A Novel Approach for Quantifying the Pharmacological Activity of T-Cell Engagers Utilizing In Vitro Time Course Experiments and Streamlined Data Analysis

CD3-bispecific antibodies are a new class of immunotherapeutic drugs against cancer. The pharmacological activity of CD3-bispecifics is typically assessed through in vitro assays of cancer cell lines co-cultured with human peripheral blood mononuclear cells (PBMCs). Assay results depend on experimental conditions such as incubation time and the effector-to-target cell ratio, which can hinder robust quantification of pharmacological activity. In order to overcome these limitations, we developed a new, holistic approach for quantification of the in vitro dose–response relationship. Our experimental design integrates a time-independent analysis of the dose–response across different time points as an alternative to the static, “snap-shot” analysis based on a single time point commonly used in dose–response assays. We show that the potency values derived from static in vitro experiments depend on the incubation time, which leads to inconsistent results across multiple assays and compounds. We compared the potency values from the time-independent analysis with a model-based approach. We find comparably accurate potency estimates from the model-based and time-independent analyses and that the time-independent analysis provides a robust quantification of pharmacological activity. This approach may allow for an improved head-to-head comparison of different compounds and test systems and may prove useful for supporting first-in-human dose selection. Supplementary Information The online version contains supplementary material available at 10.1208/s12248-021-00637-2.


Cell lines, expression vector and test items
Human cancer cell lines CX1 (AAC129) and MKN45 (AAC409) were purchased from DSMZ, and HEK 293T cells (CRL-3216) were purchased from ATCC. Identity authentication was performed on MKN45 by Microsynth AG in 2018. No authentication was performed on CX1 or HEK 293T cells.

S1.2 Lentivirus production and transduction of target cell line
Full-length cDNA encoding FolR1 was cloned into a lentiviral transfer vector and used to transfect HEK 293T cells along with packaging constructs pRSV-Rev, pCgpV, and pCMV-VSV-G by using the X-tremeGENE 9 DNA Transfection Reagent according to the manufacturer's protocol.
The virus-like particles were harvested 3 days after transfection. The supernatant was centrifuged to remove any cellular debris. Assembled lentiviral particles were isolated by filtration with a 0.22 µm Millex-GP filter and stored at -80°C.
Transduction of HEK 293T cells was performed in 48-well flat bottom plates that were pre-seeded with 1x10 3 viable cells the day before. The culture medium was replaced with 320µL purified supernatant containing the virus-like particles together with 80µL fresh culture medium (DMEM + 10% FCS) and 4 μg/mL hexadimethrine bromide.
The cells were spinoculated for 2h (800g, 32°C) in aerosol tight centrifugation buckets in order to increase transduction efficiency. The cells were exposed to the lentiviral vectors for another 16h at standard culture conditions (37°C, 5% CO 2 ), followed by washing and addition of fresh medium. Two days after transduction, 1 µg/mL Puromycin was added. After initial selection, the cells were sorted for low (designated FolR1 low ) and high (designated FolR1 high ) surface expression of FolR1 by using the BD FACS ARIAIII Cell Sorter (BD Bioscience). The sorted cells were cultured to establish stable clones and to guarantee monoclonality. The FolR1 expression density was confirmed and tested for stability by performing flow cytometric analysis with an APC-conjugated anti-FolR1 antibody over a period of 4 weeks.

Supplemental Section S2 Automated dose-response analysis with Python
A semi-automated workflow has been developed in Python in order to perform time-independent analyses. The AUCE calculations, curve fitting, simulations, and data plotting were conducted using several libraries including scikit-learn, scipy, numpy and matplotlib. The curve fit function uses nonlinear least squares optimization. Outputs are the optimal values for the parameters, so that the sum of the squared residuals of model output and data as well as the variance-covariance matrix are minimal. The parameter estimates were provided along with their RSE% values. These can be used to flag any identifiability issues with the model parameters. Not all the profiles will fit appropriately with the models explored in this analysis. Nevertheless, the automated analysis provides all the necessary information to investigate whether the model fits and the parameters are reliable. The outputs include plots of raw data per experimental conditions (denoted by the readout and experimental condition; e.g., 'CD8_CD25 over time cibisatamab CX1.jpeg'), EC 50

Supplemental Section S3 Calculating quasi-equilibrium trimer concentration
The quasi-equilibrium equations for calculating the trimer concentration in function of drug concentration, binding affinities and target availability have been derived by Schropp and colleagues. In short, the following equations are required to calculate the trimer concentration at quasi-equilibrium (Eq. S1-S7). KD 1 is the binding affinity between the drug (C) and the free tumor target (R). KD 2 is the binding affinity between C and free CD3 (CD3). Quasi-Equilibrium calculations (Eq. S1-S7): (Eq. S1) Equations S1-S3 are required to calculate the concentration of free R and CD3 (Eq. S4-S6): The trimer concentration can then be calculated based on the concentration of free drug (C), free R and free CD3, and the respective binding affinities. The explicit equation for trimer concentration at quasi-equilibrium becomes: The quasi-equilibrium model above assumes rapid binding and a constant receptor pool (R tot (t) = R tot (0), CD3 tot (t) = CD3 tot (0)). A Python script is provided in the GitHub repository    Figure S1. Flowchart of the automated workflow developed for dynamic PK/PD analysis in Python. The program reads a user-provided data sheet then plots each readout or condition over time for all reported drug concentrations. The cumulative AUCE is computed for each readout and for all reported drug concentrations. The AUCE dose-responses are then used to fit a sigmoidal model. A hockey-stick model is simultaneously fitted to determine a threshold concentration in order to attempt to provide a usable pharmacology metric in case a sigmoidal model cannot be fitted. Finally, the dynamic potency or threshold concentrations are reported in .txt format for each readout and experimental condition.