expected, when the tumor cells were stimulated with both EGF and IGF, treatment with either IMC-A12 or IMC-11F8 only inhibited phosphorylation of the individual receptor (Fig. 4, lanes 6 and   FIG. 1. Construction and produc
The effect of IMC-11F8, IMC-A12 and the Di-diabody on the two major downstream signal transduction molecules associated with both EGFR and IGFR, Akt and p44/p42 MAP kinases, were also studied in MCF-7 cells (Fig. 4). Stimulation with IGF results in significant phosphorylation of Akt (Fig. 4, lane 2), whereas EGF causes strong phosphorylation of p44/p42 MAPK (Fig. 4, lane 3). As expected, combination of IGF to EGF leads to activation of both Akt and p44/p42 MAPK (Fig. 4, lane  4). In the presence of both EGF and IGF, IMC-11F8 significantly inhibited the activation of MAPK but only moderately reduced the activation of Akt (Fig. 4, lane 7), whereas IMC-A12 strongly reduced Akt phosphorylation but was less effective in p44/p42 MAPK activation (Fig. 4, lane 6). Similar to the observation with the receptors, a combination of IMC-11F8 and IMC-A12, either as a mixture or as the Di-diabody, effectively blocked phosphorylation of both Akt and p44/p42 MAPK induced by EGF and IGF (Fig. 4, lanes 5 and 8).
The Di-diabody Triggers Efficient IGFR Internalization and Degradation-We previously demonstrated that IMC-A12 was capable of down-regulating tumor cell surface expression of IGFR by inducing rapid and efficient receptor internalization and degradation (23). Here we investigated whether the Didiabody retained the receptor modulation activity of IMC-A12 on tumor cells. Consistent with our previous observation, IMC-A12 triggered significant IGFR internalization and degradation in MCF-7 cells after incubation at 37°C for 4 h (Fig. 5, A  and B). Incubation with the Di-diabody led to significant degradation of IGFR in both MCF-7 and BxPC3 cells (Fig. 5). This IGFR modulation effect of the Di-diabody is both dose-depend-

FIG. 2. Bispecific and dose-dependent binding of the Di-diabody to EGFR and IGFR.
A, receptor cross-linking assay. Various antibody preparations were first incubated with a biotin-labeled IGFR in solution and then transferred to a microtiter plate coated with EGFR, followed by incubation with streptoavidin-HRP to measure the plate-bound biotin activity. B and C, dose-dependent binding to immobilized EGFR and IGFR by the Di-diabody. Various amounts of antibodies were added to 96-well plates coated with human EGFR (B) or IGFR extracellular domain (C) and incubated at room temperature for 1 h, after which the plates were washed three times with PBS containing 0.1% Tween 20. The plates were then incubated at room temperature for 1 h with a rabbit anti-human IgG Fc-HRP conjugate. The plates were washed, peroxidase substrate was added, and OD 450 nm was read. The data shown are the representative of three similar experiments and are the means Ϯ S.D. of triplicate samples. Various amounts of the antibodies were added into the culture and incubated with the cells for 4 days, after which 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide was added and incubated for additional 4 h. The plates were washed twice with PBS and incubated with HCl/isopropanol at room temperature for 10 min, followed by optical density reading at 570 nm. The data shown are the representative of three similar experiments and are the means Ϯ S.D. of triplicate samples. ent, with the maximum effect, similar to that of IMC-A12, achieved at an antibody concentration of 100 nM (Fig. 5A), and time-dependent; greater than 90% of the IGFR was degraded in BxPC3 cells after 4 h of incubation at 37°C (Fig. 5C). The ligand IGF, IMC-11F8, and the control antibody, IMC-1121, did not show any effects on IGFR modulation. In contrast to their activity on IGFR, none of the treatments, including the ligand EGF, the Di-diabody, and IMC-11F8 (alone or in combination with IMC-A12), demonstrated any modulation effects on the total level of EGFR in both MCF7 and BxPc3 cells (Fig. 5).
The Di-diabody Mediates Effective ADCC on Tumor Cells-In addition to blocking growth signals by interfering with growth factor/receptor interaction and down-regulating receptor surface expression, antitumor IgG antibodies can also cause direct tumor cell killing via mediating effective ADCC. Here we examined whether the Fc-containing Di-diabody is capable of mediating tumor cell killing in the presence of human effector cells. As shown in Fig. 6, IMC-11F8 showed good lysis activity to both A431 and BxPC3 cells but was ineffective to MCF-7 cells. On the other hand, IMC-A12 was only effective toward MCF-7 cells but failed to kill A431 and BxPC3 cells. The Di-diabody, similar to the combination of both IMC-11F8 and IMC-A12, was equally potent in mediating cell killing to all three tumor lines.
The Di-diabody Inhibits the Growth of Human Tumor Xenografts in Nude Mice-We previously showed that both IMC-A12 and IMC-11F8, when used alone, were very efficacious agents in inhibiting the growth of human tumor xenografts in nude mice (23,30). Here we compared the antitumor efficacy of the Di-diabody with its parent antibodies, alone and in combination. In these models, nude mice bearing established xenografts of ϳ 200 -300 mm 3 were treated with the various antibodies twice a week by intraperitoneal injections. In the first model of BxPC3 pancreatic tumor xenografts (Fig. 7), both IMC-11F8 and IMC-A12 alone (at 40 mg/kg) yielded significant (77 and 58%, respectively; p Ͻ 0.05 compared with the PBS group, p Ͼ 0.05 between IMC-11F8 and IMC-A12) tumor growth inhibition at 6 weeks post-treatment initiation. The Di-diabody (at 40 mg/kg) demonstrated similar antitumor activity (52% tumor growth inhibition, p Ͻ 0.05 compared with the PBS group, p Ͼ 0.05 among the Di-diabody, IMC-11F8 and IMC-A12) to the individual parent antibodies when given at the same dose (40 mg/kg). Combination of both IMC-11F8 and IMC-A12 (at 40 mg/kg of each antibody) resulted in the best antitumor activity (Ͼ90% tumor growth inhibition) among the groups ( Fig. 7; p Ͻ 0.05 compared with the Di-diabody or the individual antibody group).
In the second xenograft model, a colorectal carcinoma cell line, HT29, was used. HT29 xenografts were less responsive to individual anti-EGFR and anti-IGFR antibody therapies (Fig.  7). Treatment with either IMC-A12 or IMC-C225, an anti-EGFR antibody that is functionally equivalent to IMC-11F8 regarding both in vitro and in vivo antitumor activity (31) treatment, respectively (p Ͻ 0.05 compared with the saline and the human IgG groups). The Di-diabody yielded a tumor growth inhibition rate of 58%, which is comparable with that achieved by the combination of both IMC-C225 and IMC-A12 (63% tumor growth inhibition, p Ͻ 0.05 compared with both the PBS and the human IgG groups). There is no statistically significant difference in overall tumor inhibition; however, between groups treated with the Di-diabody, the antibody combination, or the individual antibody at the end of the study. DISCUSSION Compelling evidence suggest that both EGFR and IGFR play important roles in the growth and progression of a variety of human cancers; thus they may represent excellent targets for effective cancer intervention. In fact, consistent antitumor effects have been observed both experimentally and clinically with a number of strategies that antagonize either individual receptor activity, including the use of antagonistic antibodies and small molecular tyrosine kinase inhibitors (16 -26). We hypothesize that a strategy that targets both EGFR and IGFR simultaneously, by using either a combination of two antagonistic antibodies or a BsAb, may yield greater antitumor activity than other approaches that address only a single receptor. In our previous study we produced a BsAb, using IMC-11F8 (anti-EGFR) and IMC-A12 (anti-IGFR) as the building blocks, in a (scFv)4-IgG format (27,32) and demonstrated that the BsAb was as potent as the combination of IMC-11F8 and IMC-A12 in neutralizing both EGF-and IGF-stimulated receptor activation and downstream signal transduction (27). The BsAb was, however, difficult to produce in mammalian cells because of its low level of expression, thus preventing any further studies especially the in vivo animal testing. Here we engineered a novel IgG-like BsAb, the Di-diabody, that can be efficiently produced in mammalian cells and be purified by conventional protein A chromatography in a single step. This Di-diabody format should be readily applicable to the construction of other BsAb from antibodies recognizing any pairs of antigens.
In this study we demonstrated that simultaneous blockade of both EGFR and IGFR activation, either by a combination of an anti-EGFR and an anti-IGFR antibody or by a BsAb may lead to broader and enhanced antitumor activity. Tumor cells may gain their growth advantage and/or resistance to apoptosis by (over)expressing a number of growth factor receptors including EGFR and IGFR (10 -15). Binding of ligands/growth factors leads to receptor activation and downstream signal transduction, resulting in cell proliferation, invasion, and increased resistance to apoptosis (10 -15). Because of the redundancy of growth signaling pathways in tumor cells, inhibition of one receptor function (e.g. EGFR) could be effectively compensated by up-regulation of other growth factor receptor (e.g. IGFR)mediated pathways. For example, a recent study has shown that malignant glioma cell lines expressing equivalent EGFR had significantly different sensitivity to EGFR inhibition depending on their capability of activating IGFR and its downstream signaling pathways (33). Other studies have also demonstrated that overexpression and/or activation of IGFR in tumor cells might contribute to their resistance to chemotherapeutic agents, radiations, and antibody therapy (34 -37), and consequently, inhibition of IGFR signaling has resulted in increased sensitivity of tumor cells to these therapeutic agents (38,39). Taken together, these observations strongly suggest that simultaneous blockade of both EGFR and IGFR may provide significant antitumor benefits over individual receptortargeted therapies. Here we showed that the Di-diabody almost completely inhibited, as efficient as the combination of IMC-11F8 and IMC-A12, both EGF and IGF-stimulated activation of EGFR and IGFR, as well as the downstream signaling molecules, including Akt and MAPK p44/p42 (Fig. 4). In contrast, treatment with IMC-A12 or IMC-11F8 alone only inhibited the activity of its respective receptor and receptor-associated signaling molecules (Fig. 4). When tested in vivo, the combination of IMC-11F8 and IMC-A12 yielded better antitumor activity than did each individual antibody in both BxPC3 and HT29 xenografted models, demonstrating the benefits of dual receptor targeting. Although less potent than IMC-11F8 in inhibiting EGFR-dependent tumor cell proliferation in vitro, the Didiabody was equally efficacious to IMC-11F8 in vivo in BxPC3 xenografts and showed a trend of enhanced antitumor activity (similar to the combination of IMC-11F8 and IMC-A12) in the more antibody-resistant HT29 xenografts. The levels of receptor expression, the activation status of the receptor, and its downstream signaling pathways may be partly responsible for the different sensitivities of various tumor xenografts to antibody (alone and in combination) and Di-diabody therapies (40,41). Taken together, these results indicate that the anti-EGFR ϫ anti-IGFR Di-diabody may represent a novel and powerful approach to more effective cancer treatment with a broad antitumor spectrum and thus may be applicable to tumors with etiology based on EGFR, IGFR, or both.
In addition to direct blockade of growth factor/receptor interactions and inhibiting the activation of the subsequent signaling cascades, several other mechanisms of action may also play important roles in the antitumor activity of anti-receptor antibodies. Binding of anti-receptor antibodies to the tumor cell surface may trigger receptor internalization followed by degradation of the receptor in lysosome and/or proteasome compartments, leading to down-regulation of receptor expression, and ultimately, cell growth inhibition and/or apoptosis. In addition, antibodies may also recruit host effector mechanisms, such as ADCC and complement-mediated cytotoxicity, via its Fc region, to directly kill target tumor cells. Our Di-diabody inherited from its parent antibody, IMC-A12, the capability of inducing rapid and efficient IGFR down-regulation, and from both IMC-11F8 and IMC-A12, the efficacy in mediating ADCC activity toward tumor cells that express either EGFR or IGFR. It is interesting to note that although IMC-11F8 only showed good ADCC activity to A431 and BxPC3 cells and IMC-A12 was only effective toward MCF-7 cells, the Di-diabody was equally effective in killing all three tumor cell lines. This phenomenon may be explained by the difference in the levels of EGFR and IGFR expression on these tumor cells under the assay conditions. It is believed that for an effective ADCC to occur, the tumor cells should express the target on their surface (and maintain the expression, i.e. no significant modulation) at a level that is above a certain threshold. We have previously shown that both A413 and BxPC3 cells express a huge number of EGFR (in millions) but much fewer numbers of IGFR, whereas MCF-7 expresses much more IGFR with minimal expression of EGFR (23,27). Further, as shown in Fig. 5, incubation with IMC-A12 at 37°C for 4 h (the assay condition for ADCC) resulted in almost complete depletion of IGFR from BxPC3 cell surface but only caused ϳ50% IGFR down-regulation in MCF-7 cells. It is therefore reasonable to believe that both A413 and BxPC3 may not have sufficient IGFR expression during the assay conditions to be efficiently engaged by IMC-A12 to trigger measurable ADCC activity. On the same note, MCF-7 cells may express too few EGFR to be targeted by IMC-11F8. Conceivably, the Di-diabody thus, like the combination of IMC-11F8 and IMC-A12, may mediate its ADCC effect on BxPC3/A431 and MCF-7 cells via engagement of different receptors on the tumor cell surface.
A major obstacle in the development of BsAb has been the difficulty in producing the materials in sufficient quality and quantity via the traditional methods, including the hybrid hybridoma and chemical conjugation (42). In contrast to rapid and significant progress with various recombinant BsAb fragments (43-44), only limited success has been achieved in the past years in both engineering and production of full-length IgG-like BsAb (27,32,(45)(46)(47), for review, see Ref 48). BsAb fragments are smaller than full-length IgGs, so they have better solid tumor penetration rates, but their small size and lack of an intact Fc also results in their being cleared rapidly from circulation, leading to a short in vivo half-life. Further, BsAb fragments do not require glycosylation, so they can be produced in high yield in bacteria. Compared with the full-length IgGlike BsAb, these fragments are, however, incapable of promoting effector function such as ADCC. Here we described a novel approach to the efficient production of an IgG-like BsAb, the Di-diabody. There are several noteworthy characteristics associated with this BsAb molecule. Firstly, each Di-diabody is a tetravalent molecule comprising two binding sites to each of its targets. Bivalency in some instance may be required for certain antibodies to exert their therapeutic functions, e.g. cross-linking the receptors on target cell surface to stimulate activation, to induce apoptosis, or to promote receptor internalization (as in the IGFR case). In addition, antibody bivalency usually leads to higher binding avidity that is often desirable and may even be necessary for each arm of a BsAb destined for human therapy to demonstrate its activity (49). Secondly, the Di-diabody contains a full IgG Fc region and possesses approximately the same molecular mass as an IgG (150 kDa). Compared with the other Di-diabody format we reported earlier in which the diabody was fused to the CH3 domain only (i.e. no CH2 domain) of the Fc region (50), the full Fc-containing new format confers the Di-diabody not only a long serum half-life but also the capability of supporting secondary immune function, such as ADCC (Fig. 6). The Di-diabody had a clearance half-life in nude mice of ϳ7 days (as determined by an anti-human Fc enzymelinked immunosorbent assay), which is comparable with that of an IgG (for example, both IMC-11F8 and IMC-A12 demonstrated a half-life in nude mice of ϳ5 days). The Fc region also allows the Di-diabody to be purified, like other IgG mAb products, directly via conventional protein A affinity chromatography. Finally, the Di-diabody can be efficiently produced in mammalian cells at a high expression level. The Di-diabody is being routinely produced at 300 -400 mg/liter from a stable NS0 cell line cultured under unoptimized conditions in a spinner bioreactor. In contrast, our previous version of the anti-EGFR ϫ anti-IGFR BsAb, the (scFv)4-IgG construct (27), expressed poorly in mammalian cells and presented significant challenges to the production of sufficient material for in vivo testing.
A number of issues associated with the Di-diabody construct remain, however, to be further addressed. First, in addition to the active tetravalent BsAb, there is evidence of production of nonactive scFv-Fc dimer that is devoid of the other cross-over scFv polypeptide. Careful control of the balance of production of both the scFv-Fc fusion and the partnering cross-over scFv, for example via genetic manipulation of the expression vector, to favor the formation of the tetravalent molecule may reduce the secretion of the nonactive component. Second, although the Di-diabody maintained its binding activity to both EGFR and IGFR when incubated in vitro in mouse serum at 37°C for up to 7 days (not shown), the molecule appeared to be much less stable in vivo. For example, although both circulating IMC- FIG. 7. Inhibition of growth of human tumor xenografts in nude mice by the Di-diabody. Female athymic nu/nu mice were injected subcutaneously on the lateral dorsal surface with BxPC3 (2 ϫ 10 6 /mouse) or HT-29 (5 ϫ 10 6 /mouse) tumor cells. When tumors reached ϳ200 -300 mm 3 , the mice were randomized by tumor size and divided into treatment groups. The mice were treated by IMC-11F8 (or the equivalent IMC-C225), IMC-A12, IMC-11F8 (or IMC-C225) plus IMC-A12, the Di-diabody, or the control articles (saline or normal human IgG). Each antibody was administered by intraperitoneal injections at 40 mg/kg (or 80 mg/kg for the Di-diabody in the HT29 model) twice a week. Tumor volume and body weight of each animal was measured twice a week. The data represent the means Ϯ S.E. of tumor sizes from 10 to 12 animals in each group.
11F8 and IMC-A12 retained Ͼ90% of their antigen binding activity 7 days after administration in mice, the Di-diabody appeared to lose its activity starting at 6 h post-injection; it retained Ͼ90, ϳ30 -40, and Ͻ10% of binding efficiency to both EGFR and IGFR at 6, 24, and 168 h post-administration, respectively. Based on the fact that there is no covalent linkage between the two polypeptides forming the bispecific diabody on each arm of the Di-diabody (Fig. 1A), this loss of antigen binding activity in vivo is most likely a result of dissociation of the two polypeptides in the circulation followed by the rapid clearance of the smaller cross-over scFv fragment (ϳ25 kDa) through the kidney, leading to the production and circulation of the nonfunctional half-molecule, the scFv-Fc dimer. This also explains the apparent discrepancy between the physical halflife (i.e. ϳ7 days) and the biological half-life (i.e. Ͻ24 h) of the Di-diabody in vivo, because the nonfunctional scFv-Fc was detected by the anti-human Fc enzyme-linked immunosorbent assay (used for the determination of the physical half-life) but not by the receptor-binding enzyme-linked immunosorbent assay (used for the determination of the biological half-life). Structural modifications in the Fv interfaces between each cognate VL and the VH domains within the diabody molecule, for example the installation of new disulfide bonds (51), may offer an efficient solution to enhancing the stability of the Di-diabody. It is plausible that a more stable Di-diabody should lead to an improved pharmacokinetics in vivo, hence enhanced antitumor activity.