World Bispecific Antibody Summit, September 27–28, 2011, Boston, MA

Abstract

With more than 30 therapeutic monoclonal antibodies (mAbs) approved and annual global sales of the products at ~$50 billion in 2010, these products have proven to be successful in many ways. Nevertheless, there is room for improvement in performance, and substantial unmet medical needs remain. As a consequence, numerous organizations are devoting resources to engineering novel mAbs such as bispecific antibodies that have increased functionality compared with unmodified IgG molecules. The World Bispecific Antibody Summit, organized by Hanson Wade, drew over 100 participants to Boston to discuss engineering novel bispecific antibodies, generating lead candidates and clinical study and commercialization of the molecules. Approaches such as the trifunctional antibody (TRION), dual variable domain-Ig (Abbott), two-in-one (Genentech), dual affinity retargeting (MacroGenics), kappa-lambda body (NovImmune), bispecific T-cell engager (Micromet) and chemical generation (CovX/Pfizer) were discussed in detail. In addition, posters describing bispecific Affibody® molecules for targeting of EGFR and HER2 (Affibody), T-cell receptor based bi-specifics that target HLA-peptides (Immunocore), a novel mAb-Fv bispecific antibody format utilizing Fc region (Xencore), generation of a tetravalent bispecific antibody against IL4 and IL13 for the treatment of idiopathic pulmonary fibrosis (Sanofi), Combining Affibody® molecules and the Albumod™ technology to create long acting multispecific protein therapeutics (Royal Institute of Technology, Affibody) and COVA301 as a highly potent bispecific inhibitor of IL-17A and TNF-α (Covagen) were presented.

Abbreviations

Ang	=	angiopoietin
CD	=	cluster of differentiation
CEA	=	carcinoembryonic antigen
EGFR	=	epidermal growth factor receptor
HER2	=	human epidermal growth factor receptor
IL	=	interleukin
TNF	=	tumor necrosis factor
VEGFR2	=	vascular endothelial cell growth factor receptor 2

Day 1: September 27, 2011

Janice M. Reichert presented an overview of the clinical development of bispecific antibodies. To provide context, she noted that an average of ∼20 therapeutic mAbs per year were entering clinical trials in the beginning of the 2000s, but this number steadily grew during 2002–2008 and is currently ∼54. Despite the impressive increase in new mAbs, the number of FDA approvals per year has so far not followed such a dramatic upward trend and has varied from 0–4 since the late 1990s. The main factors affecting the rate of antibodies entering study in the past decades are advances in (1) antibody engineering and design, (2) manufacturing processes, (3) understanding of the mechanism of action and (4) understanding of the targeted molecular pathways. She also mentioned that ten mAb blockbusters generated more than $1 billion each in 2010.

Next, Professor Reichert discussed the particular advantages of bispecific mAbs and the trends in development of bispecific antibodies compared with more conventional formats. Bispecific constructs can potentially enable: (1) simultaneous inhibition of two cell surface receptors, (2) simultaneous blocking of two ligands, (3) crosslinking of two receptors and (4) recruitment of T cells to proximity of tumor cells. Due to their structure and mode of action, bispecifics could potentially be more efficacious and less costly to develop. The number of bispecifics entering clinical studies has varied though the years, but the number is likely to increase. As evidence of this, Professor Reichert noted that five new bispecific antibodies entered clinical studies during January 2010–August 2011 alone. Most bispecific mAbs are in early stage clinical study; ten bispecific mAbs are currently in Phase 1 and two are in Phase 2 (Table 1). However, one mAb product, catumaxomab (Removab^®), was approved by the European Medicines Agency in 2009 for the treatment of malignant ascites. Catuxomab is a mouse IgG2a/rat IgG2b triomab developed by Trion/Fresenius. It simultaneously targets EpCAM and CD3, which results in Fc-mediated activation of macrophages, natural killer (NK) cells and co-stimulation of T-cell response.

To conclude, Professor Reichert briefly discussed to the clinical development and approval time needed for therapeutic mAbs, which averages 8 y, and the factors that could possibly affect the future trends in the sector of bispecifics. Due to the large number and variety of constructs (there are more than 35 ways to generate bispecific mAbs), it is likely that attrition will occur in the preclinical phase and the future clinical pipeline will be dominated by a few types of bispecifics that show superior efficacy (Table 1).

Patrick Baeuerle (Micromet) discussed the case of blinatumomab as an example of advancing a bispecific antibody from research to the clinic. He started his presentation by highlighting the growing interest of the pharmaceutical industry in bispecific antibodies, illustrated by the number of deals that have taken place in recent years. Blinatumomab is a CD19xCD3 bispecific T-cell engaging (BiTE) antibody that functions by activating T cells, leading to cell lysis.1–5 CD19 activates the PI3 kinase pathway and is expressed in several hematologic malignancies.6

Blinatumomab is currently undergoing evaluation in a Phase 1 study of 62 patients with non-Hodgkin's lymphoma (NHL). The primary aim of this study is to evaluate the safety and tolerability of blinatumomab intravenous infusions administered over a total period of 4 or 8 weeks. Most adverse effects (lyphopenia, leucopenia, thrombocytopenia) were transient and not clinically relevant; four deaths were recorded due to disease progression or sepsis. Neurological and psychiatric adverse effects were noted mostly at the onset of treatment and were all fully reversible. Dr. Baeuerle showed images from patient biopsies before and after blinatumomab treatment that demonstrated depletion of bone marrow from lymphoma cells.7 CT scan images from another patient showed response of a bulky mantle cell lymphoma to blinatumomab administration over 58 d. Approximately 75% of patients with diffuse large cell B-lymphoma (DLCBL) have ongoing responses as of the last follow-up in June 2011. Median duration of response in all patients was 508 d, with responses ongoing in 11 patients as of June 2011.

B-cell acute lymphoblastic leukemia (B-ALL) is an aggressive disease with poor prognosis;8 the presence of minimal residual disease (MRD) is the most important negative prognostic factor.9,10 Dr. Baeuerle presented the results of a Phase 2 study of blinatumomab in chemotherapy-resistant B-ALL patients with MRD. The primary endpoint was molecular complete response (CR), while secondary endpoints were time to relapse and occurrence of adverse effects. Molecular CR with blinatumomab was achieved in 80% of patients, mostly within the first cycle of treatment. Relapse-free survival was achieved in 14 out of 20 patients. Most adverse events occurred early and resolved later during treatment; two patients discontinued the regimen.

Dr. Baeuerle also presented interim results from an ongoing Phase 2 trial of blinatumomab in patients with relapse or refractory (r/r) B-ALL. In this study, 75% of patients treated with blinatumomab achieved CR or CR with partial hematological recovery (CRh). All patients with CR or CRh also achieved molecular response with tumor cells content in the bone marrow being below 0.01%. Positive response was reached even in patients expressing high-risk biomarkers. The safety profile of blinatumomab in this group of patients was favorable when compared with existing approaches. In concluding, Dr. Baeuerle highlighted the success of blinatumomab in hematological malignancies in achieving and maintaining high response rates in patients with ALL and NHL subtypes while having a favorable safety profile. Based on clinical data collected to date, the company has begun a focused development program in ALL; the first of several pivotal studies are ongoing.

Peter Ruf (Trion) discussed the clinical development and of catumaxomab (Removab^®), which was developed using the Triomab^® platform. This platform represents a cost-effective approach to the production of bispecific antibodies by chimeric mouse/rat quadroma cells. The mIgG2axrIgG2b quadroma cell line possesses several advantages, including reduction of H/L chain mismatch variants down to 4–10% due to preferential intra-species H/L pairing; efficient heterologous H/H chain pairing; and easy product purification due to the different affinities of mouse and rat IgG for protein A.11 The supernatant from mouse/rat quadroma cell culture is purified through protein A (to remove parental and mismatched mAbs) and ion exchange (to remove final impurities) chromatography in a GMP-compliant production platform.

Because of their trifunctional design, Triomab^® antibodies bind to cancer-specific antigens and recruit both T cells as well as accessory cells, such as macrophages, dendritic cells and NK cells, thereby activating both the adaptive and innate immune system mechanisms. Tumor cell killing is achieved though cell lysis, antibody-dependent cell-mediated cytotoxicity (ADCC), phagocytosis or apoptosis. The apoptotic tumor particles are processed by antigen-presenting cells and subsequently presented to T cells by stimulated Fcγ-RI, RIIa or RIII positive accessory cells, thereby achieving induction of a long-term anti-tumor activity. Dr. Ruf showed that catumaxomab binds preferentially to activating FcγRIIa in comparison to its inhibitor counterpart FcγRIIb.

The lack of relevant animal models for Triomab bispecific antibodies has created the need for surrogate antibodies against homologous mouse proteins (such as the mouse T-cell receptor CD3). Triomabs BiLu (against human EpCAM and mouse CD3) and Surek (against tumor antigen GD2 and mouse CD3) are two such surrogate antibodies used in animal studies. Administration of surrogate Triomab BiLu significantly improved survival in mouse models of B16 melanoma, A20 lymphoma and CT26 adenocarcinoma at doses 4–10 µg/mouse.12–14

Dr. Ruf then discussed results obtained during the CASIMAS study [NCT00822809], a Phase 3b randomized trial of catumaxomab for the treatment of malignant ascites due to epithelial cancers. The monitoring of anti-EpCAM IgG plasma levels revealed increases in anti-EpCAM titers in 12 out of 25 patients following three intraperitoneal catumaxomab injections, indicating active immunization in about 50% of malignant ascites patients with EpCAM-positive tumors. The immune reaction was not restricted to EpCAM, but spread to non-targeted HER-2 tumor-specific antigen, as demonstrated by the increase of anti-HER2 IgG plasma titer. An anti-HER2 IgG booster reaction was observed in 3 out of 4 patients that received a second cycle of catumaxomab treatment.

Gabriele Schaefer (Genentech) described the development of the bispecific mAb MEHD7945A, which targets the cell surface receptors epidermal growth factor receptor (EGFR) and human epidermal growth factor receptor (HER)3. Dr. Schaefer's presentation began with a brief review of the HER/ErbB receptor family and their ligands. Upon ligand binding, the receptors undergo conformational changes that allow for receptor heterodimerization and intracellular signal transduction.15 Although several mAbs have been developed against EGFR, HER2 and HER3, Dr. Schaefer argued that combined blockage of EGFR and HER3 inhibits all major HER family signaling pairs. In particular, HER3 plays a critical role in tumors as an activator of the PI3K signal and several compensatory pathways (EGFR, Met, FGFR2, IGFR1) linked to drug resistance. Next, Dr. Schaefer presented the example of the development of a dual-specificity antibody based on the marketed anti-HER2 mAb trastuzumab (Herceptin^®). Dr. Schaefer showed top-down alanine scanning images of trastuzumab paratopes used for the design of the bispecific mAb. Complementarity-determining region (CDR) heavy and light chain libraries were generated for Fab-phage display and dual action Fab respectively. Using this approach, trastuzumab can be evolved to simultaneously bind two antigens, in this case HER2 and vascular endothelial cell growth factor receptor (VEGFR), with high affinity. The efficacy of this proof-of-concept bispecific mAb has been confirmed in experimental models that are sensitive to anti-HER2 or anti-VEGFR treatment.16

The first step in the generation of bispecific MEHD7945A was the production of the anti-EGFR mAb D1.5. A binding assay using labeled EGFR was used to determine the EC₅₀ of D1.5, which was in the range of 0.2 ± 0.01 nM. D1.5 could reduce EGFR levels and inhibit phosphorylation of AKT, MAPK and Tyr in NR6 cells. The tumor-repressing efficacy of D1.5 in a xenograft model was comparable to that of cetuximab. Next, induction of mutations in the light chain CDRs of D1.5 lead to the generation of a new clone (clone 1) that had acquired binding and inhibitory properties against HER3, but had lost potency against EGFR. Improvement of potency against EGFR was achieved through an affinity maturation protocol that resulted in the production of bispecific IgG1 MEHD7945A, which can bind EGFR or HER3 with either Fab. The generated bispecific mAb blocks ligand binding to both HER3 and EGFR, with K_d (hu HER3) = 0.39 nM and K_d (hu EGFR) = 1.9 nM. Dr. Schaefer presented crystallography structures of MEHD7945A binding complexes with the domain II of EGFR or the domain III of HER3. Due to bispecific Fabs, the binding of MEHD7945A to EGFR is ablated by the extracellular domain of HER3 and vice-versa, as demonstrated by ELISA binding assays. MEHD7945A inhibited EGFR-and HER3-dependent downstream signaling as shown by protein gel blot. In addition, MEHD7945A inhibited the combined MAPK and PI3K signal in A431 epidermal carcinoma and BxPC3 pancreatic cancer cell lines. MEHD7945A demonstrated superior growth inhibition activity compared with cetuximab and anti-HER2/3 monospecific mAbs in H1666 (EGFR⁺, HER2⁺, HER3⁺) and A431 (EGFR⁺) cell lines. MEHD7945A administration (6.25 or 12.5 mg/kg) in a non-small cell lung cancer (NSCLC; H292) xenograft model induced reduction of tumor volume to levels comparable with cetuximab. In a breast xenograft model (MAXF449), MEHD7945A achieved superior reduction of tumor volume compared with cetuximab and a monospecific anti-HER3 mAb. In addition, MEHD7945A showed increased efficacy in a FaDu animal model compared with cetuximab and a monospecific anti-HER3 mAb. Finally, MEHD7945A demonstrated efficacy in a range of tumor in vivo models including NSCLC, colon, vulva/squamous/epidermidoid, head and neck, pancreas and breast.

Dr. Schaefer noted that, in addition to blocking tumor cell growth, MEHD7945A mediates ADCC similarly compared with cetuximab, as shown in A431 cells. MEHD7945A-N297A, a variant of MEHD7945A that lacks ADCC function, has decreased efficacy in xenograft models. The bispecific mAb MEHD7945A enhances gemcitabine-induced cytotoxicity in a range of cell lines. Although most anti-EGFR inhibitors show skin rashrelated toxicity, a study in cynomolgus monkeys showed reduced skin toxicity with MEHD7945A.

Paul Moore (Macrogenics) discussed the development of dual-affinity re-targeting (DART) bispecific proteins and their applications in cancer and autoimmune diseases. DART technology is highly flexible and focused on dual specificity antibody-like proteins capable of targeting multiple different epitopes. Product stability, reduced immunogenicity due to minimal linker size and content, expression in mammalian and prokaryotic systems, and adjustable half-life and avidity are some of the advantages of this platform. To date the company has produced more than 70 DART proteins for more than 35 targets. DART proteins function through various mechanisms, e.g., cell killing via activation of effector cells, blocking of cytokines or cellular receptors, modulation of intracellular signals by cis-targeting membrane receptors, and binding to pathogenic epitopes.

The cis-acting inhibitory mechanism is of particular interest and is being used in novel approaches for the treatment of inflammatory diseases. Dr. Moore presented an example of inhibition of B-cell activation by DART molecules constructed by pairing the Fv region from a mAb targeting CD32B with the Fv region of a mAb directed against CD79B (the b-chain of the invariant signal-transducing dimer of the B-cell receptor complex).

The simultaneous binding of a CD32BxCD79B-DART to the two receptors generates inhibitory signals that block B-cell activation. The CD32BxCD79B-DARTs are stable molecules both in vitro and in vivo, and can be expressed at high levels in mammalian cells. In addition, the CD32BxCD79B-DARTs retain the binding properties of parental mAb Fvs against each receptor.17 Each antigenic region of CD32BxCD79B-DART binds specifically to its target, as demonstrated by flow cytometry using CD32B-transfected CHO and CD79B⁺ Ramos cells. Side by side comparison of Ramos and Daudi cells showed that CD32BxCD79B-DART binds in cis to double-positive cells. Cis binding of a CD32BxCD79B-DART induces phosphorylation of CD32B and SHIP1 within the first 5 min of exposure, as it was shown by protein gel blot. Longer CD32BxCD79B-DART exposure times induce decrease in phosphorylation levels of ZAP70, Syk, ERK1/2 and Akt, indicating attenuation of Syk signaling pathway. CD32BxCD79B-DART inhibits B-cell receptor Ig secretion and reduces primary B-cell proliferation at concentrations as low as 0.032 nM.

Dr. Moore described the development of a DART against mouse CD32 and CD79B that was used as a surrogate in animal studies. The mouse-specific CD32xCD79B DART, but not the parental anti-CD32 and anti-CD79B mAbs, was capable of inhibiting proliferation of spleen mouse B-cell proliferation. Administration of the mouse-specific CD32xCD79B DART also reduced the severity of symptoms in a collagen-induced arthritis animal model.

In addition to inflammatory diseases cis-acting DARTs are also being explored in cancer. A new DART that targets IGF1R and EGFR is being developed to explore the hypothesis that simultaneous blocking of both receptors will exhibit enhanced anti-tumor activity compared with single pathway inhibition. The IGF1RxEGFR Ig-DART incorporates Fc effector functions and can be expressed in mammalian cells under GMP-compliant conditions up to 100 L scale. The IGF1RxEGFR Ig-DART inhibited colon cancer cell growth in vitro; its efficacy was similar to a cocktail of parental anti-IGF1R and anti-EGFR mAbs, but superior to the efficacy each parental mAb alone.

Sean McKenna (EMD Serono) discussed the development of bispecific antibodies based on the strand-exchanged engineered domain (SEED) platform. Dr. McKenna started his presentation by reviewing the two general categories for bispecific antibodies: (1) homodimeric Fc and (2) heterodimeric Fc. Homodimeric Fc-based bispecifics tend to be multivalent and have symmetrical design, but are in general large molecules and difficult to express. Heterodimeric Fc-based bispecific antibodies have asymmetric design, are bivalent, and are the size of IgG, but stabilizing the heterodimerization of the comprising domains is challenging. The SEED platform was designed to generate asymmetric bispecific antibody-like molecules with improved therapeutic potential compared with natural antibodies. The stabilization of the bispecific heterodimer is achieved by exchanging structurally related sequences of immunoglobulin within the conserved CH3 domain. Strand exchange between IgG and IgGA CH3 domains creates two non-identical chains (namely GA and AG) with asymmetric interface, which allows the formation of the heterodimer. Each chain of the generated heterodimer can be fused to different functional domains.18 The generation of the heterodimer was confirmed by comparing the sizes of AG/GA, GA/GA and Fc, by protein gel blot, size exclusion chromatography, and isoelectric focusing gel separation. Thermal stability analysis showed a monophasic melting curve with Tm of the SEED framework within an acceptable range. SEED is a flexible platform that can be used to generate: (1) monovalent single arm molecules that reduce receptor downregulation or block agonist activity, (2) bi-paratopic molecules that target two epitopes on the same molecule for improved selectivity and combined mechanism of action, (3) bispecific molecules that target two antigens on the same cell or connect two (or three) cells and activate effector cell engagement and (4) ligand traps that target two cytokines simultaneously. Fab, svFC or VHH domains can be used to construct the exchangeable strands in mono- and bi-valent formats.

Next, Dr. McKenna described the development of anti-EGFR SEED-Fab variants of C225 (cetuximab; Erbitux^®), which exists in both mono- and bi-valent formats. EGF binding assays showed that Fc and SEED-based bivalent versions had similar blocking capacity while monovalent binders were 10- to 15-fold less efficient. Both SEED formats induced ADCC-related cell lysis of A341 cell line at concentrations of 1–2 ng/ml, with the monovalent form slightly less competent. Cell-based assays also demonstrated that the Fab-SEED fusion is able to induce complement dependent cytotoxicity. Pharmacokinetic studies in mice showed that Fab-SEED constructs had similar pharmacokinetic (PK) properties with Fc-fusions.

Monovalent SEED-C225 formats may contain three Fab chains (C225 VHCH1-AG + GA + C225 VLCL), two Fab chains (C225 VHCH1-AG + C225 VLCL-GA), or one Fab chain (C225 svFC-AG + GA). Thermal stability analysis revealed that only svFc-based monovalent molecules showed marginal Tm values. The latter format also has lower expression yields compared with the other variants. A341 cell membrane-binding assays showed that the bivalent SEED format has a higher antigen affinity (K_i = 0.14 nM) and is more efficient (IC₅₀ more than 1 order of magnitude lower) in blocking EGF binding compared with the monovalent versions.

In concluding, Dr. McKenna described a paradigm of SEED bi-paratopic antibodies (BPAb). C225 (cetuximab) and C425 (matuzumab) are mAbs that bind to non-overlapping epitopes of EGFR and block EGF-induced signaling. Various reports have shown that simultaneous administration of both mAbs can have synergistic effect in tumor models.19,20 A number of bi-paratopic anti-EGFR SEEDs have been constructed by combining Fab or svFC domains of C225 and C425. Isoelectric focus analysis showed that AG/GA heterodimers are predominant in all formats. The bi-paratopic format had significantly increased avidity compared with bivalent 225/225 and 425/425 SEEDS. However, binding assays showed similar blocking capacity between bi-paratopic and bivalent SEEDs. Bi-paratopic SEED, but not the combination of monovalent SEEDs, induced EGFR downregulation in MDA-MB-468 cells. EGFR downregulation was accompanied by dramatic downregulation of Akt and MAPK pathways.

Nicholas Fisher (NovImmune) described the development of a novel bispecific antibody format called kappa-lambda body. This type of bispecific antibody is designed to embody the ideal characteristics of a therapeutic antibody, e.g., fully human, multiple modes of action, high expression (>1 g/L), Fc effector function and extended half-life. The κλ-bodies are human IgG molecules assembled naturally without the requirement of mutations or linkers. The immunoglobulin complex contains one common heavy chain and two light chains (κ and λ). Due to their structure the κλ-bodies display two specificities (A and B). They are compatible with all modes of action enabled by the bispecific antibodies, including immune cell retargeting to tumors, dual inhibition, increased killing specificity and central nervous system delivery. The κλ-body design requires the identification of two antibodies with common heavy chain. Although this design could initially appear counterintuitive, as the VH is the main contributor to antigen binding energy, Dr. Fisher presented evidence supporting that light chain diversity is sufficient to achieve specificity.

The process of Fv identification implements fixed VH libraries that are combined with separate κ and λ libraries to produce the bispecific κλ-bodies, reaching a total diversity of 2 × 10¹⁰ possible configurations. As an example, Dr. Fisher described the phage display selections against interferon (IFN)γ and IL6RC using ELISA screenings of λ and κ libraries, respectively. The identified κ and λ chains were reformatted into IgG with affinities ranging between 0.2–400 nM without affinity maturation. The chains H, Lκ and Lλ were co-expressed in PEAK cells after transfection using proprietary tri-cistronic vectors. The antibodies produced are κλ-bodies (50% of the total IgG molecules), mono-λ (25%) and mono-κ (25%) bodies. The free light chains, the mono-κλ bodies and other contaminants are removed during a 3-step industrial scale compatible process, generating a purified κλ-body product. Purification efficiency is controlled by protein gel blot. Use of the proprietary vectors ensures the controlled relative expression of the three chains (H, Lκ and Lλ) and maximizes the assembly efficiency of the κλ-body. The produced κλ-body was characterized in a binding assay simultaneously against hIFNγ and hIL6R.

NovImmune has generated κλ-bodies against five targets to date. The large-scale manufacturing of κλ-bodies is performed in CHO cells similarly to other standard industrial methods. Briefly, the CHO cells are transfected by electroporation and the stable or semi-stable clones are transferred to bioreactors that generate up to 25 L of cellular supernatant. The number of viable CHO cells and κλ-body titer reaches maximum values at approximately 200 h in culture. The κλ-body yield during the process is 2–6 g/L, similar to standard mAb productivity. The product is then purified; the purity control and characterization of κλ-bodies obtained from CHO cells is performed using protein gel blot and chromatography techniques. Overall, the κλ-bodies represent a unique class of unmodified, fully human, bivalent, bispecific, Fc-bearing molecules that enable multiple modes of action.

Modular bispecific antibodies (mAb^2TM) were discussed by Max Woisetschläger (f-star). The mAb² represent a new generation of therapeutic mAbs with longer half-life, immune effector functions and additional binding sites in non-CDR-loops. The favored formats for this category of bispecific molecule are engineered Fc (Fcab) and bispecific IgG (mAb²). The Fcab (∼1/3 of the IgG size) are Fc with two antigen binding regions via engineered binding sites in CH3 domains. Dr. Woisetschläger presented data on the Fcab HER2-1, which binds with high affinity to receptor HER2 (EC₅₀ approximately 2 nM). An epitope recognition assay demonstrated that HER2-1 Fcab binds to a different epitope than trastuzumab. In addition, HER2-1 Fcab can kill HER2⁺ HCC1954 cells through ADCC.

Dr. Woisetschläger then briefly discussed the role of the LewisY (LeY) antigen, a tetrasaccharide cell surface receptor normally expressed during embryogenesis that is also expressed in the majority of cancers. VL311 is an anti-LeY mAb that kills LeY-expressing cells via ADCC. The Fc of VL311 was replaced with an Fcab binding HER2 (Fcabs HER2-1 or HER2-2), generating a bispecific mAb² that binds simultaneously to both HER2 and LeY. The LeY antigen binding in VL311-HER2-2 mAb² is unaffected by the binding of HER2, as demonstrated with a MCF7 cells binding assay. In addition, the presence of HER2-2 Fcab does not compromise the ADCC activity of the mAb², which is comparable with that of parental VL311. Interestingly, VL311-HER2-2 mAb² only kill cells that express both antigens and has no effect on cells expressing only LeY or HER2. Cell toxicity assays using HCC1954 cell line showed that co-crosslinking of LeY and HER2 is required for ADCC-mediated cell death. Forced co-crosslinking of Fcab HER2-1 and VL311 with secondary mAb or multivalent HER-2 crosslinking with mAb² trastuzumab-HER2-1 is not sufficient for cell killing. Annexin V staining and TUNNEL assay revealed apoptotic events (membrane flipping, DNA fragmentation), but no caspase pathway activation in cells exposed to mAb² VL311-HER2-1. Results obtained using the MitoScreen assay suggested that VL311-HER2-1 induces de-polarization of mitochondrial membrane potential, providing further evidence of an apoptosis-mediated cell death.

The Thomsen-Friedenreich (TF) glycoepitope is another cell surface receptor recognized by mAb BW835 mentioned by Dr. Woisetschläger. The mAb BW835 was used to generate a bispecific antibody similarly to VL311-HER2-1 mAb². Interestingly, primary human mammary epithelial cells were resistant to cell toxicity induced by BW835-HER2-1 and VL311-HER2-1. PK studies in mice showed no differences between VL311-HER2-1 and the parental antibody. Although LeY and HER2 are expressed in a range of tumors, their levels appear to be particularly high in the gastric tumor GFX281, rendering this cancer a suitable model to test VL311-HER2-1 mAb² in vivo.

Figures and Tables

Table 1 Bispecific antibodies in clinical study sponsored by commercial firms

Company	INN or code name	Targets	Clinical status*
Sanofi	SAR156597	IL4 x IL13	Phase 1
Affimed	AFM13	CD30 x CD16A	Phase 1
Trion	FBT-A05, Bi20	CD20 x CD3	Phase 1
Micromet	MT110	EpCAM x CD3	Phase 1
Micromet	MT111, MEDI 565	CEA x CD3	Phase 1
Immunocore	IMCgp100	gp100 x CD3	Phase 1
Immunomedics	TF2	CEA x IMP288 hapten	Phase 1
Genentech	MEHD-7945A	EGFR x HER3; dual IgG1	Phase 1
Merrimack	MM-111	HER2 x HER3	Phase 1
Pfizer	CVX-241	Ang2 x VEGF; dual IgG1	Phase 1
Ablynx	Ozoralizumab, ATN103	TNF x albumin	Phase 2
Micromet	Blinatumomab, MT103	CD19 x CD3	Phase 2

* Most advanced phase of clinical study listed on company web page or clinicaltrials.gov as of September 2011.

Day 2—September 28, 2011

Chairman Jonathan Davis (Adnexus/Bristol-Myers Squibb) reminded the audience that reports describing the preparation of bispecific antibodies were published in the early 1980s, but this initial work on bispecifics as therapeutics did not lead to a marketed product. Advances in antibody engineering and design in the past decade have rekindled interest in this promising area of research, as evidenced by the research presented at the conference.

Horst Lindhofer (TRION) presented data on trifunctional antibodies as next generation immunotherapeutics for cancer. He first reminded the audience that catumaxomab (Removab^®, TRION Pharma/Fresenius Biotech) is a first-in-class molecule in many ways. It is the first approved treatment of malignant ascites, the first approved anti-EpCAM antibody, the first approved bispecific, trifunctional antibody, as well as the first approved therapeutic antibody that is active in the µg-range.1 Triomab^® molecules such as catumaxomab are mouse/rat hybrids composed of mouse IgG2a and rat IgG2b that bind a tumor-associated antigen (TAA) and CD3, as well as Fc receptors. The format has advantages in production due to the species restricted heavy/light chain pairing and the hybrid Fc-pairing that occurs due to homologous sequence. The effector functions also confer advantages; Triomab^® molecules bind to activating FcγRI, RIIa and RIII-positive accessory cells and have minimal or no binding to inhibitory FcγRIIb. Dr. Lindhofer explained that trifunctional antibodies induce a highly potent immune response through multiple mechanisms such as physiological co-stimulation via T-cell activation and recruitment of accessory cells that induce lysis, apoptosis, antibody-dependent cell-mediated cytotoxicity (ADCC) and phagocytosis.2–4 The molecules also have an active immunization effect.5

In addition to catumaxomab, TRION is currently developing three other Triomab^® constructs. Ertumaxomab (rexomun™) targets breast cancer cells with low as well as high expression levels of HER2 as the tumor-associated antigen,6 and is being evaluated in Phase 2 studies of patients with advanced or metastatic breast cancer. FBTA05 (lymphomun™) targets CD20 and is undergoing evaluation in a dose escalating Phase 1/2 study as a treatment of relapsed or refractory disease in patients with either low- or highgrade NHL. TRBS07 (ektomun™) targets GD2/GD3 and is in preclinical development as a potential treatment for melanoma, small cell lung cancer, glioma and neuroblastoma. Dr. Lindhofer summarized preclinical data that indicated that trifunctional antibodies are characterized by simultaneous activation of different types of immune cells, destruction of tumor cells at low ng/mL concentrations, elimination of tumor cells with low target antigen expression, elimination of apoptosis-resistant tumor cells, and induction of long-lasting anti-tumor response.

Dr. Lindhofer then focused on clinical study results for catumaxomab, with an emphasis on the results indicating that the product induces humoral tumor-associated immune responses in patients with malignant ascites.7 He described the study design for the on-going Phase 3 CASIMAS study [NCT00822809] of patients with malignant ascites due to epithelial cancer. Patients were initially tested for EpCAM expression then administered four 3-h infusions of 10, 20, 50 and 150 µg catumaxomab over 11 d. Plasma samples were taken at various intervals in the pretreatment, treatment and follow-up periods, and anti-EpCAM and anti-HER2 immunoglobulin (Ig) responses were quantified by ELISA. A 3–4x increase of the initial anti-EpCAM IgG titer was observed in 12 of 25 patients (48%), and development of HER2-specific IgG antibodies was observed in 14 of 25 patients (56%). In a result comparable to a booster reaction to repeated vaccination, three of four patients who received a second treatment cycle showed a clear anti-HER2 IgG response that was stronger and faster compared with patients who received one treatment cycle (median IgG concentration at day 18 = 58 ng/ mL and 11 ng/mL, respectively). Dr. Lindhofer indicated that the vaccination effect of Triomabs is characterized by anti-tumor responses that are target antigen induced, polyvalent indicating antigen spreading, and humoral and cellular. In concluding, he noted that Triomabs act as vaccine adjuvants by their xenogenic nature, and that the rat/mouse origin of the molecules has several advantages, e.g., the unique isotype combination preferentially binds to human activating Fc receptors on antigen presenting cells; the rat/mouse origin serves as a “danger signal” at the tumor site, which amplifies the anti-tumor response.

Ulrich Brinkmann (Hoffmann-La Roche) discussed his experience with resolving stability issues associated with bispecific antibodies and with digoxigenin-binding (< Dig >) bispecific antibodies for targeted payload delivery. He first noted that production of stable bispecific antibodies that do not aggregate can be a technical problem, and that one solution is the use of intrachain disulfides.8 Dr. Brinkmann presented data showing bispecific disulfide-stabilized Fvs (dsFvs) have low aggregation at concentrations up to 5.0 mg/mL and temperatures up to 40°C. He discussed several formats for bispecific antibodies, including the bivalent Crossmab 1 + 1 format.9 Crossmabs are Y-shaped IgG format molecules. The knobs-into-holes technology is used to enable heterodimerization of the heavy chains; correct association of the light chains with heavy chains is achieved via domain exchanges within the Fab of one half of the bispecific antibody. Antigen-binding affinity is retained, but the two arms are sufficiently different that light-chain mispairing does not occur.

Dr. Brinkman then focused on the tetravalent < Dig >-bispecific format, which has an antigen binding module as well as a Dig binding module for payload delivery.10 The < Dig >-bispecific antibodies can bind digoxigeninylated compounds, including small molecules, peptides, proteins, liposomes, under physiological conditions; antibody-mediated delivery can be achieved via sequential administration of a < Dig >-bispecific antibody and Dig-payload. The functional, stable, recombinant Dig-binding module is a dsFv, i.e., single chain Fv with additional VH44-VL100 disulfide stabilization. Dr. Brinkman presented data showing the < Dig > bispecifics can be expressed and purified via standard technologies, and the binding stoichiometry is 1 antibody: 2 payloads. His data indicated that the < Dig > bispecifics simultaneously bind cell surface targets and payload, and they were shown to accumulate fluorescent and protein payloads on targets in vitro, as well as in vivo.

Insight into bispecific T-cell engaging (BiTE) antibodies in early development was provided by Patrick Baeuerle (Micromet). Three BiTE antibodies are currently in clinical development (anti-CD19 blinatumomab, anti-EpCAM MT110, and anti-CEA MT111) and six are in preclinical development. Micromet's BiTE antibodies selectively bind to a tumor-associated antigen (TAA) and to CD3 on T cells, leading to T-cell activation and destruction of the tumor cell. Dr. Baeuerle discussed key learnings from clinical studies of the company's lead product candidate blinatumomab,11 such as (1) eradication of the entire CD19-expressing B-cell compartment by T cells with a BiTE antibody along with malignant cells is well-tolerated and reversible; (2) the BiTE has high activity in blood at 0.005 mg/m² per day, and in bone marrow at 0.015 mg/m² per day; to date, response rates of 75% in relapsed/refractory acute lymphoblastic leukemia (ALL) and 80% in minimal residual disease positive ALL have been observed; (3) BiTEs can durably eliminate tumor cells in bone marrow below the limit of detection and (4) long lasting responses of more than 3 y in elderly NHL and ALL patients after 4–8 weeks of single-agent treatment have been observed.

Dr. Baeuerle then focused his presentation on the anti-CD33 BiTE MT114 as a potential treatment for acute myeloid leukemia (AML). He indicated that there is high unmet medical need for novel AML treatments, especially those with the potential to extend disease-free survival, effectively treat patients in remission and establish efficient second/third line treatment regimens. Dr. Baeuerle pointed out that the potential hurdles to conventional antibody-based therapies include the aggressive growth of blasts, a wide range of genetic factors contributing to outcome, issues with expression of targets on leukemic/pluripotent normal stem cells, little impact of selected modes of action on leukemic stem cells and severe immune suppression by AML blasts. The factors taken into consideration in the selection of CD33 as a target were then discussed by Dr. Baeuerle. CD33 (SIGLEC-3) is a sialic acid-binding Ig-like lectin with an immunoreceptor tyrosine-based inhibitory motif (ITIM). It is broadly and significantly overexpressed on bone marrow cells from patients with AML, chronic myeloid leukemia (CML) or myelodysplastic syndromes (MDS) compared with those from healthy subjects. CD33 is consistently expressed on human AML cell lines, and circulating soluble CD33 has recently been identified as a marker for progression of AML.

As a BiTE molecule, MT114 engages patients' T cells to kill target cells, which are AML blasts and normal myeloid cells expressing CD33. M114 is a ∼55 kDa, highly stable monomer comprising two single-chain antibodies of human sequence tethered together via a non-immunogenic linker. The molecule is cross-reactive with cynomolgus monkey CD33 and CD3, and can be produced at a gram/liter scale by a CHO cell clone and purified in a scalable 3-step process. Dr. Baeuerle presented data showing that MT114 mediates lysis by unstimulated PBMC of all 10 human AML lines tested, with EC₅₀ values in the 0.4–7 pM range, and supports serial lysis of tumor cells as demonstrated at very low effector-to-target (E:T) cell ratios. Most notably, MT114 induces massive infiltration of CD8 T cells into tumor tissue in vivo in an established HL-60 subcutaneous xenograft model. Dr. Baeuerle also indicated that levels of soluble CD33 in AML patients do not affect MT114 activity, and potent activation of AML patients' T cells and lysis of blasts in autologous ex-vivo settings by MT114 has been observed.

Single light chain (VL) antibodies and MeMo™, a transgenic mouse suitable for discovery and production of human bispecific antibodies, were discussed by John de Kruif (Merus). He first noted that the canonical antibody structure contains two variable domains and six CDRs, but it is now well-known that functional antibodies can use one domain or fewer CDRs, and affinity and specificity can be captured using VH diversity only. Use of the single VL format allows expression of multiple IgG in one cell without heavy and light chain mispairing, and production of bispecific antibodies in the absence of engineering that therefore behave like conventional IgG.

Single VL IgGs can be derived from phage display libraries of human single VL Fabs or MeMo™ mice. Dr. de Kruif described Merus' work using the phage display approach to generate single VL antibody collections against more than ten targets. Semi-synthetic and immunized/natural repertoires were used, and affinities were as low as 4 pM. A total of 50–150 unique clones (based on V-D-J gene recombination) per antigen were isolated; these provided broad coverage of functional epitopes as established by domain mapping. The non-formulated single VL antibodies demonstrate excellent stability exemplified by a low propensity toward aggregation, with a mean aggregation level of only 1.4% as measured by HP-SEC analysis.

As an example demonstrating that single VL mAbs are functionally active, in vivo study results for two anti-respiratory syncytial virus (RSV) single VL antibodies were discussed by Dr. de Kruif. The study was RSV prophylaxis in the cotton rat; intraperitoneal (ip) administration of mAb occurred one day prior to intranasal challenge with RSV A2 and animals were sacrificed at day 4. Mean reduction in the virus titer was similar in the cohorts of animals administered palivizumab, a benchmark humanized IgG1 that targets the F protein of RSV, or two Merus single VL IgG1 that targeted respectively the F or the G antigen of RSV.

Dr. de Kruif then discussed the design of the transgenic MeMo™ mouse, which is a knockout for murine Ig loci while including a single rearranged human VL and a rearranging human VH locus containing human VH, DH and JH segments. Murine CH and control elements are used that were shown to ensure normal B-cell development in central and peripheral lymphoid organs and robust immune responses. He noted that an analysis of randomly picked human VH gene sequences from spleen B cells of MeMo™ mice showed all human DH, JH and VH gene sequences were used, normal heavy chain CDR3 length and distribution, and extensive diversification of CDR3 through deletions and N insertions. Immunization of the mice results in a robust serum antibody response.

A VH mining approach can be used with MeMo™. Using this approach, RNA from lymphoid organs of immunized mice are isolated, VH regions are deep sequenced (the VL sequence is constant and known), CDR3 heat maps are constructed and used to identify immunogen-driven clones. The immunogenspecific VH with the single VL are transfected in a eukaryotic cell line to provide single VL IgG in therapeutic formats for functional screening.

Dr. de Kruif explained that the MeMo™ and single VL concepts work well together for bispecific IgG discovery and production. Two single VL IgG expressed in a single cell yield 50% bispecific IgG; with minimal Fc engineering, >90% bispecific IgG can be obtained. The bispecific antibodies are purified using cation exchange chromatography. The separation is based on isoelectric point differences between the bispecific and parental mAbs. Dr. de Kruif discussed an example where the bispecific was isolated in >99.5% purity at a yield of 70%, and noted that the yield can be improved to >95% by applying semi-continuous ion exchange chromatography.

Gunasekaran (Guna) Kannan (Amgen) presented novel applications of engineered antibody Fc heterodimers and recent studies with designed monomeric form of Fc. Guna explained that he would focus on engineering CH₃-CH₃ domain interaction because Fc crystal structures show extensive protein-protein interaction surfaces between the CH₃-CH₃ domains, and hinge region disulfides are known to form only when the heavy chain is expressed with CH₂ and CH₃ domains,12 and work by Paul Carter and colleagues,13 as well as Jonathan Davis and colleagues14 and other groups, has indicated that CH3 domain engineering is a suitable approach. He discussed a computational analysis scheme that might improve on the results seen using previous approaches. The scheme involves steps such as (1) identifying known Fc crystal structures from the Protein Data Bank (PDB); (2) identifying interface residues through contact and solvent accessibility analyses; (3) examining structural conservation of side chain conformations and (4) identifying buried interface residues and comparing these with structurally conserved positions. Guna then discussed the residues in the CH3 domain interface that are structurally conserved, and specifically the charge complementary pairs Lys409-Asp399′, Lys392-Asp399′, Lys439-Asp356′, and Lys370-Glu357′. He also noted that Lys392, Asp356 and Glu357 are exposed residues. Literature evidence suggests that charged residues occurring even outside of the protein-protein interface region can influence the rate and formation of protein complexes, electrostatic steering effects can be effective in enhancing the rate of association between proteins and engineering protein-protein interaction specificity is possible, which further suggests that charged residues can be used to promote Fc heterodimer (and hinder homodimer) formation. He then discussed a charge pair mutation (CPM) strategy to enhance the formation of Fc heterodimers, e.g., K409D and D399′K mutations. As a proof-of-concept, the K409-D399′ pair was chosen because it is the only charge pair that is structurally conserved and buried. Dr. Gunasekaran stated that the approach does not cause large changes in the amino acid side chain volume and the hydrophobic core of the interface remains intact. He presented data demonstrating that the Fc heterodimer designed using the CPM approach retained FcRn binding and thermal stability and was made in good yield. And the properties of the molecule were comparable to one designed according to a knobs-into-holes strategy. The CPM approach was used to produce a bispecific scFv-Fc that was 98% heterodimer. He presented data showing the bispecific scFv-Fc bound targets on T cells and cancer cells (J45.01 cells and U87 cells transfected with a tumor antigen), mediated killing of tumor cells as measured in a CTL assay using human PBMC and direct T-cell-mediated killing of tumor antigen-bearing cells in vivo as determined using a U87-TARTK xenograft model in NOD SCID mice. PK analysis in mice confirmed that the serum half-life is not affected due to the CPM strategy. He also showed data that the CPM approach could be applied to formation of monovalent IgG15 and possibly to other IgG subclasses.

To conclude, Guna briefly discussed formation of stable monomeric form of Fcs, which have potential applications as fusion modules (e.g., monovalent, mono/bispecific, half-life extension). The design goal is 2-fold: discourage dimer formation and stabilize monomer. To discourage homodimer formation, charge residue polarity could be changed, while structurally conserved hydrophobic residues could be mutated to small polar restudies to stabilize monomer formation. He noted that threonine is preferred in β-strands, which are rich in the CH₃ domain interface. Examination of a variety of Fc-delta hinge -CH₃ mutants lead to identification of a K392D, K409D and Y349T mutant, which formed homogeneous monomeric Fc that retained FcRn binding in vitro. DSC and 37°C storage stability studies suggested that the monomeric Fc is likely to be stable in solution. Fusion proteins comprising the monomeric Fc with Fab or scFv were expressed at 18 µg/mL and 60 µg/mL, respectively. He also noted that PK analysis reveals that a Fab-monomeric Fc fusion had >3–6x more exposure that a Fab-Ig-like dimer, which was the size-matched control, or Fab alone.

Syd Johnson (MacroGenics) discussed maximizing the potential of the dual affinity retargeting (DART) bispecific platform. DARTs utilize a diabody format that includes a disulfide linkage between the C-terminus of each chain to stabilize and constrain the conformation of the molecule. Macrogenics has produced over 75 different DART molecules that have potential applications as treatments for cancer, autoimmune disorders or infectious diseases. Dr. Johnson presented results for an anti-CD16xCD32B DART that redirects the activity of macrophages and natural killer cells toward B cells.16 The DART could be expressed at grams per liter levels; the purified protein displayed very little aggregation and did not lose binding activity when incubated with human serum at 37°C for up to 8 weeks. Potent, dose-dependent cytotoxicity against B-lymphoma cell lines (Daudi, Raji, RPMI-8226), and in vivo protective activity (Raji cell xenograft model) consistent to that of rituximab at the same dose, were observed. Dr. Johnson also noted that DARTs that retarget T cells toward tumor-associated antigens (e.g., anti-CD19xCD3, anti-CD19xTCR DARTs) have been produced and in vitro and in vivo results suggest these molecules may be more potent than BiTE molecules.17

In the final presentation of the meeting, Venkata Doppalapudi (CovX) provided an overview of the chemical generation of bispecific antibodies. The CovX technology that he described involves joining two different peptides using a branched azetidinone linker, and then fusing the resulting bispecific pharmacophore to a human IgG1 scaffold antibody through a heavy chain lysine located in the hydrophobic binding pocket of each Fab.18 The pharmacophores confer functional activity and the antibody scaffold provides long half-life and Ig-like distribution to the fusion protein, termed a CovX-body. Dr. Doppalapudi discussed the generation and properties of CVX-241, a bispecific CovX-body that binds angiopoietin-2 (Ang2) and vascular endothelial growth factor (VEGF). The two targets were selected based on data showing that increased levels of both VEGF and Ang2 are found in many solid tumors and correlate with the lowest survival in cancer patients. VEGF inhibition is clinically validated (e.g., target of bevacizumab); AMG 386, a selective angiopoietin 1/2-neutralizing peptibody, is undergoing evaluation in a Phase 3 study [NCT01204749] in women with recurrent partially platinum sensitive or resistant epithelial ovarian cancer, primary peritoneal cancer or fallopian tube cancer.

To produce CVX-241, the anti-Ang2 and anti-VEGF peptides were synthesized separately, and then each was condensed with a PEG carboxylic acid spacer. The two molecules were then joined through maleimide-thiol ligation, and the resulting molecule coupled with the azetidinone linker. The peptide heterodimerlinker was then coupled site-specifically to the two Fabs of the scaffold antibody at room temperature. Dr. Doppalapudi emphasized that the position of the peptides relative to the linker and antibody needed to be carefully selected to maintain the binding activity of the peptides as well as the stability of the final molecule. The K_d of CVX-241 for Ang2 and VEGF was 5.8 and 4.9 × 10⁻¹⁰, respectively. The half-life was 84–91 h for Ang2 binding and 73–92 h for VEGF binding in both mice (higher range) and cynomolgus monkeys (lower range) after a single iv dose of 10 mg/kg. In a Colo205 xenograft model, CVX-241 showed a greater and prolonged anti-tumor effect compared with monospecific anti-VEGF CovX-body CVX-3051, monospecific anti-Ang2 CovX-body CVX-060, and to the combination of the two monospecific molecules. Enhanced anti-tumor efficacy and delay of tumor growth were also observed when 10 mg/kg CVX-241 was administered weekly with irinotecan in the Colo205 xenograft model. Dr. Doppalapudi concluded by noting that CVX-241 is undergoing evaluation in a Phase 1 study of safety, tolerability, and PK in patients with advanced solid tumors [NCT01004822].