**4. Discussion**

The multifactorial nature of immunogenicity requires that an integrated preclinical risk assessment should be a key element of biotherapeutics development. As T cell-dependent responses are major drivers of immunogenicity, in vitro T cell assays are frequently used as tools to identify and measure CD4+ T cell-dependent responses to biotherapeutics. The DC:CD4+ T cell restimulation assay described here assesses the propensity of a biotherapeutic to trigger a CD4+ T cell response that may result in B cell activation and ADA production. This assay plays a key role in our integrated approach to therapeutic protein immunogenicity risk estimation, which could accelerate drug development.

While a number of assays probing T cell activation in the context of immunogenicity have been published in recent years [6], we believe that the DC:CD4+ T cell restimulation assay described here provides a more comprehensive insight into the role of dendritic cells (taken here as the archetypal APC) in the context of their activation of T cells [17]. The immune response follows a three-signal rule for activation (TCR:MHC/peptide interactions, costimulatory interactions such as via CD28, and cytokine production); the assay published herein captures the interplay of all three signals [18]. In addition, the number of preexisting T cells specific to biotherapeutics is very low, ranging between 1 out of 108 (e.g., trastuzumab and etanercept) and 1 out of 10<sup>7</sup> T cells (e.g., rituximab) [9], but the assay format of the DC:CD4+ T cell restimulation assay allows screening of more CD4+ T cells than in a classical PBMC-based assay. We also believe that the re-stimulation step increases the likelihood of capturing a sustained T cell response [19].

A key part of validating the DC:CD4+ T cell restimulation assay was assessing repeatability and reliability, and its potential to categorize biotherapeutics according to their risk of inducing an unwanted immune response in the clinic. To this end, we carried out a detailed analysis of 24 biotherapeutics with various levels of clinically-detected ADA rates as a proxy for immunogenicity risk. Our robust assessment comprised several levels of repetitions, including repeated assessment of some compounds in multiple assay screens, to provide insights into both the potential and inevitable limitations of the assay. Variance component analysis showed that the primary factors governing the experimental setup, i.e., the screen and the batched donor processing within each screen, did not have a major systematic impact on the readouts. Notable variability arose, however, when compounds were re-tested in another screen, presumably related to subtle variation in compound preparation or the used production batches. Nonetheless, the compound batch effect is not specific to this assay. It is likely that the handling and storage of the sample plays a role here, influencing post-translational modifications and aggregation. Additionally, non-product related factors (e.g., DNA and host cell protein contaminations) have an impact on the risk of immunogenicity and could also influence the assay readout. General donor specific inducibility and the donor specific response to individual compounds also explained parts of the signal variability. However, there was a rather substantial residual unexplained variance, which should caution the user with regard to overinterpretation of individual readouts. Notably, quantification of SI changes in a strict sense were not directly informative, as even clinically-tested compounds with low immunogenic risks (for example, bevacizumab) resulted in a few positive readouts. We presume that additional insights might be gained by fundamental and costly changes of the lab protocol, i.e., performing replicated measurements in different experimental batches for every condition. In our

experience, donor cohorts of 30 individuals per screen offer a reasonable tradeoff between the cost, timelines, and statistical power of the assay. While not specifically discussed in the manuscript, we found it important to test all compounds using a panel of donors that showed an HLA-DRB1 allele frequency that broadly reflects the world population. It has been demonstrated that certain HLA alleles were associated with an increased immunogenic response towards certain biopharmaceuticals [20–25]. Nevertheless, in the context of use of this preclinical assay carried out in 30 donors, we primarily investigated whether compounds may be at risk of inducing an enhanced immunogenic response in a general population. As each screen usually comprised different sets of donors, an arbitrary selection of pre-typed donors with respect to their allelic HLA-DRB1 composition enabled a higher comparability of the data in the long term.

While tempting, it is problematic to compare SI values measured in a DC:CD4+ T cell restimulation assay with actual ADA rates in the clinic, although it is one of the few available measures directly related to clinical immunogenicity. Assays used to measure ADA in clinics are based on different methodologies sensitive to sample handling, the timing of sample collection, concomitant medications utilized in the study, and the underlying nature of the treated disease [26]. Furthermore, while we believe we have used the most recent information available on the FDA database, most labels may not be updated on a regular basis: in a recent review, Borrega et al. showed that 57% (39/69) of the biological drugs authorized before 2012 did not have updated summaries of product characteristics, especially in the immunogenicity section [27]. In our study, we collected the ADA rates of the 24 assessed marketed compounds as a starting point to build a database to benchmark newly developed immunogenicity estimation methods and to have a retrospective and comprehensive overview of the immunogenicity of marketed antibodies. We used the available data to create two categories of compounds, at high (≥20% reported ADA rate) and low (<20% ADA rate) risk for immunogenicity, on which we calibrated the assay's linear mixed model. Thus, this binary high/low risk paradigm is the most reasonable for implementation in preclinical risk evaluation for therapeutics. To facilitate this process, we found it essential to add in the panel a few standard compounds (at minimum, a negative control, such as bevacizumab, and a positive control, such as KLH; any additional comparators also provide useful comparisons) to help set precise boundaries of low and high risk of immunogenicity while mitigating intrinsic donor variability. Accordingly, in this context of use, one of the most useful applications of the DC: CD4+ T cell restimulation assay is to provide a relative ranking for compounds with similar amino acid sequences and mode of action, or compounds that have different formulation or have been produced in different batches.

While assays measuring T cell activation in response to novel biopharmaceuticals are not yet required by regulatory agencies, there is added value in presenting the results of such assays as part of the risk assessment submitted in the Integrated Summary of Immunogenicity [28]. A current challenge is that none of the published assays is considered to be fully validated. We here propose a new assay format that captures the interaction between DCs and CD4+ T cells by monitoring the production of IFN-γ by CD4+ T cells in response to biotherapeutics processed by DCs. We tested the predictive power of this assay vs. clinical ADA rate by assessing 24 marketed antibodies, which resulted in 83% accuracy. Predicting the actual rate of ADA-positive patients in a clinical setting with a single in vitro assay is unlikely to be possible, given the myriad contributing factors. However, the DC:CD4+ T cell restimulation assay can help flag potentially immunogenic biopharmaceuticals in preclinical drug development, allowing for selection or de-immunization before a clinical trial starts, improving both patient safety and the cost of pharmaceuticals. Implementation of this assay as part of a comprehensive risk assessment has the potential to provide a more robust and informative immunogenicity risk assessment in early drug development.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/pharmaceutics14122672/s1, Figure S1: Heatmap showing the relative distribution of the HLA-DRB1 supertype frequencies among the donor cohorts per screen of the DC:CD4+ T cell restimulation assay.

**Author Contributions:** Conceptualization, M.S., G.S., L.C.F., K.B.-F., T.P.H., N.S. and C.M.-D.; Formal analysis, M.S., G.S., L.C.F., A.D. and C.M.-D.; Investigation, M.S., G.S., L.C.F., F.C., J.H., K.H., C.M.L., N.S. and C.M.-D.; Writing—original draft, M.S., G.S., L.C.F., A.D. and C.M.-D.; Writing—review & editing, F.C., J.H., K.H., C.M.L., K.B.-F., O.R., T.P.H. and N.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the local Research Ethics Committee (reference number: 21/LO/0474).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Data are contained within the article or Supplementary Materials.

**Conflicts of Interest:** M.S., G.S., L.C.F., K.H., C.M.L., A.D., K.B.F., T.P.H. and C.M.D. are employees of F. Hoffmann. La Roche (Switzerland); F.C., J.H. and N.S. are employees of Lonza Biologics (United Kingdom). The company had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.
