Risk-based approach of bioanalytical methods for clinical immunogenicity assessment of multidomain biotherapeutics

Antibody engineering has dramatically evolved in recent years, resulting in current conventional antibody therapeutics having fewer immune-related adverse effects observed in the clinic. However, the advanced antibody engineering technologies have also created a great variety of novel antibody-based therapeutics with bispecific and multispecific domains entering to the clinic [1]. While having multiple specificities in the same molecule that are capable of binding to multiple different molecular antigens, or two different epitopes on the same antigen, may provide additional efficacy benefit over combination therapy, more bioengineering modifications and complex manufacturing processes are required for multidomain biotherapeutics compared with the conventional monospecific antibody therapeutics. As a result, multidomain biotherapeutics may potentially hold greater risk for immunogenicity and present unique challenges for immunogenicity analysis. Some clinical data have highlighted immunogenicity as one of the key challenges in the development of multidomain biotherapeutics. A recent publication reported that the bispecific antibody LY3415244 showed increased immunogenicity as compared with the two parental antibodies, in which all 12 patients developed treatment-emergent antidrug antibodies (TE-ADAs) to LY3415244 [2]. In addition, pre-existing antidrug antibodies (PE-ADAs) to LY3415244 were also detected in serum samples from normal human donors and the enrolled patients unexposed to the drug [2,3]. Another clinical trial also reported that strong immunogenic responses were observed for a bispecific antibody, which were unexpected as similar findings were not observed in the two previously studied parental molecules [4]. Therefore, there is an increasing need for immunogenicity investigation of multidomain biotherapeutics as part of drug development in the pharmaceutical and biotechnology industry. To guide the industry on the development of this rapidly growing class of biotherapeutics, the US FDA in 2021 published regulatory guidance on bispecific antibody development programs [5]. Although the guidance is intended for bispecific antibodies, the principles discussed in the guidance may also be applied to other types of bispecific protein products and multispecific products. In general, many aspects of a bispecific antibody development program will be like monospecific monoclonal antibody development programs. However, the guidance notes that an immune response to one domain may inhibit a specific function while leaving others intact. Therefore, it recommends that when examining immune responses to bispecific antibodies, it may be appropriate to develop multiple assays to measure immune responses to different domains of bispecific antibodies. This recommendation is in alignment with other guidelines from the FDA and European Medicines Agency on bioanalysis for bispecific/multispecific biotherapeutics [6,7]. The guidance also encourages the industry and other stakeholders to engage the FDA to discuss their individual bispecific antibody development programs. As such, the challenge facing the bioanalytical community is how to apply the FDA’s guidance in the immunogenicity assessment process. To address the challenge, at the 2021 Workshop on Recent Issues in Bioanalysis (WRIB), a hot-topic discussion was devoted to multidomain biotherapeutics and bispecific antibody immunogenicity. Given the industry’s limited experience with immunogenicity of multidomain antibodies, the resultant 2021 WRIB white paper recommends that customized bioanalytical strategies and methodologies be planned based on the target biology and mechanism of action, phase of development and key reagent availability [8]. At the 2023 WRIB, continued discussion on the topic was held to provide the global bioanalytical community with the current industry/regulator consensus in support of clinical immunogenicity assessment of multidomain biotherapeutics [9]. Although the industry and regulatory agencies are generally aligned in a risk-based approach to the immunogenicity assessment of multidomain biotherapeutics, at this time the bioanalytical practices in the development of multidomain biotherapeutics have not been harmonized in terms of bioanalytical strategies and methodologies. At Boehringer Ingelheim, we have supported several multidomain biotherapeutic programs for their clinical immunogenicity assessment. In this commentary, we will present our risk-based approach on clinical antidrug antibody (ADA) detection and characterization of multidomain biotherapeutics. Additionally, case studies will be provided that describe the reagents and tools along with the methods used to characterize domain specificity and identify the location of PE-ADAs in clinical ADA samples. We will also share our strategy for the development of an orthogonal clinical ADA method to differentiate PE-ADAs from TE-ADAs in clinical ADA samples.

Clinical ADA analysis & characterization of multidomain biotherapeutics

Before it is finally decided that the molecule will advance to clinical development, a process on immunogenicity risk assessment and bioanalytical assay strategy has taken place. The immunogenicity risk assessment for multidomain biotherapeutics has been well reviewed [10,11] and is out of scope of this commentary.

Overall, our clinical ADA analysis and characterization is a tiered, clinical phase-appropriate approach. For non-pivotal (phase I and II) clinical trials, the immunogenicity assessment includes a standard Meso Scale Discovery electrochemiluminescence (MSD-ECL) drug-bridging ADA method. The ADA analysis includes the use of three tiered assays: screening, confirmatory and titration. The screening assay identifies the potential ADA-positive samples, using the screening cut point to distinguish between positive and negative samples. For the confirmatory assay, the potential positive samples are preincubated with an excess of unlabeled drug. The binding of the unlabeled drug to ADAs results in signal inhibition in the bridging MSD-ECL assay. The confirmatory assay identifies ADA-positive samples, including those with PE-ADAs that are specifically able to bind to the drug. The confirmatory cut point is used to distinguish the positive ADA samples to the drug from the potential positive samples identified in the screening assay. In the titration assay, the confirmed positive samples are diluted to determine the highest dilution that is just above the titer cut point. This allows the semiquantitative determination of ADAs present in the clinical samples. The standard ADA analysis is also able to identify the patients who have PE-ADAs, as these samples are naive/pretreatment samples that screen and confirm positive in the ADA assay.

During clinical phase I and phase II trials, the immunogenicity risk assessment and bioanalytical assay strategy should be reviewed again once clinical immunogenicity data become available. As a result of the review, for pivotal/phase III trials, in addition to the standard ADA assays, further ADA characterization may be needed based on the clinical ADA data. This may include neutralizing antibody, domain specificity and TE-ADA assessment to distinguish TE-ADAs in post-dose positive samples from PE-ADA-positive patients. As examples, in the following sections we will present two case studies for domain specificity characterization and PE-ADA assessment.

Case study 1: domain specificity characterization

In this case, BI X is a humanized IgG1-based bispecific antibody with two domains: domain A against target A and domain B with single-chain variable fragments (scFv) against target B, respectively. It is in phase I clinical development.

An MSD-ECL drug-bridging ADA method with acid dissociation was developed and optimized to measure anti-BI X antibody in human serum. Briefly, in the screening assay, study samples, positive controls and negative controls are first treated with glycine-HCl. The acid-dissociated samples are then neutralized and co-incubated with biotin-labeled BI X and ruthenium-labeled BI X. Antibodies to BI X present in the samples bind to both the biotin- and ruthenium-labeled BI X. The resulting complexes are captured via the binding of biotinylated BI X to a streptavidin-coated multiarray MSD plate. In the presence of tripropylamine-containing read buffer, ruthenium produces a chemiluminescent signal that is triggered when voltage is applied. The resulting chemiluminescence is proportional to the amount of ADA present in the samples. The confirmatory assay is run in the same manner, except that samples are screened both unspiked and spiked with an excess amount of BI X.

During the standard ADA analysis, ADAs were detected and specificity against the drug was confirmed. The purpose for domain specificity characterization of ADAs present in the clinical samples is to determine which domain(s) the immune responses are directed toward. To do this, there are bioanalytical challenges, which include but are not limited to: necessity of having the right ‘representative’ tools (positive controls) and monospecific or domain-specific ‘drugs’ to characterize specific domains; complexity of validation and time/resource consumption, which is more complicated than for a standard monoclonal antibody therapeutic; and data interpretation for ADAs is required against the whole molecule and each of the domains. For this case study, four tool monoclonal antibodies (mAbs) were generated: PC-mAb-A and PC-mAb-B were used as positive controls against domain A and domain B of BI X, respectively, and mAb-A and mAb-B contain the same domain as domain A and domain B of BI X, respectively.

In brief, our domain specificity test assays can be considered as an extension of the confirmatory assay of the ADAs against the drug. In the standard ADA analysis, polyclonal antibody against the drug was used, while in the domain-specific ADA assays, PC-mAb-A and PC-mAb-B were used. Likewise, in the confirmatory ADA assay, either mAb-A or mAb-B was preincubated with the samples instead of the drug. The percentage inhibition of the signal in the presence of mAb-A is used to determine the specificity for domain A, and in the same way, the percentage inhibition of the signal in the presence of mAb-B is used to determine the specificity for domain B. As in the confirmatory assay for the drug, the domain specificity cut points were determined for each domain of BI X to determine the domain specificity of the ADA in the sample.

Using the drug-bridging assay format with acid dissociation to improve drug tolerance, with suitable reagents (positive controls) against both arms of BI X along with versions of monospecific drugs representative of the domains of BI X, we developed a screening assay and three different confirmatory assays to characterize the ADA responses against the whole drug and each domain of BI X. If clinical data emerge that require additional characterization to understand the clinical impact of immunogenicity to BI X, the domain specificity assays described above will be validated and used to support sample analysis for the pivotal/phase III clinical trials.

Case study 2: detection & characterization of PE-ADAs

In this case study, the molecule is the same bispecific antibody used in case study 1, BI X.

During early method development, numerous high ‘ADA’ signals were detected in various naive/blank individual matrix lots. These individual matrix lots were screened, confirmed and titrated in the standard ADA assays. In addition, the magnitude of titers between individuals varied, and most individuals were confirmed to have a percentage inhibition of >60%. Due to these unexpected data, orthogonal experiments were done to investigate the ‘false-positive’ results; however, the same positive outcomes were observed in all experiments, which suggested that these results were due to PE-ADAs.

In a drug-bridging ADA assay, the target molecule could cause false-positive signals by forming dimers or polymers in the matrix. To rule out the possibility of false-positive signals detected in the assay due to potential dimerization (polymerization) of targets in the matrix, a target interference test was carried out. The experiment was done by spiking the two target molecules separately and in combination in various concentrations into pooled human serum (negative control) and testing these samples in the ADA method. The results showed that no significant differences (<20%) in percentage inhibition were observed between all target interference samples and the negative controls, indicating no target interference in the ADA method and no false-positive assay signals were observed due to the presence of drug targets.

To confirm that the positive responses from naive/blank individuals observed in the assay were due to IgG, additional experiments including an IgG depletion were performed. For this, the ADA assay negative and positive controls and naive/blank individuals with negative and positive responses observed in the assay were passed through a Protein G column plate (Pierce™ Protein G Spin Plate, catalog no. 45204, MA, USA) and the flow-through from each sample was collected in a collection plate. The before and after IgG depletion samples were tested in the ADA screening assay. The results showed that the positive controls and positive naive/blank individuals prior to depletion became negative (i.e., the same as the negative controls) after the depletion step, indicating the positive ADA results were due to IgG.

Based on the PE-ADA finding, a strategy was effectively developed to identify the portion of the molecule that binds to PE-ADA. Two tool monoclonal antibodies (mAb-A and mAb-B) were generated for the investigation. mAb-A has the same domain as domain A of BI X, and mAb-B has the same domain as domain B of BI X, respectively. mAb-A and mAb-B confirmatory inhibition tests were done by adding an excess amount of each in the ADA confirmatory assay. Previously tested positive naive/blank individual lots were used in this assessment.

When comparing BI X versus mAb-A and mAb-B confirmatory results (percentage inhibition), positive inhibitions were observed for positive individuals against BI X and mAb-B, but not mAb-A. This indicates the positive responses were due to the binding of PE-ADA specifically to domain B of BI X, which contained the scFv.

Strategy for differentiating PE-ADAs from TE-ADAs

In the above case study 2, using mAb-A and mAb-B, we were able to identify domain B of BI X as the binding domain of the PE-ADA. It remains to be understood whether the presence of PE-ADAs increases the risk of TE-ADA development. Therefore, it is critical to establish a plan to be able to distinguish between PE-ADA and TE-ADA responses.

To differentiate the PE-ADA from TE-ADA, the essential step is to generate a modified BI X in which the PE-ADA epitope has been removed. This will involve some protein engineering work on domain B of BI X, such as modifications in the linkers within scFv and between Fc and scFv, and the orientation of scFv. The modified BI X is ruthenium-labeled and used as the detection reagent in the PE-ADA and TE-ADA differentiating assay. The post-dose ADA-positive samples from the PE-ADA-positive patients will be co-incubated with biotin-labeled BI X and ruthenium-labeled modified BI X. With this bridging assay format, only TE-ADAs to BI X present in the samples will bind to both the biotin- and ruthenium-labeled modified BI X and be measured. This TE-ADA-specific assay is called the drug-bridging TE-ADA assay. Further characterization is planned on samples that are positive in the drug-bridging TE-ADA assay for domain specificity, as described in case study 1, if this information would be needed to better understand the clinical impact of immunogenicity of the TE-ADA. It should be noted that the strategy proposed here is only to be used for patients with PE-ADA at baseline as measured in the standard MSD-ECL drug-bridging ADA assay and to compare the titers between the standard MSD-ECL drug-bridging ADA assay and the drug-bridging TE-ADA assay to understand the titer changes in PE-ADAs and TE-ADAs in these patients. Additionally, the drug-bridging TE-ADA assay would not be able to detect TE-ADAs to the PE-ADA epitope, which would result in a negative response.

Points to consider

From the BI X case studies presented above, to support pivotal/phase III clinical trials, the clinical ADA assessment may require additional tiers of ADA characterization for multidomain biotherapeutics beyond the standard immunogenicity testing. A pivotal/phase III clinical ADA sample analysis workflow may include the following steps:

• Standard approach of screen, confirm and titer of ADA responses against the drug product for all samples from drug-dosed patients, including pretreatment samples for these patients.

• Neutralizing antibody and domain specificity characterization for the positive samples from the standard approach.

• TE-ADA assay to distinguish TE-ADAs from PE-ADAs in post-dose positive samples from PE-ADA-positive patients.

The implementation of these additional ADA characterization methods for multidomain therapeutics will need specialized reagents and tool molecules which can require a lot of time to develop. Therefore, it is critical to use a risk-based approach that is based on clinical ADA data and start additional immunogenicity bioanalytical strategy planning once early clinical immunogenicity data become available. The immunogenicity risk assessment and bioanalytical assay strategy should be reviewed periodically along with clinical ADA results to ensure that appropriate bioanalytical assays are in place prior to pivotal/phase III trials. Overall, as an industry we have more than three decades of experience with clinical development of conventional monoclonal therapeutics, and the rapidly growing multidomain biotherapeutics are relatively new to us; however, we are continuing to grow our experience with drug development of multidomain biotherapeutics.