*3.2. DC:CD4+ T Cell Restimulation Assay Precision Assessment and Comparators*

We investigated first the repeatability of the assay by testing the IFN-γ response of donors to KLH and Avastin (same production batch) in multiple assay screens. To this aim, we plotted the SI for KLH and Avastin for all donors, grouped by batches. All the donors analyzed over 24 screens consistently showed high SIs (with a geometric mean of 225 across all screens) upon treatment with KLH (a widely accepted positive control), while SIs obtained with treatment with bevacizumab were distributed around 1 (Figure 2a), suggesting that there was no substantial change in IFN-γ release compared to the blank. Moreover, very few donors (40/607, 6.6%) in this treatment group showed a two-fold SI change or above (our criterion for calling a positive response, see Section 2). Based on these findings, we recommend the use of bevacizumab as a negative comparator in this assay. We used KLH as the technical positive control in our analyses, as highly immunogenic biopharmaceuticals tend not to reach marketing authorization [12].

We used the DC:CD4+ T-cell restimulation assay to investigate 24 biotherapeutics developed by a range of pharmaceutical companies, comprising a broad range of drug formats and targets. Details about the molecules were extracted from the corresponding FDA label [13] and are summarized in Table 1.

However, for most of the labels, important data about the trial were missing, ultimately limiting the interpretability of the reported ADA rates. Moreover, for a number of trials, drugs were administered in combination with radiotherapy, which is known to impact the immune system and the subsequent production of ADA [14]. In other cases, biopharmaceuticals were administered with corticoid pre-treatment to dampen the immune response, which also influences the production of ADA. In this manuscript, data from combination trials were omitted, except for Alemtuzumab, Cetuximab, Daratumumab, Elotuzumab, Sarilumab, and Tocilizumab, which are always co-administered with other drugs.

Results are summarized in Figure 2b,c. Most of the tested biopharmaceuticals elicited low levels of IFN-γ release (alirocumab, avelumab, benralizumab, bevacizumab, brentuximab, certolizumab, cetuximab, durvalumab, evolocumab, galcanezumab, necitumumab, nivolumab, sarilumab, tocilizumab, ustekinumab, vedolizumab). However, we saw stronger T cell responses with alemtuzumab, elotuzumab, pembrolizumab, infliximab, and daratumumab, for which more than 10% of the donors showed a SI statistically significant above 2. Interestingly, antibodies with identical modes of action (i.e., infliximab, adalimumab, and certolizumab all target TNF-α) triggered different T cell responses with regards to IFN-γ production. In addition, when compounds were tested several times in different screens, we observed that SIs and the derived response rates showed a significant variability (Figure 2c). We observed that for adalimumab, for which screens 02 and 06 resulted in 23.3% and 26.7% of positive donors, respectively, whereas it dropped to 0% in screen 07. These discrepancies are seen for pembrolizumab, atezolizumab, and

elotuzumab, as well. One explanation for this observation could be a compound batch effect, as illustrated for adalimumab in Figure 2d.

**Figure 2.** Overview of the stimulation indexes (SI) obtained in the DC:CD4+ T cell restimulation assay. (**a**) Stability of the controls over different assay screens and donor batches. Data were generated for the set of benchmark compounds in single (**b**) or multiple screens (**c**,**d**). SI represents the number of IFN-γ positive cells over baseline. If a datapoint is significantly above the SI threshold of 2, the donor is considered as positive for this condition and appears in red.

**Table 1.** Overview of the test items and their respective clinical ADA rates. Alemtuzumab, cetuximab, daratumumab, elotuzumab, sarilumab and tocilizumab are part of a co-treatment. Therefore, consideration should be taken when looking at the reported ADA rates. The information was extracted from FDA labels [13]. If several clinical ADA rates were reported, studies mentioning a co-treatment were excluded and the mean value for the remaining study outcomes was taken. In many cases, larger deviations may be due to systematic differences in the treated patient populations, as well as different analytical methods.

