Bispecific antibodies targeting dual tumor-associated antigens in cancer therapy

Rational target selection basically determines the MOA of BsAb and is the most important step for success. The preferred BsAb should enable novel biological function and therapeutic MOA which cannot be achieved using mAbs alone or in combination. Basic science supported a key role of c-MET in NSCLC patients developing resistance to EGFR TKIs, supporting design of JNJ-61186372 (EGFR × c-MET BsAb) and patient selection criteria leading to demonstrated anti-tumor activity in NSCLC patients with resistance to EGFR TKIs (Park et al. 2020; Yun et al. 2020). Interestingly, duligotuzumab (MEHD7945A), a BsAb targeting EGFR and HER3, showed no clinical benefit in comparison to cetuximab (anti-EGFR mAb) in phase 2 trials in patients with metastatic colorectal cancer or head and neck squamous cell carcinoma. Expression of HER3 determined by RNA or protein in tumor biopsies did not correlate with the response rate to duligotuzumab. Therefore, the researchers concluded that HER3 has a minor role in EGFR inhibitor naïve mCRC patients (Fayette et al. 2016; Hill et al. 2018). However, others believe that the disappointing results of the study were mainly due to improper selection of patients that were not resistant to prior cetuximab exposure (Saba 2017). Similarly, a phase III study (NCT02134015) of patritumab (HER3 inhibitor) in combination with erlotinib (EGFR inhibitor) for the treatment of NSCLC patients had failed before duligotuzumab (Liu et al. 2019; Yonesaka et al. 2017). Thus, the rationality of selecting EGFR and HER3 as targets for BsAb development requires further investigations. So far, BsAbs targeting dual TAAs have only involved a limited number of targets, with a main focus on ErbB family proteins. It will be interesting to assess BsAbs targeting novel target combinations developed for unmet clinical need.

The affinity and biological activity of BsAb to each of the two antigens could have a critical impact on the final clinical outcome. Before JNJ-61186372, a BsAb against EGFR and c-MET (LY3164530) developed by Eli Lilly, did not enter phase II study due to toxicity and lack of data supporting a predictive biomarker. LY3164530 consisted of an IgG4 antibody targeting c-MET (emibetuzumab, LY2875358) and a single-chain variable fragment targeting EGFR (cetuximab) fused to the N-terminus of each heavy chain. By making use of these two existing antibodies (cetuximab and emibetuzumab), the affinity and activity for each individual arm in LY3164530 were fixed and the relative inhibition of EGFR versus c-MET and affinity to each individual antigen could not be adjusted to improve functionality. Significant toxicities of LY3164530 were recorded and found to be associated with EGFR inhibition but not c-MET inhibition, indicating that engineering the functionality of each arm might have improved its overall toxicity profile (Patnaik et al. 2018). In contrast, JNJ-61186372 was selected from a panel of EGFR × c-MET BsAbs based on functional activity and, similarly, zenocutuzumab came from an unbiased functional screening of aa panel of 545 BsAbs (Geuijen et al. 2018; Grugan et al. 2017). Moreover, a BsAb can sometimes exert a completely opposite activity compared to its two parental mAbs due to its format of conformation. For instance, a dual-variable-domain immunoglobulin (DVD-Ig) BsAb, generated by combining two well-validated antagonist anti-HER2 antibodies trastuzumab and pertuzumab, was shown to be a functional agonist of HER2 (Gu et al. 2014). Therefore, activity of a BsAb should not be assumed based on its parental mAbs, instead both affinity and biological activity should be investigated in an unbiased fashion following construction of the BsAb.

The format of BsAb greatly influences its final physicochemical properties and biological functions. Over 100 different BsAb formats have been invented to solve many scientific or technical issues and their diversity enabled researchers to use them for various applications. A BsAb format suitable for all applications does not exist—the best format is the one that works well for desired application specifically (Brinkmann and Kontermann 2017). BsAbs proper designed with a well-chosen backbone can demonstrate enhanced anti-tumor efficacy and/or reduced side effects. Currently, the human IgG1 backbone is commonly used for dual TAAs targeting BsAbs mainly due to its well-known capacity to confer high exposure and long terminal half-life as well as inducing strong secondary immune functions. Many studies have demonstrated that small differences in the amino acid sequence of the CH2 and CH3 domain as well as the glycosylation profile of the Fc domain highly impact antibody thermal stability, pharmacokinetic properties and FcγR-mediated effector functions (Haraya et al. 2019; Kapelski et al. 2019; Regula et al. 2016; Roux et al. 1997; Zheng et al. 2011). The human FcγRIII, expressed on macrophages, monocytes, neutrophils, mast cells, and NK cells, binds antibodies with low glycosylation more tightly, thus inducing more potent ADCC effects (Satoh et al. 2006). For instance, ADCC of MCLA-128 was enhanced by low fucose glycoengineering using the GlymaxX® technology (ProBiogen) while retaining its binding to FcRn(de Vries Schultink et al. 2018); JNJ-61186372 was produced by a CHO cell line defective for protein fucosylation to enhance ADCC (Moores et al. 2016). However, in addition to differences in Fc region, variation in the variable region presentation and flexibility of the hinge region affect the functional activity of the IgG class. As reported by Kapelskia et al. the hinge region of human IgG subclasses showed different flexibility (IgG1 > IgG4 > IgG2, IgG1 being the most flexible) which significantly influenced the T cell redirection capacity of BsAb (Kapelski et al. 2019). Furthermore, in another example, eight anti-HER2 biparatopic BsAbs were generated from the same parental mAbs by DVD-Ig platform with different variable domain orientations or linker lengths. Interestingly, four BsAbs with same variable domain orientation showed strong agonistic activity while another four BsAbs with opposite orientation were antagonists. Further experiments demonstrated that the BsAb with a particular variable domain orientation could specifically prevent the heterodimer formation of EGFR/HER2 and HER2/HER3, thus forming more HER2 homodimers which lead to the activation of HER2 signal pathway (Gu et al. 2014). Unlike the factors influencing the potency of T cell engager antibodies which are well studied and reviewed (Ellerman 2019), the factors such as IgG subclass, variable domain orientation and length of hinge influencing BsAbs targeting dual TAAs are still largely unknown due to the completely different epitope topology, target geometry and distribution.

Compared to conventional monovalent BsAbs, more and more BsAb formats designed with multi-valence for each target have appeared and showed distinct advantages in particular cases. Cibisatamab (RG7802), a 2:1 CEA × CD3 BsAb, was optimized to have two CEA binding arms with low affinity individually but high avidity when combined, to increase the specificity to CEA^high tumor cells but spare CEA^low healthy cells. This setup facilitated Cibisatamab to bind to cells with > 10,000 CEA-binding sites/cell, which are most likely tumor cells (Bacac et al. 2016). For dual TAAs targeting BsAbs, the valence for each targeted TAAs should be considered individually based on specific properties of the targeted product profile. For instance, in the case of RO7386, a 2:2 BsAb targeting FAP and DR5 using high-affinity bivalent FAP arms ensured tumor-selective targeting, whereas bivalent low-affinity DR5 arms facilitated DR5 hyperclustering and killing of tumor cells (Brunker et al. 2016). Interestingly, for two BsAbs targeting EGFR and c-MET JNJ-61186372 used a 1:1 format while EMB01 used a 2:2 format. Early evidence supporting 1:1 design was that bivalent binding of c-MET invariably induced activation rather than inhibition due to dimerization (Wang et al. 2016). However, in pre-clinical studies, with bivalent binding to c-MET, EMB01 showed no c-MET activation in the absence of HGF. Furthermore, EMB01 achieved significant and sustainable tumor regression in the NCI-H1975-HGF CDX model, which was claimed to be more striking than the one achieved by JNJ-61186372 in a similar model. Such differences may be due to clustering induced by tetravalent antibody binding, which was demonstrated to enhance internalization and degradation for many receptors, including EGFR (Gong et al. 2017).

Besides IgG formats, the IgM format is also used for the development of BsAbs which by design provides more antigen binding sites than IgG format (Kaveri et al. 2012). For instance, IgM-2323 is a CD20 × CD3 bispecific IgM antibody developed by IGM Biosciences currently under clinical evaluation in Phase I for the treatment of patients with B cell Non-Hodgkin’s lymphoma (NHL) and other B cell malignancies (NCT04082936). In contrast to BsAb in other formats, IGM-2323 has 10 binding units to CD20 and one binding unit to CD3. Due to its 10 binding units for CD20, IGM-2323 is speculated to display very high avidity for CD20 expressing cancer cells including those with low CD20 expression that would escape from conventional anti-CD20 therapy (Keyt et al. 2020).