T and B cell epitope analysis for the immunogenicity evaluation and mitigation of antibody-based therapeutics

ABSTRACT

The surge in the clinical use of therapeutic antibodies has reshaped the landscape of pharmaceutical therapy for many diseases, including rare and challenging conditions. However, the administration of exogenous biologics could potentially trigger unwanted immune responses such as generation of anti-drug antibodies (ADAs). Real-world experiences have illuminated the clear correlation between the ADA occurrence and unsatisfactory therapeutic outcomes as well as immune-related adverse events. By retrospectively examining research involving immunogenicity analysis, we noticed the growing emphasis on elucidating the immunogenic epitope profiles of antibody-based therapeutics aiming for mechanistic understanding the immunogenicity generation and, ideally, mitigating the risks. As such, we have comprehensively summarized here the progress in both experimental and computational methodologies for the characterization of T and B cell epitopes of therapeutics. Furthermore, the successful practice of epitope-driven deimmunization of biotherapeutics is exceptionally highlighted in this article.

KEYWORDS:

• Anti-drug antibody
• antibody
• B cell epitope
• deimmunization
• epitope
• immunogenicity
• T cell epitope
• therapeutic protein

1. Introduction

Because of their superior bioactivity, specificity, and safety profile over small-molecule agents, therapeutic monoclonal antibodies (mAbs) are one of the most promising categories of therapeutic proteins (TPs) and now represent an essential component within the portfolio of the biopharmaceutical industry. Despite many advantages, as foreign macromolecules mAbs and other antibody-based TPs (Figure 1a) may elicit undesired humoral immune responses by triggering the development of anti-drug antibodies (ADAs),^1,2 which manifest diverse clinical implications by modulating the efficacy,^3–6 pharmacokinetics and pharmacodynamics (PK/PD),^7,8 and immune-related safety profile of biotherapeutics (Figure 1b).^9–11 Depending on the genetic origin of the mAbs (i.e., chimeric, humanized, human), the non-germline content located in either framework (FR) or complementarity-determining regions (CDRs) may give rise to significant ADA emergences following treatment. The magnitudes of ADA (both preexisting and treatment-emergent) need to be closely monitored throughout drug administration as mandated by the regulatory authorities.^12,13 For clinical investigations involving biotherapeutics with high immunogenicity risk, extended ADA monitoring beyond the end of studies may be required due to the sustained presence of ADAs in some individuals.^14,15

Figure 1. The effects of immunogenicity on antibody-based therapeutics. (A) The structural diagrams of different antibody-based therapeutics. mAb, monoclonal antibodies. KIH, knob-into-hole. BiTE, bispecific T cell engagers. ADC, antibody-drug conjugate. Fv, variable fragment. (B) The impacts on antibody-based therapeutics. The ADA bound to the paratope of mAb elicit neutralization capacity and therefore lead to the attenuation of therapeutic efficacy. The ADA may also trigger the onset of type I (IgE-mediated) and II (IgG-mediated) hypersensitivity. Once immunocomplexes are formed by ADAs (IgM or IgG) and therapeutics, the adverse events associated with type III hypersensitivity may take place. The large sizes of immunocomplexes also lead to faster clearance of therapeutics from blood stream. The graph was created on BioRender.com.

In recent years, many engineered antibody-based therapeutics (e.g., antibody-drug conjugates, multispecific antibodies) and antibody-like molecules (e.g., heavy-chain variable domains [VHHs], designed ankyrin repeat proteins) are being developed as drug or TP components and many products have entered clinical stages. Notably, the complexity and idiosyncrasy of ADA risk has intensified with the burst of innovative biological modalities such as multi-domain biotherapeutics (MDBs),¹⁶ cell therapies, and gene therapies.^17–22 Hence, there is a demand for an integrated strategy for immunogenicity assessment and characterization that can facilitate nonclinical discovery and clinical development, as well as post-marketing surveillance of biopharmaceuticals.

The generation of ADAs involves two mechanisms: T cell-independent (Ti) and T cell-dependent (Td) B cell activation (Figure 2).²³ The Ti process is initiated by the multivalent crosslinking of B cell receptors (BCR) by the repeated surface antigen of therapeutics, which is followed by the activation of BCR signaling and rapid secretion of antibodies. The antibodies generated by the Ti pathway are predominantly the short-lasting pentameric immunoglobulin M (IgM), which exhibit a low binding affinity to their targets due to the absence of somatic hypermutation and affinity maturation processes. One well-elucidated scenario involving the Ti-mediated antibody generation witnessed in biopharmaceutical research is the emergence of anti-polyethylene glycol (PEG) IgM subsequent to the administration of PEGylated nanomedicine.²⁴ On the other hand, the ADAs of IgG isotype, which harbors high affinity and long durability, are generated primarily via the Td pathway, which involves uptake and hydrolysis of TPs via endosomal-lysosomal system by antigen-presenting cells (APCs) such as dendric cells (DCs). The short peptides derived from TPs can possibly complex with the major histocompatibility complex class II molecule (MHC II, also referred to as human leukocyte antigen class II [HLA II] in human) in the antigen processing compartments and then be transported to the surface of DCs, allowing the engagement of peptide-MHC II complex (pMHC II) with the epitope-specific T cell receptor (TCR) expressed by follicular helper T lymphocytes (Tfh) in the germinal center.²⁵ DCs undergo maturation during antigen processing and produce cytokines that are necessary for CD4⁺ helper T cell activation. The MHC II-restricted peptides associated with complementary TCRs are defined as T cell epitopes (TCEs). The activated, antigen-licensed Tfh population further engages B cells that have encountered and presented the identical antigens, and produce the essential cytokines to induce the maturation of B cells toward the ADA-secreting plasma cells. A fraction of B cells, at this point, gains the potential to differentiate into memory population for the more rapid response after secondary immunization. During the process of antigen processing in B cells, specific regions within the antigens, known as B cell epitopes (BCEs), are recognized by BCRs to facilitate their uptake and processing. While TCEs can be distributed throughout the entire antigen protein theoretically, BCEs are limitedly located on the surface area that is directly accessible by BCRs/antibodies. The presence of TCEs and BCEs collectively dictates the probability of ADA generation. Thus, researchers are expected to explicitly evaluate the immunogenicity risk of biotherapeutics by examining both TCEs and BCEs, which may or may not overlap.

Figure 2. The schematic illustration of T cell-dependent and -independent B cell activation pathways. DC, dendritic cell. pMHC II, peptide-MHC II complex. TCR, T cell receptor. BCR, B cell receptor. The graph was created on BioRender.com.

Running tiered immunoassays for detecting the total and neutralizing ADA (Nab) is standard practice in the development of TPs, as endorsed by regulatory guidance as well as industrial white papers (or recommendations).^{12,13,26–29} However, immunoassays based on intact therapeutics do not provide in-depth information such as epitope profiles, which harbor critical insights outlining the implications of anti-therapeutic immunity. Numerous methods have thus been developed for TCE/BCE characterization to meet different research purposes. In this endeavor, we aim to offer a strategic overview of immunogenicity assessment by spotlighting the methodologies that have been leveraged to profile in particular the TCEs and BCEs of antibody-based therapeutics.

2. Experimental strategies for TCE discovery and immunogenicity derisking

The immunogenicity of TPs is primarily dictated by the Td immune responses. Therefore, the TCE-focused immunogenicity assessment, which can be conducted either preemptively in early discovery stages for lead evaluation or post hoc for clinical immunogenicity characterization, is of critical importance in delineating the developability of the candidate drugs. As some key steps of Td immune response against exogeneous can be recapitulated by cultured cells, many in vitro assay platforms were developed to monitor the T cell-mediated immunogenicity of biotherapeutics.^30,31 These methods involve the measurement of physical or functional binding of therapeutics-derived peptides with HLA II molecules as summarized in Table 1. Examples in the following paragraphs illustrate different methods for TCE characterization. It is worth noting that these strategies, namely the in vitro binding, immunopeptidomics, and cell-based approaches, are usually exploited in an integrated manner.

Table 1. Summarization and comparison of various strategies applied for TCE identification.

Category	Method	Cost	Time	Commercial availability	Ref
In vitro binding	HLA binding assay	++	+	Yes	^Citation32–35
Immunopeptidomics	MAPPs	+++	++	Yes	^Citation35–41
Cell-based functional assay	ELISpot	++	++	Yes	^Citation34,Citation38,Citation42
	T cell proliferation assay	++	++	Yes	^Citation38–40,Citation43

2.1. In vitro HLA binding

The peptide-MHC II binding stands as the prerequisite of Td immunogenicity prior to the CD4⁺ T cell activation. Considering the preserved molecular characteristics of peptide-MHC II interaction, researchers sought to interrogate the TCE profiles of antigens by examining the binding of synthetic peptides with MHC II molecules.

The in vitro HLA binding assay was performed for the identification or validation of the TCE clusters of therapeutic mAbs such as infliximab, a chimeric tumor necrosis factor (TNF)-targeting mAb with considerable immunogenicity in humans. Overlapping 15-mer peptides derived from the heavy and light chains of infliximab were synthesized and divided into subpools for the in vitro binding assay in the study by Hamze et al. Consequently, peptides near CDR2 and CDR3 of the heavy chain (HCDR2, HCDR3), FR1 and FR2 of light chain (LFR1, LFR2) were found to bind sufficiently with many HLA II alleles.³⁵ As the binding assay is generally conducted along with other methods instead of being used solely for the de novo identification of TCEs, we discuss more examples involving the immunogenicity analysis of antibody-based drugs in subsequent sections.

The use of in vitro binding assays also facilitated the understanding of immunogenicity risks posed by many TPs for enzyme replacement therapies. Erythropoietin (EPO) is commonly used to restore red blood cell generation of pure red cell aplasia patients who develop autoantibodies against EPO. As such, protein engineering strategies such as introducing N-glycans or abolishing TCEs were proposed to circumvent the immunogenicity issue.⁴⁴ Tangri et al. performed the in vitro HLA II binding screen with a synthetic peptide library derived from wild-type human EPO. Surprisingly, nearly all peptides exhibited binding to one or more HLA II alleles with varying affinities. On the foundation of the TCE landscape elucidated from the peptide screening, deimmunization of EPO was achieved by modifying the most immunogenic peptide and subsequently confirmed by the peripheral blood mononuclear cells (PBMCs) proliferation assay.³² The HLA II binding assay was also used for characterizing and diminishing the immunogenicity of Factor VIII (FVIII), a key coagulation factor that is absent in hemophilia A patients. Similar to EPO, the clinical investigations indicated that a substantial number of patients subjected to the wildtype FVIII develop inhibitory antibodies during the course of treatment. Steinitz et al. used a synthetic peptide library and identified eight T cell epitope clusters.⁴⁵ In a subsequent study, after locating one essential epitopic peptide of FVIII by the in silico approaches, the investigators subsequently located the key residues by running the HLA II binding assay with peptides with single mutations. The deimmunized FVIII mutant was generated accordingly and validated by the T cell functional assay.³³

The mechanistic understanding of peptide-MHC II association was substantiated by large-scale in vitro peptide binding studies. For instance, Wang et al. conducted one of the most comprehensive peptide-MHC II binding screens with a total of 180 libraries, each having one position with a fixed amino acid and the other 19 positions with an equal mixture of all 20 natural residues. The libraries were tested for binding affinity and yielded about 40,000 data points of binding affinity data with 26 HLA II alleles at high frequencies in the global population.⁴⁶ Based on the thorough elucidation of peptide-MHC II binding as well as the accumulation of experimental binding affinity data, many in silico TCE prediction algorithms have been implemented to substitute the in vitro MHC II binding assay for the initial epitope screening. The in vitro binding assay, instead, is often used to validate the candidate peptides obtained from alternative strategies. Nevertheless, it is important to note that binding epitopes are not necessarily “functional epitopes” that elicit immune response subsequently. Therefore, the MHC II binding assay is usually coupled with in vitro T cell assays for the functional validation of antigen-derived peptides.

2.2. Immunopeptidomics

The in vitro binding assay offers an unbiased approach to comprehensively examine the potential TCEs from any given TPs, but it does not account for antigen processing under physiological condition. MHC-associated peptide proteomics (MAPPs) assay is a cell-based in vitro approach to identify the MHC II-restricted peptides that are naturally presented by DCs.⁴⁷ To establish this assay, CD14⁺ monocytes need to be initially isolated from PBMCs and treated by the differentiation cocktail to acquire DCs with high HLA II expression. Upon pulsing by TPs followed by cell lysis, pMHC II complexes can be immunoprecipitated for the identification of the HLA-associated peptidome by mass spectrometry (MS). Primary cells collected from multiple donors are usually required to cover the polymorphism of HLA II.

MAPPs assay has been extensively used in the determination of TCEs derived from antibody-based therapeutics. For instance, the CD4 TCEs of infliximab were revealed by multiple studies involving the MAPPs assay. TCE clusters were identified at several regions, including HCDR2, HCDR3, LFR1 of the light chain of infliximab.^35,37,41 As expected, most of the peptides identified by the MAPPs assay were confirmed as HLA II ligands according to the in vitro data.³⁵ In the work by Karle et al., the immunogenicity of secukinumab (clinical approved anti-interleukin [IL]-17A mAb for the treatment of plaque psoriasis) was evaluated by the MAPPs together with cell proliferation assays. In comparison with five marketed mAbs with various immunogenicity risks, secukinumab developed the lowest density of TCE clusters in the MAPPs assay. Intriguingly, a correlation was found between the levels of peptide presentation and the clinical ADA incidences of all six mAbs, underscoring the predicting power of the MAPPs assay for clinical immunogenicity risk assessment.³⁷ Casotta et al. conducted an in-depth characterization of the Nab against natalizumab, an anti-integrins 4 mAb for multiple sclerosis.³⁹ Through the MAPPs assay with pMHC II purified from patient-derived, natalizumab-specific B cells, a single TCE slightly upstream of LCDR2 was found to dictate the immunogenicity against natalizumab. Intriguingly, this epitope appeared to be engineerable, which prompted the generation of deimmunized variants to be tested in vivo.

To enhance the efficiency and consistency of the MAPPs assay, Lee et al. developed a semi-automated workflow to streamline the sample preparation steps (i.e., cell lysis, immunoprecipitation, washing, and elution). A panel of clinical mAbs was tested by this workflow, and the result indicated that peptide clusters were identifiable from the antibodies with clinical immunogenicity risk. More importantly, some of unveiled peptides aligned well with previous publications from other groups, suggesting the robustness and reliability of this novel format of assay workflow.⁴⁸ Additional investigations also suggested the impacts of other factors such as the selection of DC culture medium, immunoprecipitation beads, and antibodies on the results of the MAPPs assay.^41,49,50

Unlike the in vitro binding assay which relies on synthetic peptides of given lengths, the MAPPs assay as a powerful tool exhibits additional physiological relevance since it reveals only the naturally presented TCEs. This also enables the examination of parameters such as flanking amino acids and post-translational modifications beyond the sequences of binding cores in terms of their impacts to TCE formation. Nonetheless, the peptide detection sensitivity remains the biggest limiting factor of the MAPPs assay especially when molecules with low immunogenic propensities are tested. Factors including, but not limited to, cell culture, pMHC II immunoprecipitation, chromatography isolation, and MS detection need to be optimized to achieve satisfactory immunopeptidomics results.

2.3. Cell-based functional assays

The in vitro functional assays offer solid validation of the immune response induced by antigenic peptides identified from the in silico prediction, in vitro binding screen, and MAPPs assay. Moreover, cell-based experiments can be leveraged to confirm successful deimmunization through epitope elimination.

Within the Td immunogenicity generation pathway, expansion of antigen-specific CD4⁺ T cells follows the pMHC II-TCR engagement (Figure 2). Given so, measuring the T cell proliferation and activation represents an advantageous approach to monitor the immunogen-specific T cell response. The T cell proliferation assay can be executed in the context of PBMCs or with purified DCs and autologous CD4⁺ helper T cells (also termed DC-T cell assay). In the former case, tested compounds are subjected to PBMCs for sufficient time to stimulate T cell proliferation, which reflects the immunogenicity of analytes. For the DC-T cell assay, purified DCs (differentiated from CD14⁺ monocyte population) need to be pulsed by therapeutics prior to the stimulation of autologous CD4⁺ T cells. As indicated by Gokemeijer et al. in a survey, the T cell proliferation assay in either format was performed quite differently in terms of assay parameters such as the donor number, control selection, readout, etc., by researchers from the biopharmaceutical industry.³¹

Numerous examples have underscored the value of the T cell proliferation assay in advancing the developability of antibody-based therapeutics. For instance, the crystallizable fragment (Fc) of therapeutics mAbs can be engineered to accommodate different pharmacological properties, such as reduced antibody-dependent cell-mediated cytotoxicity (ADCC). In the study by Wilkinson et al., ADCC-abolished human IgG1 Fc variants were created, and the peptides carrying the corresponding mutations were scanned for immunogenic neoepitopes by the T cell proliferation assay. Compared to the peptide representing the wild-type Fc sequence, the peptide containing L234/235A mutations, a commonly adopted ADCC depletion strategy in mAb engineering, demonstrated no incremental T cell response as specified by the PBMC proliferation assay.⁴⁰ Steven et al. reported the deimmunization of an anti-human serum albumin (HSA) shark variable domain (VNAR) via the in vitro maturation process. Random mutations were introduced to generate variants from the previously constructed huE06v1.10, a humanized anti-HSA VNAR, and subjected to the DC-T cell assay for immunogenicity evaluation. Consequently, the clone BA11, which exhibited intact target binding capacity and minimal immunogenicity, was established.⁴³

The T cell proliferation assay has been continuously advanced to enhance its robustness in immunogenicity assessment. For instance, the sensitivity of the T cell proliferation assay is somewhat unsatisfactory due to the naturally low abundance of antigen-specific CD4⁺ T cells. In respect of this limitation, Siegel et al. developed an innovative DC-T coculture assay that involves a restimulation step to enhance the assay sensitivity. In detail, following the coculture of CD4⁺ T cells and antigen-pulsed DCs for several days, freshly prepared, antigen-challenged autologous monocytes were added to the coculture for additional T cell stimulation. Through this workflow, the authors successfully reproduced the immunogenic effect provoked by the TP-derived TCE peptides identified in previous publications. Moreover, the deimmunized counterparts of those peptides demonstrated attenuated DC-T assay response, suggesting the robustness of this assay platform.⁵¹ Another study suggested the use of additional T cell activation markers such as CD134 and CD137 to specifically outline the antigen-responsive CD4⁺ T cell population in the context of the PBMC assay. The researchers demonstrated, through testing 14 biologics of various immunogenic potentials, that the proportion of CD134⁺/CD137⁺ population within the CD4⁺ T cells following the in vitro stimulation correlated well with the clinical ADA incidences.⁵² These reports underscored a need for further harmonization of the in vitro assays to more accurately predict the clinical immunogenicity of TPs.

Alongside the T cell proliferation assay, the enzyme-linked immunosorbent spot (ELISpot) assay is also used to visualize antigen-specific T cell activation. Cytokines associated with T cell activation (e.g., IL-2, interferon [IFN]-γ) are often used as readouts of ELISpot assay to examine the Td immunogenicity of TPs. In comparison to enzyme-linked immunosorbent assay (ELISA) which measures the total cytokine levels from the bulk cell cultures, ELISpot is more favorable when cytokine secreting (functional) cell populations need to be quantified. In the study by Spindeldreher et al. mentioned above, the ixekizumab-derived peptides identified from the MAPPs assay were validated by the IFN-γ ELISpot assay using CD4⁺ T cells challenged by antigen-loaded PBMCs.⁵³ PE38 is a bacteria-derived toxin used to construct the recombinant CD22-targeting immunotoxin moxetumomab pasudotox. Due to the non-human origin of PE38, moxetumomab pasudotox demonstrated considerably high immunogenicity in clinic as anticipated,⁵⁴ which prompted the effort to evolve a deimmunized version of PE38. Mazor et al. identified the TCEs of PE38 by ELISpot assay using synthetic PE38-derived peptides. The deimmunized PE38 with TCE depletion was subsequently generated and verified by the abolished T cell immune response.⁴² The incorporation of unnatural amino acids (unAA) in therapeutic peptides has been broadly implemented partially for the interest of immunogenicity management. Azam et al. conducted a proof-of-concept study by introducing unAA into the model antigen, influenza hemagglutinin (HA) peptide HA_{306 − 318}. As confirmed by ELISpot and in vitro peptide binding assays, the modified peptides exhibited compromised T cell priming capacity along with decreased HLA II binding to various extents.³⁴

Notably, either epitopic peptides or the intact TPs can be used as the input of T cell proliferation assay. Peptides undoubtedly provide more of an epitope-centric view, whereas the intact TPs offer more comprehensive insights regarding the overall immunogenicity potential. However, the leading concern of using the intact TPs for cell-based immunogenicity analysis derives from the mechanism of action of the compounds. Most of the immunomodulatory therapeutics naturally convey stimulatory signals to T cells, which confer vigorous T cell activation and proliferation that is hard to differentiate from the immunogenicity-triggered cell expansion. In this case, the DC-T cell assay might be more feasible since it circumvents the direct T cell activation by the tested agents. Otherwise, peptides (or overlapping peptide pools) may serve as the input material of PBMC assay to highlight epitope-specific T cell activation. Rather than exploiting a generic procedure, a fit-for-purpose approach should be planned for each program by integrating a spectrum of in vitro assays. Xue et al. reported the systematic immunogenicity assessment of ATR-107, a highly immunogenic anti-IL-21 receptor (IL-21 R) mAb. In this study, the authors observed the stronger proliferation of T cells when pulsed by ATR-107 rather than control mAb, and identified a light chain CDR2 (LCDR2)-located TCE cluster derived from ATR-107 by the MAPPs assay. More importantly, by running the internalization assay, it was suggested that the immunogenic response of ATR-107 was somewhat attributed to its efficient engagement with the antigen processing and presentation machinery of DCs, which was explained by the high expression of the drug target (IL-21 R) on DCs.³⁶ In another study, researchers compared the T cell expansion, DCs activation and antigen processing of three humanized mAbs (namely mAb1, mAb2 and mAb3) against the same target, and identified the TCEs by the MAPPs assay. The superior T cell proliferation induced by mAb1 was attributed to the activation of APCs. mAb3, which appeared to be immunogenic in 90% of patients, failed to promote T cell proliferation in vitro, likely due to the absence of DC activation. However, the synthetic peptide cluster arising from the MAPPs data elicited robust CD4⁺ T cell responses.⁵⁵ These findings indicate the inadequacy of relying on any single platform to estimate the potential immunogenicity risk of candidate molecules.

3. Scope of BCE mapping methodologies

Unlike TCEs that can be identified via in vitro measures, the humoral immune response is completed by multiple cell types in different tissues, which cannot be easily recreated by in vitro or ex vivo systems. As such, BCE analysis is usually performed through cell-free experiments with ADA samples acquired from animal or human studies. Highly diversified strategies have been established to characterize the BCE profiles of TPs as reported.^56–58 Given the heterogenic nature of ADAs, we specifically discuss the epitope mapping of polyclonal antibodies (pAbs), which could be dramatically more challenging to characterize compared to mAbs. Experimental approaches such as ligand binding assay (LBA), MS, and structure analysis unveil the antibody-antigen binding interface through different principles. The adaptability, expense, as well as the time consumption, of each strategy discussed in this section are listed in Table 2. Each technique offers distinct advantages and limitations, underscoring the necessity for an integrated strategy for explicitly understanding the epitope spectrums of biotherapeutics.

Table 2. The summary and comparison of various strategies applied for pBCEs identification.

Category	Name of methods	Epitope type		Cost	Time	Commercial availability	Ref
		Linear	Conformational
LBA	Domain specificity	No	No	+	+	Yes*	^Citation59–61
	Peptide screening	Yes	No	++	++	Yes	^Citation62–65
	Peptide display	Yes	No	++	++	Yes	^Citation66,Citation67
	DMS	Yes	Yes	+++	+++	No	^Citation68–72
MS	PROTEX	Yes	No	++	++	No	^Citation73,Citation74
	HDX-MS	Yes	Yes	++	++	Yes	^Citation75
	FPOP-MS	Yes	Yes	++	++	Yes	^Citation76
Structure biology	nsEMPEM	Yes	Yes	+++	+++	Yes	^Citation77–80

*Domain specificity analysis can be incorporated into the validated ADA/Nab assay and executed by facilities compliant with regulatory standards, but it is usually not provided as a service independently.

3.1. LBA-based methods

The underlying principle of LBA-based epitope mapping is to delineate the antibodies’ binding preference among the distinct segments or mutants of the antigen (qualitatively or quantitatively). Many academic and industrial laboratories have reported the development of their in-house epitopes mapping assays, and many commercial options are available for investigators with various research objectives and goals. Through several case studies, we elaborate on how LBA-based approaches could help reveal the epitope information of domain, peptide, and residue-level resolution.

3.1.1. Domain specificity characterization

The US Food and Drug Administration recommends the assessment of ADA binding specificity to different clinically relevant domains of MDBs.⁸¹ A generic, stepwise strategy was proposed by Gorovits et al. to meet the requirement of domain specificity characterization of MDBs-associated ADA.¹⁶ As a representative study of ADA domain specificity analysis, Stubenrauch et al. reported the practice of domain detection and competition assay (DDA and DCA, respectively) to examine the domain specificity of ADAs induced by a targeted immunocytokine (TIC) in a non-human primate study.⁵⁹ In brief, three fundamental modules of the TIC, namely antigen binding fragment (Fab), Fc, and cytokine domain, were individually purified and used as either bait (for DDA) or competitor (for DCA) to analyze ADA binding preferences. In accordance with the well-aligned DDA and DCA results, all 12 subjects exhibited ADAs against Fab and/or the cytokine domains, whereas only one subject developed Fc-specific ADAs. Another exemplary study involved the domain specificity analysis of moxetumomab pasudotox, a highly immunogenic CD22-targeted immunotoxin composed of a binding domain (BD) and the toxin PE38. Vainstain et al. established the bridging assay employing each domain as the inhibitor to measure the relative abundances of domain-specific ADAs.⁶⁰ Their method was also validated and applied in the clinical investigations and revealed that 36/67 and 66/67 of Nab-positive subjects demonstrated ADAs against BD and PE38, respectively.⁵⁴

Domain specificity analysis has also been implemented in studies involving multispecific antibodies, which contain two or more target binding domains. Luong et al. developed a cell-based, domain-specific Nab assay to support the clinical development of PF-07257876, a bispecific antibody co-targeting PD-L1 and CD47.⁶¹ Notably, the authors proposed a unique bead-based Nab extraction prior to the assay and observed dramatically decreased false-negative from the non-neutralizing ADAs. More insights could be obtained from the domain specificity of the Nab to better interpret the clinical observations regarding PK, efficacy, and safety.

Unlike the immunoassays based on endpoint measurements, surface plasmon resonance (SPR), which monitors the kinetics of ligand-receptor binding, could be used to quantify the occupation of drug molecule by ADAs as well. Lewis et al. reported an SPR-based strategy to characterize the presence of C2 domain-specific antibody of blood coagulation Factor VIII in patients’ plasma.⁸² Nevertheless, domain specificity analysis represents a rudimentary approach of epitope mapping due to the relatively low epitope resolution it gains. More sophisticated assays should be deployed in a case-specific manner to pinpoint the BCEs that mediate ADA-TP interaction.

3.1.2. Peptide binding methods

Similar to the peptide-based TCE mapping, BCE can be investigated by the in vitro binding assay with synthetic peptide library. The coverages of peptide library range from one single protein to the whole proteome depending on the research purposes.⁸³ The first endeavor of peptide-based epitope mapping was reported by Barteling et al. who dissected the foot-and-mouth disease virus capsid protein VP1 into 208 synthetic peptides. A seven-residue region was identified as the dominant epitope that antibodies were raised against.⁸⁴ Brand and colleagues developed a peptide screening assay to examine the BCEs on human α-galactosidase A (AGAL) recognized by ADAs from Fabry disease patients who failed to possess functional AGAL. As indicated, many patients developed ADAs against the peptides shared by AGAL as well as its homologue α-galactosidase B, suggesting the severe concern in regards of ADA cross-reactivity.⁸⁵

Several studies leveraged peptide screening to characterize the epitope profile of ADAs to high-immunogenicity therapeutic mAbs such as infliximab and adalimumab. By screening 15-mer peptides spanning the full sequence of infliximab, researchers identified six ADA epitopes, of which four were located within the TNF binding interface, supporting the immunogenicity-related efficacy loss of infliximab in clinic.⁶² Epitopes of a similar distribution pattern were later identified by the same group in another study involving adalimumab.⁶³ In addition, the peptide microarrays have been extensively used to explicitly pinpoint polyclonal BCEs (pBCEs) derived from pathogens and food allergens.^86,87

Peptides can also be presented in cell display systems to bypass the chemical synthesis process. Rockberg et al. created a bacteria-displayed library of randomly fragmented cDNA of human epidermal growth factor receptor 2 (HER2) and used it to map the domain specificity of several anti-HER2 pAbs. By sequencing each clone of pAbs-bound bacteria, the authors revealed one to five domains that can be targeted by each pool of pAbs.⁶⁶ Similar strategies based on phase-display system can also be established to unveil the pBCEs induced by pathogens or toxins.^88,89

A major caveat of peptide screening for epitope mapping is its inability to recapitulate the conformational and discontinuous epitopes. Some chemical strategies have been proposed aiming to resemble the conformation epitopes on peptide arrays. Timmerman et al. established a conformational peptide array by chemically crosslinking intra-peptide cysteines that were introduced at specific positions. This technique has been successfully commercialized and leveraged to identify the conformational epitopes of rituximab.⁹⁰ Nevertheless, the extensive complexity of non-linear epitopes makes them extremely difficult to reconstitute by any generic chemical strategies.

A secondary screening against a series of peptides harboring the point mutation can provide residue-level epitope information.⁹¹ In the work by Geyson et al., after pinpointing a single 6-mer peptide as the high-potential linear epitope, an in-depth library encompassing 120 hexapeptides, each containing a single point mutation from the parental peptide, were generated and revealed Leu148 and Leu151 as two essential residues dictating the anti-viral antibody binding capacity.⁸⁴

Despite the notable limitations, the fine mapping of BCE by peptide screening is widely adopted because of the adequate resolution, relatively low cost, and easy interpretation. Integration of additional complementary strategies can be beneficial in bridging the gaps associated with peptide screening.

3.1.3. Mutagenesis screening

Deep mutational scanning (DMS) coupled with cell sorting enables the high-throughput identification of function-determining residues on intact proteins.^92,93 For epitope mapping particularly, a DNA library of the target protein with single AA mutations needs to be created and expressed. The hypothesis here is that the antibodies no longer bind to antigen once the key epitopic residues are missing. The population of cells exhibiting negative antibody binding signal will be isolated and subjected to deep sequencing to subsequently reveal those key residues. As indicated, DMS facilities the discovery of epitopes that could not be identified by a residue-to-alanine library screen.⁶⁸ Notably, DMS contributed considerably to the research related to the recent outbreak of SARS-CoV-2. By screening the yeast-presented mutagenesis library of receptor-binding domain (RBD) of SARS-CoV-2, Greaney et al. identified three key RBD-resident epitopes recognized by neutralizing pAbs. Mutations that occurred within these epitopic regions, to different extents, restored the viral infection by diminishing the capacities of antibody-mediated neutralization.⁶⁹

A divergent way to map the pBCEs is to dissemble the pAbs repertoire into mAbs for individual BCE characterization. Onda et al. characterized a total of 60 mAbs from animals immunized by immunotoxin PE38 and identified seven groups of epitopes through mutual competitive binding and point mutation screening.⁹⁴ Subsequently, the same group developed a series of variants with the removal of BCEs identified in the previous work and confirmed the reduced immunogenicity along with sustained efficacy through the in vivo study.⁹⁵ The success in the proof-of-concept study translated to the development of LMB-100, which is a new immunotoxin generated by linking a mesothelin-targeting Fab with a truncated, BCE-depleted toxin PE24.⁹⁶ LMB-100 is currently undergoing clinical trials for the treatment of mesothelin-expressing tumors (e.g., NCT02798536, NCT02810418). A similar strategy was reported in the previously mentioned study involving natalizumab.³⁹ The key CDR-located residues determining the Nab binding were examined by studying the monoclonal Nabs isolated from two natalizumab-treated patients. Nevertheless, while this strategy provides the highest level of epitope resolution, the considerable workload involved may not be feasible in many circumstances.

3.2. MS-based methods

MS strategies represent another broadly used strategy to deconvolute the high-order structural information of protein complexes. As a specific analyzer of superior sensitivity, MS has been extensively leveraged for biological and pharmaceutical research for decades, and generations of MS-based methodologies have been used to decipher structural information about the antigen-antibody recognition.^97–99 In this section, we delve into how these methodologies facilitate the efforts of mapping pBCEs.

3.2.1. Proteolytic excision/extraction

The proteolytic excision/extraction (PROTEX)-MS strategy relies on MS identification of proteolytic antigenic peptides captured by immobilized antigens. Depending on the order of antigen-antibody complex formation and enzyme-mediated antigen degradation, this approach can be categorized into epitope excision (binding preceding digestion) and epitope extraction (digestion preceding binding).¹⁰⁰ Epitope excision/extraction in combination with MS detection provides valuable epitope information with minimal material input. Some studies have reported its use against pAbs, as surveyed by Opuni et al.¹⁰⁰ Among which, Roth et al. indicated through epitope excision-MS that patients with active antineutrophilic cytoplasmic antibody-associated vessel vasculitis possess differential epitope specificities of anti-myeloperoxidase autoantibodies in comparison to healthy subjects.⁷³ Considering the application of PROTEX-MS in numerous studies across a broad diversity of experimental setups, it is reasonable to include it as a potential strategy to profile the epitopes of ADAs. One potential limitation of PROTEX-MS is its reliance on proteolytic digestion to release peptides with appropriate length to precisely pinpoint the epitope location. Thus, intrinsic factors of the antigens such as size, sequence, and accessibility at native conformation will all affect the output after enzymatic cleavage. Besides, discontinuous epitopes might not be efficiently recovered, especially when the three-dimensional structure of an antigen is unavailable.

3.2.2. Protein footprinting

An advanced MS-based strategy known as protein footprinting has been increasingly used to study protein-protein interaction. In principle, the solvent-exposed area of proteins can be modified (reversibly or irreversibly) by specific reagents, and altered modification profiles (footprints) induced by protein-protein interaction can be examined by MS subsequently. Liu et al. thoroughly reviewed the methodology and applications of MS-based protein footprinting.¹⁰¹

Hydrogen-deuterium exchange coupled with MS (HDX-MS) is widely used to investigate protein-protein interaction and protein conformational changes.¹⁰² The solvent-exposed areas of target proteins are marked by the exchangeability of hydrogen to deuterium. As the HDX efficiency is almost exclusively determined by exposure status in solvent since each AA (except proline) has a single backbone hydrogen, HDX-MS is generally considered an unbiased footprinting approach to outline the high-order protein structure. Unlike some case studies using purified pAbs for analysis,¹⁰³ Stander et al. showed that HDX-MS can be exploited with a nonenriched pool of pAbs generated after immunization, further extending the robustness and adaptability of this strategy.⁷⁵

One major technical hurdle of HDX-MS is the loss of deuterium atoms due to the reversibility of the atom exchange process. To overcome this drawback, many irrepressible labeling methods have been developed by covalently modifying one or several preferred residues. Fast photochemical oxidation of proteins (FPOP) rely on tagging protein with hydroxyl radicals generated from KrF laser-pulsed hydrogen peroxide.¹⁰⁴ A recent proof-of-concept work by Genentech involves FPOP-based ADA epitope mapping of a clinical candidate bispecific antibody.⁷⁶ As reported, the CDRs of both heavy and light chains in both arms could be recognized by ADAs as per the FPOP-MS data. Interestingly, two additional regions on both arms were incrementally exposed in the presence of ADAs, suggesting the possibility of ADAs serving as allosteric modulators of TPs.⁷⁶

3.3. Structural biological approaches

Unveiling the structure of antigen-antibody complex provides atomic level resolution of BCEs. X-ray crystallography and nuclear magnetic resonance (NMR) are extensively used to acquire the structures of antigen-mAb complexes, which offers significant benefit to understand the specificity and mode of actions of mAbs.¹⁰⁵ Cassotta et al. demonstrated the crystal structure of the natalizumab-monoclonal ADA complex,³⁹ but, to the best of our knowledge, no study has reported the successful implementation of either technique to capture the antigen-pAbs structure.

Electron microscopy (EM) is currently the most feasible way to visualize the structure of macromolecular complexes that possess extensive size and heterogeneity, which are the bottlenecks of NMR and X-ray crystallography, respectively.¹⁰⁶ The utilization of EM polyclonal epitope mapping (EMPEM) for pBCE visualization was first reported in 2018.⁷⁷ In this study, IgG purified from the sera of rabbits immunized by an HIV envelop (Env) vaccine was digested to release the Fab, which was subsequently incubated with Env to form Env-pAbs complexes prior to the negative staining EMPEM (nsEMPEM). The authors observed the temporal alteration of pAbs diversity throughout the immunization by examining the staining pattern of samples from various time points. Besides, an unreported class of antibody that binds between the Env blades was also identified at a low frequency. A similar experimental workflow was later used in the epitope mapping study of additional types of viruses and vaccines.^79,80,107 To date, no study has demonstrated the application of nsEMPEM for visualizing TP-ADA complexes. Nevertheless, we remain optimistic for the future implementation of EMPEM in immunogenicity assessment, considering the growing accessibility of this technique.

4. Computational epitope prediction

Drug developers harbor profound curiosity and interest in predicting the immunogenicity risk associated with biotherapeutics. Given the large number of mAb candidates being generated for candidate screening, coupled with the formidable technical barriers and high costs of experimental methodologies, it is not practical to rely on experimental strategies for immunogenicity assessment during early-stage development. With the recent rapid evolution of immunoinformatics, however, many computational platforms have emerged to expedite immunogenicity assessment based on TCE and BCE prediction.^108,109 It should be acknowledged by all developers that the anti-drug immune response is a multi-factor process in which many compound-related (e.g., formulation, aggregation) and patient-related (e.g., immune status, genetic background) attributes play critical roles.¹² Thus, the insights gained from sequence-based computational immunogenicity analysis should be applied with caution for decision-making. The availability and performance of the current TCE and BCE prediction tools are introduced below.

4.1. T cell epitope prediction

In silico T cell prediction is routinely executed by antibody developers to provide fast and high-throughput immunogenicity assessments at earlier stages.³¹ As the CD4⁺ T cell response serves a decisive role triggering anti-drug immune response, extensive efforts have been made toward advancing the computational prediction of peptide epitopes presented by MHC (HLA) II. HLA class II genes are encoded by three loci, namely DR, DP and DQ, and all of them display extreme polymorphism in human population.¹¹⁰ Despite the extensive variability of HLA II, the antigen-derived peptides always use a 9-mer core region to interact with the binding groove of MHC II molecule, although the flanking residues outside the binding groove seem to contribute to the binding somewhat.¹¹¹ The side chains of P1, P4, P6 and P9 face toward the MHC II binding pocket, making these four positions the key “anchor residues” that determines the selectivity and affinity of peptide-MHC II interaction. Each different MHC II molecule presents its unique preference of the 9-mer peptide core, particularly the four anchor sites. The highly restricted mode of peptide-MHC binding makes the sequence-based prediction a feasible approach of immunogenicity risk assessment.

Over 44,000 MHC II ligand binding affinity data have been accumulated by The Immune Epitope Database (IEDB, http://www.iedb.org) to enable model training for algorithm development. A variety of quantitative matrix-driven (e.g., SMM-align¹¹² and machine learning-driven platforms (e.g., NN-align¹¹³ have been developed for the prediction of TCE against selected HLA II alleles with known sequences (refer to the review by Doneva et al. for the comprehensive collection of TCE prediction methods.¹⁰⁹ Ethician was developed by Nielsen et al. on the foundation of NN-align to achieve a more accurate pan-specific HLA-II epitope prediction to bypass the hurdle that only a limited number of HLA alleles have been experimentally investigated for their epitope specificity.^114–116 This method was further improved upon by the inclusion of peptide elution data from MAPPs assays.¹¹⁵ As of September 2023, Ethician 4.1 has been designated as the most recommended method based on the automated benchmarking results by IEDB.

The capability of an algorithm to discriminate immunogenic versus nonimmunogenic peptide is usually reflected by the area under the curve (AUC), which ranges from 0.5 (random prediction) to 1.0 (fully accurate prediction). To date, many MHC II epitope prediction methods can achieve AUC > 0.8 for the preferred alleles, which is mostly satisfactory. All current MHC II epitope prediction tools are based on the peptide-MHC II binding affinity calculated by computational algorithms, which usually leads to overprediction of immunogenicity. Depending on the uptake and processing of antigens, the epitopes that can bind to MHC II in theory may not be functionally presented by the cellular machinery.³⁶ Some methods have been proposed to identify the protease-cleavage sites of target proteins by checking the N- and C-terminal motifs,¹¹⁷ but these approaches failed to show significant improvement of the MHC II-restrict epitope prediction. This is much less ideal compared to MHC I epitope prediction, which greatly benefits from the incorporation of proteasome processing analysis tools like NetChop.¹¹⁸

A subset of T cell epitopes recognized by regulatory T cells, known as Tregitopes, serves inhibitory functions to induce immune tolerance of effector CD4⁺ helper T cells. Tregitopes were first identified on the Fc region of IgG and later expanded to more non-IgG proteins.¹¹⁹ As reported, the overall performance of immunogenicity prediction benefits by including the Tregtitope correction in the context of MHC II binding prediction.^120,121 As such, the incorporation of Tregitopes to induce Treg-mediated immune tolerance represents an alternative strategy for TP deimmunization.¹²²

For antibody-based therapeutics, as the neoepitopes of mAbs are usually found in the CDRs that define the binding affinity and specificity of mAbs, the in silico TCE prediction is generally used for candidate selection rather than protein engineering. However, the value of the in silico TCE identification in drug optimization has been manifested in other modalities. Dulaglutide, an anti-diabetic therapy developed by Eli Lilly, was made by fusing the Fc domain of human IgG4 with a glucagon-like peptide-1 (GLP-1) analog peptide. A strong epitope identified near the C-terminus of the parental peptide was eliminated by a point mutation R36G to create the deimmunized GLP-1 analog.¹²³ Case studies involving FVIII deimmunization also represent examples demonstrating immunogenicity risk mitigation driven by the in silico platforms.^33,124

The robustness of the computational paradigm of MHC II epitope prediction should be evaluated properly to inform the decision-making processes. The correlation relation between in silico assessment and clinical observation has been reported, but it remains doubtful whether the conclusion will be challenged by the increase of samples size. While many sponsors tend to leverage the in silico data to support development actions during the early stage (prospectively), only the minority used it to understand the immunogenicity-related events that have been observed in clinic (retrospectively).³¹ The in silico analysis can be integrated with the in vitro assays to support candidate selection at early stages (as exemplified by Xue et al..³⁶ It should also be noted that the current algorithms are designed and optimized based on amino acid sequence only. The effects of non-protein elements such as post-translational or chemical modifications cannot be evaluated by the available methods.

4.2. B cell epitope prediction

Comparable to TCE, the prediction of BCE can be achieved by a range of methodologies built on various algorithms. Sequence-based platforms like BepiPred scrutinize the chemical and physical properties of amino acids (including factors such as hydrophobicity, polarity, and flexibility) to delineate potential linear epitopes within target antigens.^125,126 Conformational epitopes, which account for approximately 90% of all BCEs,¹²⁷ can be predicted through methods such as DiscoTope and ElliPro, which were trained on the basis of the known antibody-bound regions of proteins.^128,129 The recent advances in deep learning-based tools enable the accurate prediction of high-order protein structures.⁷⁸ In the latest version of DiscoTope (3.0), the power of structure prediction was embedded into the previously establish training models, making it possible to realize the structure-based BCE prediction on a broad spectrum of antigens without solved structures.¹²⁸

Many enhanced in silico methods with advanced predictive robustness have been published recently, and some exhibit superior performances over the methods outlined by IEDB. However, their performance in real-world case studies remains unsatisfactory. Cia et al. presented an independent and rigorous benchmarking study of the most popular BCE prediction methods, and used the SARS-CoV-2 spike protein as an example to evaluated the performance of those methods. None of the methods achieved the desired performance owing to the insufficient quantity of characterized antigen-antibody co-structure for model training.¹³⁰ Furthermore, the current structure collection is largely biased toward pathogens and drug targets that do not represent the naturally existing antigens, especially for TPs, well. In the work by Lin et al. who sought to deimmunize the VHH scaffold, both Ellipro and BepiPred failed to discriminate the immunogenicity potentials of multiple VHH variants.¹³¹ Indeed, given the lack of thorough understanding of the intrinsic natures of antibody-BCE interaction, most BCE prediction platforms offer AUCs lower than 0.6, which is far less desirable compared to TCE analysis. These undesirable practices prompt the necessity of combining experimental and computational approaches for the comprehensive profiling of pBCEs. There is currently no consensus among drug developers regarding the value of BCE prediction so far. Furthermore, a dispute persists regarding whether the presence of BCE is determined solely by protein-intrinsic characteristics.¹³² In other words, it is questionable whether certain regions of proteins, especially TPs, are immunodominant in terms of BCE recognition. Indeed, caution needs to be exercised when using the information acquired from BCE prediction for decision-making purposes.

5. Conclusion and future perspectives

TP developers are expected to evaluate the risk factors for immunogenicity assessment of their TP products as per the guidance by the regulatory authorities.^12,13 Among all risk factors, it has been realized that the immunogenic profile, especially the TCE and BCE landscapes, serves fundamental importance dictating the developability of antibody-based therapeutics. In this review, we presented a roadmap of epitope-driven immunogenicity evaluation by introducing the current collection of experimental and computational methodologies for TCE and BCE identification, and this toolbox is continuously expanding on the foundations of the advances in genomic and proteomic analysis. As each investigation tool has unique advantages and limitations, we hereby present in Figure 3 a recommended schedule indicating when different immunogenicity evaluation strategies may be applied throughout the lifecycle of drug development.

Figure 3. The recommended schedule of actions to be conducted for immunogenicity assessment of TPs. Different computational and experimental strategies can be performed to support the investigation of immunogenicity from early discovery to marketing stages. Note: the in silico BCE analysis is outlined in dashed line due to the lack of solid evidence justifying its value in candidate evaluation.

Evidence suggests that drug development benefits from epitope analysis, especially TCE-focused investigations (experimental and computational). Advanced therapeutics with diminished immunogenicity risks can be constructed via the identification and removal of TCEs as manifested by many examples we discussed. To date, TCE characterization has been recognized by the industry as a standard procedure for nonclinical and clinical immunogenicity assessment since: 1) TCEs play decisive roles governing the immunogenicity propensity of TPs, and 2) methodologies are more accessible and mature to yield reliable results. B cell-focused investigations, on the other hand, are typically carried out at later stages of development due to the availability of ADA samples. Nonetheless, BCEs are linked more closely to the pharmacological and toxicological outcomes, as it mediates the physical interaction between TPs and ADAs. Therefore, the importance and benefits of BCE characterization in drug development should not be underestimated. Overall, the proactive immunogenicity derisking strategy based on TCE removal has been successfully used in early stages of development since they are easier to identify, manipulate, and validate. In contrast, BCE probing appears to be substantially more intricate both experimentally and computationally. Rather than directly informing drug optimization, the BCE information is used more for monitoring and characterizing the immunogenicity-related adverse effects in the clinic. Nonetheless, there is a lack of consensus of the in silico and in vitro immunogenicity risk assessment strategy for TP development. A case-specific, risk-based workflow is required for each program under evaluation.

Despite the regularly implemented approaches described above, the emergence of innovative platforms offers alternative strategies that allow investigators to evaluate the potential immunogenicity of TPs at early stages. Taking advantage of the organoid culture on a microfluid chip system, Goyal et al. reassembled the lymphoid follicle, the key organ where adaptive immunity is formed, by coculturing primary human blood T and B lymphocytes.¹³³ Using this system, the authors observed successful antibody generation upon challenge by the influenza vaccine. Nevertheless, the whole process of adaptive immunity can hardly be recapitulated by the in vitro tissue culture. Given the poor translatability of animal-based immunogenicity investigation to human study, which is largely attributed to the distinct immune repertoire, efforts have been made to develop HLA humanized or immune reconstituted mouse models to predict the immunogenicity at nonclinical stage. Studies have shown the use of HLA transgenic mice to identify the CD4 TCEs, but the discovery is limited to a specific allele.¹³⁴ The immune reconstituted mice, on the other hand, can mimic the personalized immune response by infusing immune cells from individual donors to immune-compromised mice, but real-world evidence is needed to justify the feasibility and reliability of this type of model, leading to the concerns regarding return on investment faced by sponsors.

In summary, the field of immunogenicity research has experienced rapid maturation in recent decades. Innovative technologies are continuously being developed, validated, and commercialized on the foundation of advanced understanding of therapeutic-provoked immunogenicity. Depending on the risk factors, sample accessibility, budget and timeline, developers may consider appropriate strategies to elucidate the immunogenicity potential of therapeutic molecules to create safe and efficacious medicines.

Abbreviations

ADA	=	Anti-drug antibody
mAb	=	Monoclonal antibody
TP	=	Therapeutic protein
FR	=	Framework
CDR	=	Complementarity-determining region
MDB	=	Multi-domain biotherapeutic
VHH	=	Heavy-chain variable domain
Ti	=	T cell-independent
Td	=	T cell-dependent
Tfh	=	Follicular helper T lymphocyte
BCR	=	B cell receptor
TCR	=	T cell receptor
TCE	=	T cell epitope
BCE	=	B cell epitope
IgM	=	immunoglobulin M
PEG	=	Polyethylene glycol
MHC	=	Major histocompatibility complex
HLA	=	Human leukocyte antigen
APC	=	Antigen-presenting cell
DC	=	Dendric cell
Nab	=	Neutralizing antibody
TNF	=	Tumor necrosis factor
HCDR	=	Complementarity-determining region of heavy chain
LCDR	=	Complementarity-determining region of light chain
LFR	=	Framework of light chain
EPO	=	Erythropoietin
PBMC	=	Peripheral blood mononuclear cell
FVIII	=	Factor VIII
HAS	=	Human serum albumin
VNAR	=	Shark variable domain
ELISA	=	Enzyme-linked immunosorbent assay
ELISpot	=	Enzyme-linked immunosorbent spot
MAPPs	=	MHC-associated peptide proteomics
ADCC	=	antibody-dependent cell-mediated cytotoxicity
Fab	=	Antigen binding fragment
unAA	=	Unnatural amino acids
HA	=	Hemagglutinin
Fc	=	Crystallizable fragment
DDA	=	Domain detection assay
DCA	=	Domain competition assay
TIC	=	Targeted immunocytokine
MS	=	Mass spectrometry
IL	=	Interleukin
IFN	=	Interferon
IL-21 R	=	Interleukin-21 receptor
LBA	=	Ligand binding assay
pAb	=	Polyclonal antibody
pBCE	=	Polyclonal B cell epitope
BD	=	Binding domain
SPR	=	Surface plasmon resonance
AGAL	=	α-galactosidase A
HER2	=	Human epidermal growth factor receptor 2
DMS	=	Deep mutational scanning
RBD	=	Receptor-binding domain
HDX	=	Hydrogen-deuterium exchange
PROTEX	=	The proteolytic excision/extraction
FPOP	=	Fast photochemical oxidation of proteins
NMR	=	Nuclear magnetic resonance
EM	=	Electron microscopy
EMPEM	=	Electron microscopy polyclonal epitope mapping
nsEMPEM	=	negative staining Electron microscopy polyclonal epitope mapping
IEDB	=	The Immune Epitope Database
AUC	=	Area under the curve
GLP-1	=	Glucagon-like peptide-1

Disclosure statement

All author(s) are currently employed by Takeda Pharmaceutical Company Limited and may hold stocks of the company.

Additional information

Funding

This work was sponsored by Takeda Pharmaceutical Company Limited for preparation of the paper.