Clinical Pharmacology and Translational Aspects of Bispecific Antibodies
A Trivedi,S Stienen,M Zhu,H Li,T Yuraszeck,J Gibbs,T Heath,R Loberg,S Kasichayanula
DOI: https://doi.org/10.1111/cts.12459
2017-01-01
Clinical and Translational Science
Abstract:Development of bispecific antibodies (BsAbs) as therapeutic agents has recently attracted significant attention, and investments in this modality have been steadily increasing. This review discusses challenges, and suggestions to overcome them, associated with the development of BsAbs, specifically those pertaining to clinical pharmacology, pharmacometrics, and bioanalysis. These challenges and possible solutions are discussed by presenting several case studies of BsAbs that have gained regulatory approval or that are currently in clinical development. BsAbs, also termed "dual-targeting" or "dual-specificity" antibodies, have the ability to bind two different targets on the same or different cell(s); the targets may be cell-surface receptors or soluble ligands, as shown in Figure 1. These dual-nature antibodies have key advantages that can potentially enhance therapeutic efficacy compared with monotherapy or traditional combination therapies by: i) simultaneously blocking two different targets or mediators that have a primary role in the disease pathogenesis; ii) inducing cell signaling pathways (e.g., proliferation or inflammation); iii) retargeting to mediate antibody-dependent cell-mediated cytotoxicity (ADCC); iv) avoiding the development of resistance and increasing antiproliferative effects, specifically in oncology; and v) temporarily engaging a patient's own cytotoxic T cells to target cancer cells, thus activating cytotoxic T cells to cause tumor lysis (e.g., bispecific T-cell engagers (BiTE)). Various designs for BsAb molecules (a) Dimers inhibition: BsAbs can bind to two receptors/targets (HER2/HER3, HER2/HER4) on the same cell (e.g., MM-111); (b) Dual inhibition: BsAbs can inhibit two different cytokines simultaneously, for example, COVA322 that inhibits TNF-α and IL17A; (c) Triomabs: The antigen binding site binds to target cell receptors (EpCAM, HER2, or CD20) and the T-cell receptors (CD3). The heavy chain site binds to NK cells or dendritic cells or macrophages/phagosome (e.g., catumaxomab, ertumaxomab, FBTA05); (d) Two-ligand inactivation: two arms bind to different ligands on different cells belonging to the same population, such as DLL4 x VEGF, TNF-α x IL17A, IL4 x IL13 (e.g., OMP-305B83, COVA322, SAR156597); (e) Transmembrane/transcytosis: The BsAbs are designed specifically to cross the barriers/membrane via receptor transport (transferrin receptor) and bind to enzymes/receptors (BACE1) on the other side; (f) BiTE antibody construct: These are designed to bridge T cells and target cells by binding to CD3/CD28 or CD19/CD20/CD22/CEA/EpCAM, respectively (e.g., blinatumomab, MEDI-565, MT110). The examples mentioned above can be found in Table 2 for further information. BACE1, β-secretase 1; BiTE, bispecific T-cell engagers; BsAbs, bispecific antibodies; DDL4, delta-like ligand 4; EpCAM, epithelial cell adhesion molecule; HER, human epidermal growth factor receptor; IL, interleukin; NK, Natural Killer; TNF-α, tumor necrosis factor-alpha; VEGF, vascular endothelial growth factor. Traditional combination therapies using monoclonal antibodies (mAbs) can also modulate multiple therapeutic targets. However, the development of mAbs presents challenges not encountered with BsAbs. For example, regulatory agencies have established stringent criteria for the codevelopment of new drugs that are intended for use as combination therapies. The sponsor has to demonstrate i) the rationale for use of the combination therapy rather than individual treatments; ii) a strong justification for why the individual drugs cannot be studied and developed independently; iii) that the nonclinical and clinical studies provide adequate evidence showing that the combination therapy provides significant therapeutic gain; and iv) a reasonable toxicity profile and more durable response than the monotherapy and existing standard of care.1 These guidelines can potentially make the drug development process for combination therapy lengthy and expensive. Conversely, BsAbs are able to address the biology associated with two different targets simultaneously via a similar regulatory pathway as that required for with a single-target mAb. BsAbs may therefore offer the opportunity to benefit patients more quickly, and to access less costly development routes than can be afforded via classic combination therapies. BsAbs also offer the opportunity to modulate unexplored biology in novel ways that may not be possible with single-target mAbs. Avidity is defined as the measure of the overall strength of binding of an antigen with multiple antigenic determinants to multivalent antibodies. According to the "avidity hypothesis," BsAbs may surpass combination therapy in terms of both biology and mechanistic behavior as a result of this theoretical concept. The theory states that avidity increases when two receptors are bound to a target cell, leading to efficacy greater than which could be expected from the additive combination each single mAb. A specific example is the development of JNJ-61186372 (BsAb targeting epidermal growth factor receptor (EGFR) and c-Met), which showed that the BsAb was more potent than the combination of single receptor-binding antibodies.2 Furthermore, BsAbs are less likely than combination treatment to undergo off-target binding in the presence of a surplus of decoy cells.3 BsAbs therefore have the theoretic potential to improve therapeutic window (safety and efficacy), selectivity, and regulatory efficiency as compared with a true combination therapy approach. As a result of the aforementioned advantages of BsAbs (Table 1), BsAbs are one of the fastest growing classes of investigational drugs. In addition to the approved BsAbs, blinatumomab (BLINCYTO, Amgen, Thousand Oaks, CA) and catumaxomab (Removab, Fresenius Biotech, Homburg, Germany, initially marketed by Fresenius Biotech) for cancer immunotherapy, there are more than 50 additional BsAbs in clinical development4-10 (Table 2), with the potential for sales of up to $4.4 billion by 2023.11 Of note, almost all BsAbs that are currently in development target indications in oncology, with the following exceptions: COVA322 for plaque psoriasis, BsAbs targeting transferrin receptor (TfR) and β-secretase 1 (BACE1) for central nervous system disorders, ABT981 for osteoarthritis, ALX-0761 for psoriasis, AMG 570 for systemic lupus erythematosus, and JNJ-61178104 and MDG010 for autoimmune diseases.12 In this review we aimed at increasing awareness of the multiple facets of translational and clinical development of BsAbs. We are presenting clinical pharmacology considerations (Table 3) and modeling simulation strategies with select examples as well as bioanalytical challenges and strategies, opportunities, and approaches supported by a variety of case studies. However, published information on some of the aspects discussed here is limited, thus restricting a broader selection of case studies and underscoring the need for a more extensive application of modeling simulation approaches to support efficient drug development. The functional domain architecture of mAbs have been extensively exploited to create a number of different BsAbs formats. Spiess et al. originally classified them into five distinct structural groups:7 i) bispecific IgG,13, 14 ii) IgG appended with an additional antigen-binding moiety,15, 16 iii) BsAbs fragments,17, 18 iv) bispecific fusion proteins,19 and v) BsAb conjugates.20 The new formats are categorized based on the Fc-mediated effector functions and are classified as immunoglobulin G (IgG)–like molecules and non-IgG–like molecules, as shown in Table 4. The IgG formats are larger and undergo FcRn recycling, which results in a longer serum half-life, whereas non-IgG formats have a smaller size, which enables increased tissue penetration. These unique formats vary in antigen-binding valency properties, thereby offering a potential opportunity to optimize valency of each component antibody based on the biology of the mechanism perturbed by the therapeutic.7 Because these formats are comprised of individual functional domains, the activities of the domains can be monitored via quantification of the single domain. Elucidating the exposure of the active therapeutic for informing model-based drug development in the context of a BsAb is complex. The molecular variant chosen to illicit target interrogation can affect the likelihood of biotransformation of the molecule, a critical factor in translating active exposure to a pharmacological response. The physiochemical BsAb features must be further coupled with structural variants that may exist in free, partially bound, and bound forms21, 22 resulting from binding to a soluble target, for example. BsAbs are furthermore confounded by valency properties of each antibody component binding to its respective antigen. Partially bound forms of the molecule may alter the stoichiometry associated with the dual-target binding of the biotherapeutic. Innovative bioanalytical approaches are required to fully understand the active exposure of a BsAb, which is dependent on both its unique physical/chemical properties and the dual-targeting strategy represented by the molecule. These distinct structural groups serve to illustrate the structural complexity and diversity of BsAbs, which raise unique challenges related to the bioanalytical strategy. Several global regulatory agencies have published guidelines for the development of both small molecules and biologics. However, development strategies may differ between BsAbs, mAbs, and traditional combination therapies. As with any biologic molecule, BsAbs have the potential to elicit an immune reaction. Immunogenicity is typically assessed by detection of antidrug antibodies (ADA). Formation of ADA is not always associated with mAbs. There are cases of mAbs without ADA formation (e.g., rilotumumab, which has an ADA incidence of 0%). Therefore, the incidence of ADA associated with BsAbs cannot be lower than zero. Among BsAbs, catumaxomab has an ADA incidence of 0%, while blinatumomab has a low ADA incidence of ∼1%. Our experience in the field suggests that key factors of ADA formation related to BsAbs are similar to those associated with other mAbs. This may include, but is not limited to, the structure of BsAbs (i.e., whether it can be recognized by the immune system as "foreign"), the presence of foreign sequences (e.g., asymmetric rat-mouse hybrid BsAb), route of administration (higher incidence of ADA with subcutaneous than with intravenous (IV) administration), dose, and characteristics of the patient's immune system. It is well known that the formation of ADA may alter pharmacokinetics (PK), leading to subsequent changes in the pharmacodynamic (PD) properties of a BsAb. However, the drug-binding characteristics of an ADA may lead to differentiating impact on the individual target binding.23 Biomarker development programs for BsAbs often face several unique challenges. Standard approaches to evaluate for biological/biochemical impact via target engagement and PD assays are routinely focused on specific analytes to ensure that an appropriate dose and schedule are selected to advance farther along into clinical testing. BsAbs often present the challenge of dissecting the biologic complexity and PD coverage associated with two targets simultaneously. A BsAb targeting two antigens or ligands may create the need to define two target coverage thresholds (unique to each marker), all the while dealing with the reality that the BsAb provides a fixed ratio of exposure to either target at a given exposure. In the case of a BiTE antibody construct, it is critical in early clinical trials to demonstrate engagement of T cells (via the CD3 binder) as well as the tumor-specific target (CD19 in the case of blinatumomab). Such T-cell activation has been explored with blinatumomab by evaluating CD69 and CD25 upregulation posttreatment in acute lymphocytic leukemia patients.24 Similarly, patient stratification hypotheses for sensitivity or resistance are generally formulated in the preclinical development and then tested throughout development with the ultimate goal of identifying a predictive marker for response to a single biologic process. Patient stratification marker(s) for BsAbs may need to be tailored to each binding arm of the BsAb, or the proposed combinatorial biology elicited via simultaneous dual target PD. Molecular imaging can play an important role in identifying and improving the success rate of promising new drug candidates, including BsAbs. Goldenberg et al. described various examples of the utility of BsAbs as tools for imaging in preclinical and clinical settings. Imaging studies have also been demonstrated to be of value in identifying the immunogenicity effect of BsAbs.25 These tools can be utilized in oncology for diagnosis and detection of small tumors or lesions using differential approaches and rationales, such as Dock and Lock, click chemistry,26 or heterodimer conjugated nanomaterials.27 Drug–drug interactions (DDIs) for BsAbs are less well understood than for monotherapies or traditional combination of mAbs. If the investigational therapeutic protein (TP) is a cytokine or modulates cytokine biology, studies should be conducted to determine its effects on CYP enzymes or transporters. Lee et al. conducted a survey aimed at systematically reviewing US Food and Drug Administration (FDA)–approved therapeutic proteins and the implications of therapeutic protein–drug interactions. The survey encompassed 68 new therapeutic proteins that had been approved by the FDA by the end of 2008.28 The results showed that cytokine release followed by administration of an mAb with cytokine-modulating properties was a major reason for DDI after mAb administration. Cytokines are involved in the pathophysiology of multiple human diseases, and their levels are increased during infection and inflammation. Therefore, biologics that modulate cytokine activities can indirectly influence the expression of specific CYP enzymes and drug transporters by affecting cytokine concentrations. This concept is described extensively elsewhere.29, 30 The magnitude of cytokine-induced effects on CYP450 depends on the level of cytokine elevation, the type of cytokine (e.g., IL-6 being important), and the duration of cytokine elevation. The potential for DDI with transient cytokine elevation may be very different from that with chronic cytokine elevation.31 Kenny et al. published an article to facilitate better understanding of the current science, investigative approaches, and knowledge gaps in this field.32 Key issues discussed included translating in vitro to in vivo knowledge in DDI along with questions of whether in vitro data could add value in defining the need for a clinical DDI study, whether the acute phase response protein C-reactive protein (CRP) could be used as a potential biomarker for CYP modulation in inflammatory disease, whether TP-DDI could be quantitatively predicted from preclinical data, and how a clinical DDI study can be designed appropriately. Assessment of DDI generated with a BsAbs should be approached the same way as with any large molecule. If clinical studies are restricted to patients instead of healthy volunteers, population PK modeling provides a feasible approach for TP-DDI assessment. Population PK modeling allows less intensive sampling, incorporation of TP-DDI assessment in larger phase II and III trials involving relevant patient populations, and integration of data generated from multiple studies during different development phases. Population PK modeling also supports evaluation of the effects of combined "perpetrators" on a TP and, potentially, the effect of a TP on comedications when the analysis is prespecified and concentrations of the comedications are evaluated. Trends identified in an exploratory population PK analysis can be used to guide decisions for the need of additional DDI studies. Regulatory agencies have included recommendations in their guidance on TP-DDI assessment during drug development. The European Medicines Agency (EMA) guideline,33 published in July 2007, entitled "Guideline on the Clinical Investigation of the Pharmacokinetics of Therapeutic Proteins," describes concerns about immunomodulators such as cytokines, which have shown a potential for inhibition or induction of CYP enzymes, thereby altering the metabolism of coadministered small-molecules that are substrates of these enzymes. The guideline suggests that the in vitro and/or in vivo studies should be considered on a case-by-case basis. The 2012 FDA draft guidance on DDI similarly expands the US agency's current recommendation on TP-DDI assessment.34 Model-based approaches increasingly support decisions spanning the entire drug development process, from preclinical development through postmarketing, as shown in Figure 2. The application of such approaches to the development of BsAbs follows this standard paradigm and is applied at various stages during the development of biologic therapeutics. Predicting the PK of BsAbs generally follows the same paradigm as mAbs; i.e., allometrically scaling preclinical PK parameters to predict human PK. Although this tends to work best for antibodies with linear PK, models incorporating nonlinear clearance mechanisms have been developed. For example, preclinical PK of the BsAb MEHD7945 in cynomolgus monkeys were fitted to a standard two-compartment PK model with nonlinear and linear clearance components, and the resulting PK parameters were translated using a common method for scaling.35 More complicated target-mediated drug disposition (TMDD) models that mechanistically describe simultaneous binding to two targets have also been proposed, although it is not straightforward to translate such models from preclinical species to humans because of a lack of critical information, such as the relative density of the target in preclinical species compared with humans.36 While this is true of TMDD models in general, these challenges are exacerbated for BsAb because they bind two targets simultaneously. Still, successful development of a TMDD model that describes the PK of a BsAb along with an understanding of species differences impacting the model may help guide first-in-human (FIH) dose selection, because such mechanistic models predict the degree of target engagement for each BsAb arm. Projection of target engagement is a key application of modeling and simulation approaches, particularly for development programs that lack biomarker data. Clinical decisions regarding the FIH starting dose are based on results from toxicology studies and on expected pharmacology, a comprehensive approach recommended by the FDA.37 However, such approaches have had mixed success for traditional mAb and their application has been limited. Advances in this area represent a significant opportunity for modeling and simulation to contribute to the transition from preclinical space to the clinic.38 Mechanistic modeling and simulation approaches tailored to the interrogation of BsAb pharmacology have yielded critical insights into the mechanism of action of BsAbs and the conditions under which they offer advantages. A physiologically based pharmacokinetic (PBPK) model developed by Friedrich et al. provided a rationale for the lack of clinical success of molecules that retarget effector cells in solid tumor indications, and a platform for designing molecules with a greater potential for success. Their PBPK model suggested that effector cells dictate the distribution of the BsAb and that modifications to the tumor microenvironment were more promising approaches to boosting efficacy in solid tumor indications than optimizing the binding parameters of the BsAb.39 For a BsAb targeting both programmed death-1 (PD-1) and T-cell immunoglobulin and mucin domain (TIM)-3 to manipulate a patient's immune system into attacking a foreign entity (such as a cancer), a systems pharmacology approach was used to assess the combinations potential to provide enhanced efficacy over monotherapy or traditional combinations. The investigators used a mechanistic model and, based on its predictions suggesting that a BsAb offered no advantages over a fixed-dose combination, ultimately decided not to develop the BsAb. The model predictions also allowed for accelerated project timelines by informing the selection of individual target candidates with optimal pharmacological properties.40 The development of MM-141, a tetravalent BsAb targeting the IGF-IR and ErbB3 pathways (for the treatment of cancer), was also informed by mechanistic modeling.41 A mechanistic model that integrates information across pathways can provide insight into both the selection of the best targets for overcoming resistance mechanisms and the means by which those targets should be modulated. The simulations for MM-141 showed that a BsAb should be used to modulate the targets, predicting a superior outcome with simultaneous target binding compared with a combination of individual antibodies, for any ratio of IGF-IR and ErbB3 receptors.41 Indeed, this prediction was validated by preclinical studies comparing the antitumor potency of MEHD7945A, a BsAb also targeting IGF-IR and ErbB3, with a combination of cetuximab and an anti-ErbB3 antibody.42 Models of a similar nature were also developed for rheumatoid arthritis to elucidate the effects of an anti-TNFα/anti-Ang2 BsAb43 and for relapsed/refractory acute lymphoblastic leukemia (ALL) to provide insight into the mechanisms governing response and nonresponse after administration of blinatumomab.44 Mechanistic modeling approaches have also been used to determine the optimal characteristics of a BsAb. For example, a mathematical model addressing PK, PD, and safety outcomes was developed for a BsAb that targets the TfR and BACE1. This model was used to support preclinical candidate selection. While it may seem intuitive that increasing affinity of the TfR arm would promote delivery of the molecule to the target site (maximizing the biologic effect), the modeling showed that there is an optimal TfR affinity range for maximum reduction of Aβ peptides. Degradation of the molecule during transcytosis increased with increasing affinity, resulting in lower average brain exposures over time and reduced pharmacologic effect. Reducing affinity therefore undermined TfR-mediated transport through the blood–brain barrier. Similarly, the model suggested the relationship between TfR affinity and reticulocyte depletion was biphasic in nature. This model allowed for a rational approach to antibody engineering, preclinical study design, and candidate selection, resulting in BsAbs with optimal safety and PK/PD profiles.45 These examples show how critical model-based approaches allow for an improved understanding of the mechanism of action of BsAbs (compared with corresponding combination therapies), the conditions under which they offer advantages, and the optimal pharmacologic properties required of such agents to maximize the therapeutic effectiveness. A combination of traditional PK/PD modeling approaches and newly emerging quantitative systems pharmacology provide the best path forward for maximizing the probability of success and for accelerating the development of these molecules. Informing meaningful PK and PD assessments of both efficacy and safety in the development of these molecules demands an appropriate bioanalytical strategy with careful consideration applied to method suitability for measuring functionally relevant forms of the BsAbs. The majority of the bioanalytical methods for BsAbs are based on the principles of ligand binding assays (LBA). The advantages and principles of LBA for measuring biotherapeutics are well described.46 The choice of an appropriate LBA includes the selection of assay platform, format, and critical reagents. It also requires appropriate assessment of other contributing factors that may cause bioanalytical error and mislead the evaluation (risk assessment). Although LBA constitutes the most commonly used approach, there are other assay platforms, such as flow cytometry and mass spectrometry, which are also well suited for addressing bioanalysis questions for BsAbs. BsAbs can bind to various circulating partners, such as free target proteins, ADAs, and other endogenous serum components. In addition, BsAbs may lose their binding ability as a result of biotransformation. Consequently, BsAbs can exist in a pharmacologically active form, characterized by dual-target antigen binding sites, and in an inactive form, which has partial or no target binding sites. Whether to measure the active concentration or the total concentration (active plus inactive/partially active forms) for PK/PD assessment continues to be debated as part of the bioanalytical strategy.14, 47 The recent white paper by the AAPS Ligand-Binding Assay Bioanalytical Focus Group discussed the challenges and issues of measuring free (active), total drugs and target proteins and how these data should be used to support drug discovery and development.48 Typically, when considering which form of the therapeutic protein should be measured to achieve the intended purpose of the study, a "fit-for-purpose" approach is adopted. A working bioanalytical strategy and the selection of appropriate assay technologies to measure the intended forms is presented in Figure 3 as well as in two examples discussed later. COVA322 ((tumor necrosis factor) TNF x (interleukin) IL-17A), also known as a FynomAb, is a bispecific fusion protein consisting of an antibody and a Fynomer.19 Fynomers are small binding proteins derived from the human Fyn SH3 domain. Fynomers can be engineered to bind to target molecules with the same affinity and specificity as antibodies. It is critical to maintain the activity of COVA322 and monitor that the Fynomer is not cleaved in vivo by undergoing potential biotransformation. An enzyme-linked immunosorbent assay (ELISA) format utilizing the antigen specificity of the BsAbs is normally used to directly measure the concentration of the bioactive BsAbs. Two ELISA assays were developed: bifunctional ELISA, detecting intact and bioactive COVA322 using dual antigen, and monofunctional ELISA, detecting the anti-TNF binding antibody portion of COVA322. The assays demonstrated that the plasma concentrations obtained from PK samples by both assays were comparable, indicating that COVA322 stayed fully functional (bioactive) and that there were no indications of in vivo biotransformation. However, this may not be representative of the results observed for target binding in the bispecific formats. For instance, glucagon-like peptide-1 (GLP-1) is a 37-amino acid peptide for the treatment of type II diabetes via GLP-1 receptor binding. A CovX-Body was generated by conjugating the GLP-1 peptide to the N-terminus of a carrier mAb. After in vivo administration of the GLP-1 CovX-Body to mice,49 quantification of the therapeutic was achieved by two different ELISA assays that each used an anti-idiotypic antibody capture of the mAb portion of the construct. Two different detection antibodies were used: antihuman IgG, to measure the mAb only (total assay), and anti-GLP-1 antibody (N-terminus specific), to measure the intact GLP-1 CovX-Body. The intact assay produced drug concentrations that were significantly lower than the total assay throughout the PK time course. Unlike the FynomAb, these results indicate that, although the concentration of the mAb portion of the construct was sustained, the N-terminal region of the GLP-1 moiety was quickly degraded, demonstrating a high degree of in vivo biotransformation (e.g., cleavage of GLP-1 moiety from the antibody). Because BsAbs may present as a mixture of biologically active and inactive forms, it is important to identify the BsAb form that is most pharmacologically relevant to PK/PD assessment and to develop a validated assay that measures the appropriate form accordingly. The following case studies illustrate specific examples for each of the issues that are pertinent to the development of BsAbs. The epithelial cell adhesion molecule (EpCAM) represents a potentially attractive antigen for antitumor therapies, because the transmembrane glycoprotein is highly expressed in a broad range of solid tumor indications and correlates with a poor patient prognosis in most of them.50 Catumaxomab is a BsAb targeting EpCAM and CD3 that was approved in the EU in 2009 for the treatment of malignant ascites. Exudative ascites is fluid accumulation in the peritoneal cavity (abdominal space) that can be caused by the presence of solid tumors. Treatment with catumaxomab along with paracentesis (manual drainage) can significantly delay the need for repeated manual drainage thereafter in patients with malignant ascites caused by ovarian or nonovarian cancer.51 Catumaxomab has been generated by fusing two different hybridoma cells, a technique called "quadroma technology." The hybrid Fc region exerts binding capacity to human FcγR I, FcγR IIa, and FcγR III enabling activation of NK cells, dendritic cells, monocytes, and macrophages.52 In addition, the targeting of EpCAM on tumor cells and CD3 on T cells allows T-cell–mediated redirected lysis of EpCAM-positive target cells, which results in a more than 1,000-fold higher potency compared with monospecific anti-EpCAM antibodies. Catumaxomab is hypothesized to exert its effect via a combination of T-cell–mediated tumor cell killing, ADCC, and phagocytosis via FcγR-activated accessory immune cells. Delivery of catumaxomab in the clinical setting is limited to localized intraperitoneal (IP) administration. In an FIH study where catumaxomab was administered IV, fatal acute liver failure and cytokine release-associated systemic toxicity were observed even at low doses.53 In a pil