Immunogenicity risk assessment is an important component of biotherapeutic drug development, and part of the overall benefit risk assessment. A robust immunogenicity risk assessment process ensures that the most appropriate candidate molecules advance into the clinic, and that clinical immunogenicity is appropriately monitored. Risk assessment for multi-specific therapeutics can be especially challenging given that many molecules in this family can bind immune cells, trigger T cell activation, or bring together disparate molecules in a non-native way. Immunogenicity risk assessments include both the risk of generating an immune response and risk of clinical consequences if an immune response is generated. The latter is particularly critical to the overall benefit risk assessment and likely to inform the acceptable level of immunogenicity risk.
The potential clinical applications of multi-specific therapeutics are demonstrated by 3 approved products (Table I and II) and a breadth of molecules currently under investigation (1, 2). These include obligate concepts where having both specificities in the same molecule is critical (Table II), and combinatorial concepts where the two specificities do not necessarily need to be in the same molecule but may provide additional efficacy benefit over combination therapy (Table III). Of the 92 multi-specific therapeutics in clinical development as of March 2019 (2), 78 are for cancer and 14 in other indications such as autoimmune disorders, infectious diseases, hemophilia, diabetes, and ophthalmology. There are additional concepts that are no longer in development not represented in these tables, such as the anti-TNFα and anti-IL-17A bispecific hypothetical case study presented herein (3,4,5,6,7), and new concepts introduced since the source article was written in 2019 (2), such as the Ang-2 and VEGF bispecific faricimab (8).
As there are no formatting requirements around authoring an immunogenicity risk assessment, this article is part of a series designed to provide examples for consideration. The document structure of the immunogenicity risk assessment presented here is a result of input from the authors and their view of a fit for purpose risk assessment that may be submitted to health authorities. It does not reflect a structure endorsed by health authorities nor the experiences of any particular sponsor. This content is generally consistent with a truncated version of the integrated summary of immunogenicity structure option proposed by health authorities (14, 15), with removal of sections not pertinent depending on the development stage of the biotherapeutic program.
As with all protein therapeutics, one of the main drivers of risk of developing an immune response for a multi-specific therapeutic is the primary amino acid sequence and T cell epitope content (16). Any protein that has been engineered to change the amino acid sequence, has linker regions that introduce new linear epitopes that are not present in either parent, or that has non-natural or modified amino acids may have an elevated risk of developing an immune response. Most multi-specific therapeutics contain these elements, and consequently, these novel sequences should be evaluated for immunogenic risk using available tools such as in silico prediction, in vitro T cell assays, MHC binding assays, or ex vivo models (17).
Additionally, multi-specific therapeutics employ diverse protein engineering techniques such as quadroma, knobs-in-holes (KIH), CrossMAb, Triomab, strand-exchange engineered domain (SEED), cross-over dual variable (CODV), DART® (Dual-affinity Retargeting), TRIDENT®, dock-and-lock (DNL), BiTE® (Bispecific T cell Engager), bispecific killer engager (BiKE), trispecific killer engager (TriKE), multi-specific antibody-based therapeutics by cognate heterodimerization (MATCH), nanobody, diabody, diabody-Ig, etc. to bring together specific recombinant domains in a non-native way for simultaneous multi-target recognition (2, 18,19,20,21,22,23,24,25). The extensive protein engineering required to design and optimize such novel multi-specific therapeutics with mono- or multi-valency could inadvertently increase the product attribute-related immunogenic risk for these recombinant molecules.
Critical Quality Attributes
Product critical quality attribute (CQA)-related risk is another driver that is typically considered in protein therapeutic immunogenicity risk assessment (14, 26,27,28), and multi-specific therapeutics are no exception. The CQA risk factors such as protein aggregates, posttranslational modifications, host cell- and process-related impurities (such as host-cell proteins and DNA, endotoxin, chromatography resin, contaminants, and degradants, etc.), formulation excipients and container closure, etc. are also relevant to multi-specific therapeutics.
Given the non-native arrangement of disparate functional domains in multi-specific molecules, there is potentially a higher immunogenicity risk related to aggregation. For all other CQAs, the potential for an impact on immunogenicity risk is similar for mono-specific and multi-specific molecules.
Mechanism of Action—Immune Modulation
As the industry gains experience with immunomodulatory multi-specific therapeutics, pharmacological or mechanism of action (MoA)-based immunogenic risk factors are emerging as a crucial factor impacting the immune response to the drug. For instance, does the protein involve antagonism of more than one immune checkpoints, does it bind B cells, and/or does it agonize a costimulatory molecule? If the answer is yes to one or more of these possibilities, the molecule may have a higher than average risk of eliciting anti-drug antibodies (ADA), to be assessed during the clinical program. Data from combination studies with nivolumab and ipilimumab have clearly demonstrated this concept. When used as a monotherapy, nivolumab had a reported incidence of anti-nivolumab antibodies of 11.2%. As a combination therapy with ipilimumab, however, the incidence jumped to 37.8%, presumably based on the same antibody assay (29, 30). A similar finding was observed with durvalumab and tremelimumab (31). Multi-specific therapeutics that combine multiple immune stimulatory domains into the same molecule will carry this enhanced immunogenic risk.
When assessing MoA-based immunogenicity risk factors, direct binding to B cells of at least one domain of the multi-specific therapeutic must be considered. For example, a multi-specific therapeutic may be intended to deliver a signal to a specific subset of cells expressing target X via linkage of the signaling domain to an anti-target X antibody. In this scenario, B cell clones that recognize the antibody may also efficiently receive the signal. If the signaling domain has the ability to modulate B cell biology, then the antibody response to the therapeutic could be either enhanced or diminished.
Evidence also suggests that inclusion of a CD3 binding domain into a multi-specific therapeutic may be an immunogenicity risk factor. In general, molecules that bind CD3, such as otelixizumab or teplizumab, appear to be prone to ADA responses (32, 33) and this may hold true for T cell engagers as well, with AMG 211 eliciting antibodies in all subjects treated at > 3.2 mg in a phase 1 study (34). There are several plausible hypotheses for this finding depending on whether the mechanism is T cell engagement or T cell binding. For mechanisms engaging T cells, ADA responses may be more likely in the presence of wide-spread T cell activation. For T cell binders that are not intended to activate T cells, anti-CD3 domains may increase linkage of T cells to B cells that recognize the other domain(s) of the multi-specific therapeutic, potentially enhancing the likelihood of drug-specific B cell clones receiving T cell help.
On the other hand, the MoA of some biotherapeutics may also mitigate the risk of generating an immune response. One potential example of this is blinatumomab, which depletes CD19 + B cells. Even though it is composed of 2 murine single chain antibodies, the clinical immunogenicity incidence remains less than 2% (10, 11). Other multi-specific therapeutics may mitigate the immune response via other mechanisms, such as expanding regulatory T cells.
Mechanism of Action—Complex Formation
Another MoA-based immunogenicity risk factor is the ability of the therapeutic to form large complexes with the target(s). This means targets that are oligomeric (as mentioned earlier) and either soluble or shed at significant levels pose the highest risk of large complex formation. Large complexes have long been hypothesized to be immunogenic (35, 36).
While there are no definitive examples of mechanistically proven, large complex-mediated immunogenicity to a therapeutic, strong correlative data exist. Adalimumab and infliximab both bind trimeric TNFα and form complexes of 4,000 kDa and 14,000 kDa, respectively; both are immunogenic in clinical studies, while etanercept does not form large complexes and is relatively non-immunogenic (37, 38). Multi-specific therapeutics can further enhance this complex forming ability since there are multiple targets. For instance, ABT-122 and JNJ-61178104 are both built on an adalimumab backbone and involve the addition of an IL-17A binding domain (39, 40). In healthy volunteer studies, ABT-122 and JNJ-61178104 have a reported ADA incidence of 99% and 100%, respectively, presumably due to enhanced ability to form large complexes (3, 41). In order to assess this risk, complex formation should be evaluated preclinically so that the ratio in which the therapeutic candidate binds the target is well understood. Any downstream effector function driven by the drug Fc domains should also be explored.
In addition to soluble complexes, multimeric membrane bound targets can mediate complex formation on the surface of cells. When these targets are expressed in antigen-presenting cells (APCs), it can facilitate complex uptake and increase the T cell mediated adaptive immune response (42).
Treatment-Related Risk Factors—Patient and Regimen
As with all proteins, the intended patient population could also contribute to the immunogenic risk of multi-specific therapeutics. Important considerations include immune status of the patient population and prior exposure to related therapeutics. In an example for prior exposure, subjects with prior exposure to adalimumab have an increased immunogenic risk for a multi-specific therapeutic built on an adalimumab backbone. Immune status may be on a population level, such as autoimmune disease, or on an individual level, such as immunosuppressive methotrexate therapy.
Treatment-related immunogenicity risk factors such as administration route, dose, dosing frequency, treatment duration and concomitant immunomodulators (14, 27, 43, 44) are also critical elements in the immunogenicity risk assessment for multi-specific proteins. The extent of immune response due to treatment-related risk factors is dependent on the complex interplay between product-related intrinsic factors and patient-related extrinsic factors.
Historically, subcutaneous administration (SC) has been considered more likely to produce an immune response compared to intramuscular (IM) and intravenous (IV) routes; however, this has not been observed consistently for products with both SC and IV routes approved (45, 46). An analysis of clinical data for 27 non-oncology immunomodulatory antibodies showed no clinically meaningful difference in immunogenicity incidence between IV and SC administration routes (31). While the route of administration can be used to inform risk of positive seroconversion, it must be assessed in the clinic. SC administration may have an increased risk of protein aggregation due to dosing high concentrations in small volume (47). The aggregates and other antigenic attributes may cause an increase in the therapeutic protein uptake, processing and presentation by APCs resulting in immune responses (48). High therapeutic doses in some cases have been shown to saturate neutralizing Ab responses and restore therapeutic binding to its intended target. High doses may also induce long-term immune tolerance (49).
Longer treatment duration may also increase the risk of an immune response for protein therapeutics. The clinical immunogenicity incidence of IFNβ-1a-Rebif increased with a more frequent dosing regimen and subcutaneous administration relative to IFNβ-1a-Avonex (50, 51), with the caveat that this increase could be due to different study populations, immunogenicity assay formats, etc. While the highly potent immune-stimulatory multi-specific therapeutics may not require long treatment duration, the immune-suppressive multi-specific therapeutics may benefit from their pharmacological activity, irrespective of treatment duration. The concomitant medications (e.g., antihistamines, corticosteroids, methotrexate, interferons) and standard of care may also augment or reduce the antigen-processing of multi-specifics through their immunomodulatory effects.
Potential Clinical Consequences
The risk of consequences from an immune response needs to be considered in addition to the risk of generating an immune response.
For multi-specifics that contain non-antibody components, a key aspect for multi-specific therapeutics is if one or more components resembles an endogenous counterpart, the potential consequences of cross-reactivity of ADA are determined by the function and uniqueness of that endogenous counterpart (52). Since the multi-specific therapeutic may be more immunogenic than the native protein, any increased incidence may adversely impact patient safety to a greater extent than occurs for the mono-specific compound.
While non-clinical studies are not predictive of the clinical incidence (16, 53, 54), they can provide valuable information on potential clinical consequences of an immune response if generated. For example, if a multi-specific therapeutic antibody, which was engineered to include an immunostimulatory cytokine, drives an enhanced immune response in cynomolgus macaques relative to the parent antibody alone, that enhancement may be clinically relevant if the cytokine biology is conserved across species. An important caveat to this nonclinical risk assessment approach is that immunogenicity driven by foreign sequence (in this case the human sequence) cannot be differentiated from MoA-driven immunogenicity with any degree of certainty and this must be acknowledged in the integrated immunogenicity risk assessment.