For a long time, unwanted immunogenicity was treated – at least operationally – as a property of the molecule. Either a drug provoked an immune response, or it didn’t; either antibodies appeared, or they didn’t. The job of testing was to find out which side of that line a candidate fell on.
As therapeutic modalities have diversified and our knowledge of the immune system evolves, our thinking around immunogenicity has changed from it being something we just have to accept to needing to understand the underlying causes of immunogenicity. Is it a consequence of how a molecule is built, how it’s processed, how it engages immune cells, or of which parts of the immune system it perturbs first?
Immunogenicity is not one thing
Immune responses unfold as a sequence of events, often beginning with innate sensing before adaptive mechanisms become measurable. An innate signal is triggered, or it isn’t. Antigens are processed and presented. T cells may become activated. Antibody responses may follow – sometimes transient, sometimes persistent. Clinical consequences (through e.g. altered pharmacokinetics, loss of efficacy, neutralisation) appear only at the end of this sequence. By then, the opportunity to understand or mitigate the risk is often gone.
Different therapeutic modalities fail at different points along this chain: some activate immune cells directly through their mechanism of action, while others subtly alter antigen processing. Some never generate antibodies at all yet still provoke clinically meaningful immune activation.
Treating all of these outcomes as a single endpoint obscures the underlying biology that matters.
From assays to decisions
At Abzena, we approach non-clinical immunogenicity assessment as a decision process, rather than an assay checklist. The goal is to identify where immune risk arises, how it propagates, and whether it can be mitigated using available design or process controls.
Early risk assessment can start with in silico epitope prediction, where we use tools such as iTope-AI to scan protein sequences across a wide range of HLA alleles. These analyses don’t predict immunogenicity outright, but they are effective at identifying regions of sequence-driven risk and narrowing experimental focus early.
From there, we turn to ex vivo human PBMC assays to ask a more direct question: do human immune cells respond to this molecule? These assays, including our EpiScreen™ 2.0 platform, allow us to measure antigen-specific CD4 T-cell activation across donor-diverse populations – with phenotyping extended beyond CD4 where modality and mechanism require it – capturing diverse ex vivo biological responses.
For some modalities, this isn’t enough. Immune activation doesn’t always depend on antigen recognition and here is where cytokine release assays become critical. They detect innate immune activation and inflammatory signalling that can arise directly from pharmacologic immune engagement, independent of adaptive responses. Importantly, cytokine release assays are highly sensitive to assay configuration. Cell composition, presentation format, and exposure duration can all influence whether pharmacologic immune activation is revealed or masked. As a result, negative findings can only be interpreted with confidence when the assay format reflects the drug’s intended mechanism of immune engagement. Otherwise, a true pharmacologic immune signal may be masked by the assay itself.
When mechanistic clarity is needed, MHC-associated peptide proteomics (MAPPs) provides a direct readout of what peptides are actually processed and presented on MHC molecules. Not just what’s predicted, but what the immune system truly sees. MAPPs doesn’t capture immune activation directly, nor does it account for downstream signalling or tolerance mechanisms. Instead, its value lies in resolving how sequence or structural changes alter antigen presentation, rather than in ranking overall immunogenic risk in isolation.
Together, these approaches link structure to processing, processing to activation, and activation to risk.
Same immune system, different failure modes
The immune system is conserved, but the ways different therapies engage it are not.
Monoclonal antibodies and engineered proteins
For conventional antibodies and engineered protein therapeutics, immunogenicity most often emerges from sequence variation, post-translational modifications, aggregation, or formulation effects interacting with patient HLA diversity.
These risks are typically mediated through CD4 T-cell help. In this setting, epitope prediction combined with ex vivo PBMC assays provides the most informative view of risk, with MAPPs used when deeper insight into antigen presentation is required.
ADCs: where toxicity complicates immunity
Antibody-drug conjugates (ADCs) add another layer of complexity. For these formats, immune risk may arise not just from the antibody but also from linker chemistry, payloads, impurities, or degradation products. At the same time, the cytotoxic nature of multiple payloads complicates immune assays since extended incubation can eliminate antigen-presenting cells or T cells before meaningful immune readouts are captured.
In practice, cytokine release assays can be informative when applied with short exposure times. DC:T-cell assays and MAPPs are theoretically applicable under constrained conditions, but interpretation requires caution and data can remain limited. For these modalities, understanding the limitations of running these assays with the ADCs and considering alternative options, such as dummy payloads, is recommended.
Immune engagement by design: T-cell engagers
T-cell engagers and immune agonists often fail differently. They’re designed to activate immune cells directly and their dominant liability is often systemic immune activation rather than classical antigen-driven immunogenicity.
Here, cytokine release assays are the front-line tool for assessing non-clinical immune risk from cytokine storm. Assay configuration matters, and versions using soluble and/or immobilized sample are available to evaluate cytokine secretion risks.
Antibody–oligonucleotide conjugates
Antibody oligonucleotide conjugates (AOCs) sit at the intersection of protein and nucleic acid biology. They can engage innate pathways, drive cytokine production such as IFN-γ, and still generate adaptive responses. As a result, they often benefit from a combined approach, pairing cytokine release assays with T-cell activation and proliferation assays to capture multiple immune liabilities.
Cell and gene therapies
Gene and vector-based therapies introduce yet another immune context. Here, immunogenicity risk can arise from vector components, transgene products, or from antigen processing routes that differ fundamentally from those of therapeutic proteins. Cross-presentation can become relevant, and with it the potential for CD8-mediated immune responses alongside CD4 help. In these settings, non-clinical immunogenicity assessment requires readouts capable of capturing both arms of adaptive immunity where appropriate – for example, PBMC-based assays extended to include CD8⁺ T-cell activation and MHC-associated peptide proteomics configured to assess class I as well as class II presentation – and an understanding of how delivery and intracellular processing shape immune recognition.
In these modalities, immune responses are further influenced by delivery route, cellular tropism, and pre-existing immunity to vector components. This means non-clinical assessment extends beyond the intended therapeutic product to include how vector delivery, tissue targeting, and intracellular expression shape immune recognition in ways that differ from protein biologics.
When assay format became the risk: TGN1412
The TGN1412 incident is often cited as a failure of safety testing, but it more accurately reflects a mismatch between the biological mechanism of immune activation and the assay formats used to evaluate risk (1). TGN1412 was a CD28 superagonist antibody designed to activate T cells. In preclinical testing, nothing appeared out of place: animal studies were unremarkable, standard in vitro assays showed no alarming immune activation, and the molecule progressed. But in humans, the response was immediate and catastrophic.
What went wrong wasn’t a missing assay, but an assumption – the assumption that immune activation would only occur through classical antigen recognition and could therefore be detected using conventional formats.
Subsequent investigations showed that assay configuration was the determining factor. When TGN1412 was presented in a way that mimicked localized receptor engagement, immobilised rather than soluble, human PBMCs released large quantities of pro-inflammatory cytokines. The biology that mattered had been present all along, but the assay had failed to reveal it.
The lesson here is that we see how immune risk can arise from pharmacologic activation of T cells, independent of antigen processing or peptide–MHC presentation, and that cytokine assays can detect this, provided you use an appropriate assay format that reflects the biology.
For immune-engaging modalities, immunogenicity assessment, therefore, cannot be separated from an understanding of mechanism and context. Format matters. Exposure matters. And negative data are only meaningful if the biology has been modelled appropriately.
What the immune system actually sees: antigen presentation mapping
Not all immune risk is driven by overt immune activation – sometimes it’s quieter, and harder to predict. Small changes in protein sequence can alter how antigens are processed and presented, even when overall structure and function appear unchanged. These effects are rarely obvious from sequence analysis alone.
The Vatreptacog Alfa case illustrates this clearly. Vatreptacog Alfa differed from native factor VIIa by only three amino acids. Functionally, it behaved as intended – clinically, it did not. Unexpected immunogenicity emerged, and the program was discontinued.
In an article co-authored with FDA scientists, MAPPs was used to examine what peptides were actually presented on MHC molecules (2). The results showed increased presentation of peptides derived from the modified regions of the protein, alongside stronger T-cell responses compared to the wild-type molecule. When de-immunised variants were tested, T-cell activation was reduced.
The value of MAPPs here was seen in providing an explanation beyond just prediction. By revealing which peptides were processed and presented, the assay linked sequence modification to immune outcome in a way that neither in silico tools nor functional assays alone could.
This is where antigen presentation mapping fits. Yes, it’s a screening tool, but it’s also a way to resolve uncertainty when immune risk is real but poorly understood.
Generic peptides: when pre-clinical immunogenicity testing replaces the clinical trial
In some cases, immunogenicity assessment is about replacing an entire stage of development. For certain synthetic peptide drugs, FDA guidance allows developers to avoid clinical trials if they can demonstrate that their product is immunogenically comparable to the reference listed drug.
This shifts the focus from the peptide itself onto what else comes with it. Process-related impurities, synthesis variants, and degradation products can introduce new immune liabilities, even when the active sequence is unchanged. Regulators, therefore, expect evidence that these differences do not increase either innate or adaptive immune activation.
In this setting, non-clinical immunogenicity data become decisive. At Abzena, these programs are supported through expanded innate immune assessment using cytokine release screening, alongside adaptive immune assessment using PBMC assays. Importantly, testing is performed comparatively, evaluating the candidate product alongside the reference drug to determine whether any new immune signals emerge.
In these cases, immunogenicity assessment is for regulatory purposes and the output becomes about building a sound justification that a clinical study is not required.
Closing the loop: right assay, right model, right question
Immunogenicity assessment works best when it’s treated as a way to interrogate immune biology instead of a box to be checked. That requires choosing assays that reflect how a therapy actually engages the immune system, accepting that different modalities demand different questions, understanding where those assays are informative and where they are constrained, and building a rationale that explains risk rather than simply reporting it. ‘Good’ results begin to form a coherent story, one that connects structure, mechanism, immune activation, and consequence.
Throughout this series, one theme has surfaced repeatedly: progress in drug development doesn’t come from adding complexity for its own sake. It comes from matching the question to the biology, and the biology to the stage of development.
Part 1 focused on timing. Potency and functional bioassays only deliver value when they are phase-appropriate and aligned with mechanism of action. Too much complexity too early obscures decisions. Too little rigor too late creates risk.
Part 2 extended that logic into biological context. As mechanisms became more complex, particularly for ADCs and immune-modulating biologics, simple two-dimensional systems stopped being sufficient. More representative in vitro models were needed to understand how drugs behave in settings that resemble human disease.
Part 3 showed how these principles play out under regulatory constraints. For synthetic peptide drugs, immunogenicity assessment is not exploratory but prescriptive, with defined FDA expectations and clear consequences. In this setting, non-clinical immune data can replace clinical trials entirely, illustrating how fit-for-purpose immunogenicity assessment directly shapes development pathways.
Immunogenicity assessment follows the same logic. Immune risk cannot be reduced to a single measurement, because immunology is context dependent.
Across potency, efficacy, cellular complexity, and immune risk, the principle is the same: the right assay, in the right model, at the right time, asking the right biological question. This is how Abzena approaches design and developability – as a connected strategy that links molecular design, mechanism of action, immune activation, and downstream consequence, designed to support confident progression from early discovery through development.
References
1. H. Attarwala, TGN1412: From Discovery to Disaster. J Young Pharm 2, 332–336 (2010).
2. W. Jankowski, C. Kidchob, C. Bunce, E. Cloake, R. Resende, Z. E. Sauna, The MHC Associated Peptide Proteomics assay is a useful tool for the non-clinical assessment of immunogenicity. Front. Immunol. 14, 1271120 (2023).
Authors of Part 4:
-Dr Erika Kovacs, Sr. Director, Bioassay
-Dr Edward Cloake, Director of Immunology
-Beverley Campbell, Sr. Manager, Bioassay
Access Parts 1-3 of our Bioassay Blog Series Here:
Part 1 – Right Bioassay, Right Time: Successful Drug Development – Part 1 | Abzena
Part 2 – Modelling Complex Cellular Environments – Part 2 | Abzena
Part 3 – Part 3 of 4: Immunogenicity Assessment for Peptide Drugs | Abzena
