*2.1. Phase-Appropriate Approach*

A phased approach to the development and implementation of bioassays for biotherapeutics is widely accepted by industry and regulatory agencies, and the similar principles apply to bispecific therapeutics. It is often advantageous and preferred to start with a binding method for the early phases of product development. Most commonly implemented binding assays include enzyme-linked immunosorbent assays (ELISAs) or surface-plasmon resonance (SPR) technologies. More complex and MoA-reflective cell-based bioassays are developed by later phases, and they are validated before marketing application submission. However, it is recommended that a relevant MoA-based bioassay is developed earlier, not only to gain a greater process and product understanding but also to gain a better understanding of the method's performance prior to pivotal clinical trials. Cell-based bioassays should be qualified and monitored over the span of the clinical development to have a true understanding of the critical steps and components of the assay in most cases. The selection of the bioassay should be driven by the product's therapeutic MoA. In cases where the MoA is simply binding to a target, a surrogate method, such as a protein binding or competitive binding assay, may be sufficient to determine potency. Developing robust and quality-control (QC)-suitable cell-based bioassays is more challenging than developing non-cell based binding assays [71,72]. There are case studies of implementing surrogate, non-cell-based bioassays in the commercial control system if the surrogate assay

has demonstrated a good correlation in a bridging study using degraded product and other samples with the MoA-reflective cell-based assay.

#### *2.2. Mechanism of Action*

Design strategies for bioassays are driven by the drug's intended physiological MoA. Unlike other analytical techniques, bioassays are almost always unique for each therapeutic. A well-designed bioassay will accurately capture the biological activity of a drug candidate. As shown in Figure 2, common MoAs of bispecific therapeutics include direct binding to soluble targets (e.g., ligands, cytokines and enzymes), or to cell-surface receptors in either an inhibitory or agonistic manner.

Each MoA will require a different approach when considering the bioassay design. In the case of BsAbs and related recombinant proteins, secondary, tertiary, or synergistic MoAs may be discovered during development. This biological complexity further contributes to the challenge of developing MoA-reflective assays to capture the candidate molecule's putative therapeutic biological activity [50,73–76]. In some cases where multiple MoAs exist in a single molecule, a combination assay that measures all MoAs in a single assay may be suitable for product release and stability testing, with secondary characterization assays developed to measure the individual activities of each MoA if applicable. Otherwise, multiple bioassays would be necessary to fully characterize the molecule's activity. It is not required to have all bioassays for release, instead only the assay with the most MoArelevant and stability indicating bioassay can be selected as the release potency assay while the other assays are used for characterization.

#### *2.3. Overall BsAb Characterization Strategy*

Efficacy and safety assessments of BsAbs rely on the successful development of a pharmacologically and clinically relevant bioanalytical strategy that most importantly can reflect the biological activities of these dual-targeting antibodies and can differentiate higher order structure, potency, and efficacy.

It is most vital to develop characterization and bioanalytical approaches to study important quality attributes [77] including overall stability, fragmentation/aggregation/ immunogenicity, antigen specificity, affinity, on and off rates, avidity (for molecules with two targets on the same cell), and MoA/biological activity.

While BsAbs require bioassays to measure two binding events, the choice of the appropriate bioassay will also depend on the assay format, assay platform, critical reagents, and, importantly, the BsAb target profile. Following the successful development of the pharmacologically relevant BsAb format, the analytical strategy is outlined to first characterize the independent or simultaneous binding affinities and the preferential binding of BsAb to their dual-antigen targets. Widely used bi-functional quantitative assay formats to enable target-specific capture and detection of binding properties include flow cytometry and ligand-binding immunoassay setups. A range of other assay platforms (ELISA, SPR, ADCC, competitive flow cytometry, etc.), whose selection relies on BsAb format, MoA, and target profile, are used to address bioanalytical questions for BsAbs. These assays are listed in Table 1 and further discussed in Section 3.

Meaningful bioanalytical approaches are also needed for immunogenicity and PK/PD assessments to determine the safety and efficacy of BsAbs [78,79]. Immunogenicity is defined as the unwanted immune response of the host against the therapeutic BsAb. In addition to altering the PK of a target through changing its clearance, immunogenicity is responsible for infusion-related reactions and in some cases, reduced treatment efficacy [80]. Immunogenicity is clinically assessed by the detection of anti-drug antibodies, consisting of IgM, IgG, IgE, and/or IgA isotypes [81]. The bioassays employed to assess immunogenicity include binding immunoassays such as ELISA to detect all isotypes capable of binding the therapeutic BsAbs, and neutralization assays (in-vitro cell-based assays or competitive ligand-binding assays) directed at the biologically active site, to inhibit the functional activity of BsAb. Major histocompatibility complex-II (MHC-II)-Associated Peptide Proteomics

(MAPPs) assay can screen and quantitate naturally processed and presented MHC-II peptides on the surface of antigen-presenting cells, which are then further characterized for immunogenicity using in-vitro assays. T-cell epitope-mapping prediction tools are also used to identify the CD4 T cell epitope within the amino acid sequence of the therapeutic antibody and determine the strength of peptide binding to HLA molecules [82,83]. PK for biologics is often at least partially determined by FcRn-mediated recycling. In-vitro assays designed to measure binding of a therapeutic antibody to FcRn via its Fc domain, including SPR-based FcRn binding assays and FcRn-affinity chromatography, have been shown to be indicative of FcRn-mediated clearance and are frequently used to assess potential impacts to PK [84,85].


**Table 1.** List of bioassays for bispecific molecules.

BsAbs: bispecific antibodies; ELISA: enzyme-linked immunosorbent assay; HRP: horseradish peroxidase; IFN-g: interferon gamma; LDH: lactate dehydrogenase; MoA: mechanism of action; NFAT: nuclear factor of activated T cells; SPR: surface-plasmon resonance; VH: variable heavy domain; Ang-2: angiopoietin-2; VEGF: vascular endothelial growth factor; PD-L1: programmed death-ligand 1; CSPG: chondroitin sulfate proteoglycan.

### **3. Bioassays for Bispecific Antibodies and Case Studies**

*3.1. Bioassays for Biotherapeutics*

For biotherapeutics, a selective, physiologically relevant bioassay is essential to report on the product's potency and stability, by providing an assessment of the molecule's biological activity. Bioassays, in principle, can range from recognition of a particular antigen in a simple binding method, through systems as complex as blocking an inhibitory ligand that restores a co-stimulatory effect in a cell-based method. Selection of an appropriate method has its challenges rooted not only in the need to mimic the MoA, but also because bioassays can be costly to develop, perform, transfer, and maintain. Despite efforts to implement measures to ensure method control, cell-based bioassays can be inherently variable and often lack the precision and robustness of biophysical methods simply because they use living organisms, tissues, or cells.

While the general principles of bioassay design and strategy (e.g., measuring antigen target binding and biological activities) apply to bispecific antibodies, developing bioassays for bispecific antibodies requires unique considerations as bispecific antibodies bind two different targets with distinct mechanisms of action from monospecific biotherapeutics. A diverse range of bioanalytical assays have been developed and employed to study BsAbs, including methods designed to assess binding, potency, biological function, and purity. Figure 3 depicts a few of the methods involved in the various types of BsAb bioassays, which are further discussed in the following sections, and case studies are summarized in Table 2.

**Figure 3.** Representative bioassays for BsAb: (**a**) Reporter gene T-cell activation assay; (**b**) Single-arm binding SPR assay; (**c**) Cell proliferation assay; (**d**) Bridging ELISA. MTT:3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide.


**Table 2.** BsAb categories and potential bioassays applicable: Summary of case studies.

BLI: biolayer interferometry; TR-FRET: time-resolved fluorescence resonance energy transfer.

#### *3.2. BsAb Bioassay: Binding Assays*

ELISA and SPR are commonly used for in-vitro characterization of antigen binding for BsAbs. ELISAs are advantageous in that they are sensitive, typically fast to develop compared to cell-based assays, relatively inexpensive, and can be performed in complex matrices (e.g., cell lysates) [98,99]. Bridging ELISAs and competitive binding ELISAs can also provide information on the ability of the BsAbs to bind both antigens simultaneously, and they can therefore potentially be used as MoA-reflective potency assays, at least during initial development phases. The case studies presented below provide examples of competitive and bridging ELISAs used to confirm the simultaneous binding of two different targets for a tetravalent IgG-like BsAb. Despite their advantages, ELISA assays have drawbacks, one of which is that they are end-point assays and do not provide information on binding kinetics such as on and off rates [100]. In contrast, SPR measures continuous binding in a flow cell without the need for chemical labels, and the entire binding event can be analyzed in real time (association and dissociation). This allows for the determination of both kinetic and thermodynamic parameters through various data analyses [101]. SPR assays can also be designed to measure binding to two targets simultaneously, and are also potential candidates for MoA-reflective potency by binding assays. The case studies presented below provide two examples of assay formats for the purpose of measuring concurrent antigen binding by SPR.

Drawbacks for both ELISA and SPR are that it can be difficult to measure the impact of either target density (avidity effects) on BsAb binding, which can be important factors for a BsAb's activity in an in-vivo context. SPR allows for control of target density by depositing different amounts of capture ligand on the sensor chip. Binding of the therapeutic antibody can then be characterized under the different conditions to investigate the effects of receptor density on binding affinity/avidity [102]. However, the fact that BsAbs bind two targets, with varying levels of avidity depending on structure/format, can make this type of experiment challenging to design and interpret using label-free mass-based detection by SPR [103]. Additionally, there are open questions with respect to the in-vivo relevance of binding events [e.g., how relevant is binding to an immobilized ligand on a chip, or is solution-based association of a truncated ligand (such as a peptide or extracellular domain)

reflective of binding to a cell surface receptor?] [104]. Investigators have used SPR to measure binding of aCD20 mAbs to membrane bound CD20, and developed sophisticated software that makes it possible to extract and analyze individual binding events from heterogeneous mixtures. There have also been reports of affixing multiple ligands on a sensor surface in a solution-like context using DNA-directed immobilization using SPR, which would be useful for characterizing BsAbs. New biosensor technologies that allow for discrete detection of both binding events and precise control over surface density, as well as advances in SPR data analysis and experimental design, may provide avenues for more thoroughly investigating complex binding events under increasingly biologically relevant conditions in the future [105–109].


based on both individual readouts: two binding events can be measured, and the third parameter can be accurately calculated based on these measurements.

#### Cell Surface Ligand Binding Assays

Binding properties of investigational BsAbs to their targets can also be assessed by flow cytometry, which can be used to measure binding specificity and selectivity of BsAbs in a cellular context—information that is not captured in a traditional SPR or ELISAbased binding assay. Flow cytometry, is a fluidics and optics-based method that evaluates fluorescently-labeled cell suspensions in a single cell flow to capture receptor- or antigenbinding events in intact cells. However, flow cytometric analysis of antibody binding is an indirect measurement of kinetic values and it should be used in combination with SPR analysis to provide truly comprehensive, label-free, and accurate kinetic data for the antibodies being studied.

In addition to flow-cytometry assays, cell-based reporter assays have been developed to measure gene expression in response to disruption of an inhibitory binding interaction, such as PD-L1/PD-1 [93]. As a result, these assays provide a functional measure of BsAb binding as opposed to, for example, directly measuring binding affinity by SPR. In the case studies discussed below, variations of flow cytometric analysis are used to demonstrate the preferential binding, receptor blocking, and avidity of binding to dual-antigen-expressing target cells. In addition, application of a reporter assay to assess BsAb-mediated blockade of receptor-ligand interaction between the antigen-expressing tumor cells and effector cells is reviewed.


uate the binding activities of the investigational BsAb. Wild-type ectopically hPD-L1 expressing CHO cells (CHO.PD-L1) were incubated with test antibodies, labeled with a fluorescent secondary antibody, and analyzed by flow cytometry. Dose-dependent binding specific to CHO.PD-L1 cells was observed for the BsAb. These binding activities were replicated in several representative cancer cells endogenously expressing both CSPG4 or PD-L1. In addition, a flow cytometry-based competitive binding assay was used to assess the overall binding strength (avidity) of PD-L1xCSPG4 BsAb to CSPG4+/PD-L1+ cancer cells. BsAb binding was strongly inhibited in the presence of competing parental anti-CSPG4 mAb and only weakly inhibited in the presence of competing PD-L1-blocking mAb. These experiments demonstrate that PD-L1xCSPG4 binds to both PD-L1 and CSPG4 and that the strength of the interaction between the BsAb and CSPG4+/PD-L1+ cancer cells is primarily dominated by binding to CSGA4. To further show that the enhanced binding of the BsAb to CSPG4+/PD-L1+ cells is driven by avidity, cells were pre-incubated with a fluorescent anti-PD-L1 mAb, before being exposed to the test BsAb and a control BsAb, capable of binding PD-L1 but not CSPG4. The EC50 of PD-L1xCSPG4 for displacing the probe was substantially lower compared to the control BsAb. Performing the experiment in the presence of an anti-GSPG4 mAb increased the EC50 of the PD-L1xCSPG4 BsAb to a level similar to the control BsAb. Together these flow cytometry-based binding assays demonstrated that the PDL1xCSPG4 BsAb has enhanced selectivity for CSPG4+/PD-L1+ cancer cells driven by avidity binding.

• Case Study: Cell-based reporter assay to measure cell surface binding of PDL1xCSPG4 BsAb [93]: The authors further evaluated the role of CSPG4 in mediating the PD-L1 blocking capacity of the PDL1xCSPG4 BsAb using a PD-1/PD-L1 blockade reporter bioassay. The assay relies on co-culturing of Jurkat.PD-1-NFAT-luc reporter T cells (Jurkat cells engineered to express luciferase under the control of a NFAT response element and PD-1) and CHO.PDL1/CD3 cells (CHO cells engineered to express PD-L1 and a membrane-linked agonistic anti-CD3 antibody). Upon successful interaction of PD-1 and PD-L1 between the two cell types, TCR signaling and downstream NFATmediated luciferase activity in the Jurkat cells is inhibited. In contrast, interrupting the PD-1/PD-L1 interaction leads to NFAT-mediated luciferase activity. Addition of the PDL1xCSPG4 BsAb to the co-culture disrupted the PD-1/PD-L1 interaction between the two cell types in a dose-dependent manner, as measured by luminescence detection. Next, they tested the role of CSPG4 mAb in PD-1/PD-L1 blocking capacity of PDL1xCSPG4 BsAb by replacing the CHO.PD-L1/CD3 cells with a CSPG4+/PD-L1+ cancer cell line (the CD3 stimulation of T cells was achieved by pre-treating the cells with BIS1; an EpCAM-directed CD3-agonistic bsAb). Stimulated reporter T cells were co-cultured with the double-positive cells in the presence of PDL1xCSPG4 BsAb or controls, with and without anti-CSPG4 mAb. The ability of the PDL1xCSPG4 BsAb to block PD-1/PD-L1 interaction was reduced in the presence of anti-CSPG4 mAb. These findings suggest that the BsAb's PD-1/PD-L1-disrupting activity will be enhanced against CSPG4+/PD-L1+ cells compared to CSPG4-/PD-L1+ cells.

#### *3.3. BsAb Bioassay: Potency Assays*

In particular, the strategy of using a potency assay for BsAbs is challenging due to its complicated MoA with two target bindings, and it should be tailored to be MoA-reflective while meeting QC and regulatory expectations to be robust and sensitive methods to detect any structural changes in stability. One interesting question with respect to the BsAb potency assay is if two assays are needed for each target binding or if one potency assay would suffice. Depending on its MoA, either one or two potency assays would be suitable, but it is preferred to have one potency assay to measure synergistic biological effects of two target bindings or a dual read-out of the binding assays in a single assay.

A single assay that can fully capture the bioactivity of the therapeutic molecule is advantageous from both a cost/labor perspective and from a control perspective—synergistic

effects resulting from dual antigen binding may be missed if data from multiple assays measuring discrete events are used. However, in order to show the assay is suitably MoA-reflective, the key events in the MoA must be relatively well understood, and characterization assays designed to measure each event (e.g., binding to either antigen, receptor activation, etc.) are needed. The two case studies below describe the development and justification of single QC potency assays to measure changes in bioactivity for (1) a TDB and (2) a DAF that inhibits ligand binding to two distinct cell surface receptors.


#### *3.4. BsAb Bioassay: Effector Function Assays*

Some BsAbs target cell surface proteins or receptors with the intent of enhancing effector function. One arm often targets a tumor-associated antigen while the other targets an immune system-evading surface protein (such as CD47 or CD55/59), increasing susceptibility of the tumor cell to lysis by complement or NK cells, or phagocytosis by macrophages.

Other BsAbs have a primary MoA that does not involve effector function (e.g., TDB, or receptor blocker) but have an effector-competent Fc domain and can also exert cell killing activity through effector function. Depending on the MoA and other molecule-specific factors, effector function can be associated with unfavorable safety events, and so, effectorsilenced Fc domains are preferred [110,111]. In other cases effector function enhances a molecule's activity [54,57].


#### *3.5. BsAb Bioassay: Impurities Assays*

Impurities assays for BsAbs are often physicochemical assays such as size-exclusion chromatography (to measure aggregates and fragments), imaging capillary isoelectric focusing/ion exchange chromatography (to measure charge variants), and mass spectrometry (to sensitively identify and/or quantify post-translational modifications and other trace variants) [112], which are commonly used to characterize impurities for conventional mAbs. However, the unique structure of BsAbs can produce unique product variants with impacts to safety and/or bioactivity that are not fully addressed by physicochemical assays. The nature/activity of such impurities is rooted in the structure of the molecule, its production

process, and MoA. In order to illustrate this point a case study describing the development of a bioassay to measure T-cell activating impurities for a TDB is described below.

• Case Study: Luciferase reporter T cell activation assay to measure functional effects of impurities on CD3e-targeting TDB [95]: A CD3e-targeting TDB produced by knobsinto-holes technology and assembled in vitro contains a number of product-related impurities with the potential to activate T-cells in the absence of target cells. For example, aggregates and aCD3 homodimer, which result from the mispairing of aCD3 half antibody fragments during production, are characterized by multivalent binding to CD3 and can crosslink the TCR resulting in activation. These impurities are a safety concern because T-cell activation is linked to adverse events such as cytokine release syndrome. While aggregates and aCD3 HD can be measured using analytical methods, a bioassay is needed to assess their biological impact, such as target-independent T-cell activation. To address this need, the authors developed a reporter gene assay that measures T-cell activation in the absence of target cells using Jurkat T-cells engineered to express luciferase when activated. T-cell activation of product-related impurities present in the TDB formulation was quantified relative to T-cell activation by aCD3 HD standard. Using this assay, the authors were able to characterize the T-cell activating activities of aggregates and other product-related impurities, in order to get an idea of their potential impacts to safety and inform on the overall control strategy. Additionally, because the assay is a "catchall" assay that measures the combined T-cell activating activity of product-related impurities that may be present in a given sample, the method is able to provide reassurance that combinations of impurities are not leading to unexpected T-cell activation. Such combination effects would not be identified using physicochemical methods alone.
