A mechanistic PK/PD model to predict cytokine release and tumor cell killing associated to T-cell bispecific therapies
Apolline Lefèvre (1), Zinnia P Parra-Guillen (1), Iñaki F. Trocóniz (1),(2),(3), Christophe Boetsch (4), Nicolas Frances (4)
(1) Pharmacometrics & Systems Pharmacology, Department of Pharmaceutical Technology and Chemistry, School of Pharmacy and Nutrition, University of Navarra, Pamplona, Spain; (2) IdiSNA, Navarra Institute for Health Research, Spain; (3) Institute of Data Science and Artificial Intelligence, DATAI, University of Navarra, Pamplona, Spain, (4) Roche Pharma Research and Early Development pRED, Pharmaceutical Sciences PS, Roche Innovation Center Basel, Switzerland
Objectives:
T-cell bispecifics (TCB) are promising anti-tumor therapies with remaining challenges associated with their clinical safety profiles. In particular, Cytokine Release Syndrome (CRS) is one of the most common adverse events following TCB dosing and is expected to be associated with the levels of cytokine released [1]. On the other hand, it appears that tumor cell killing can still happen in absence of cytokine release, at least during subsequent dosing of the first dose [2]. In a recent study [2], it was observed that no or little cytokine release was occuring while efficacy (tumor cell killing) was maintained when a second TCB dose was administered close to the first one (1-7 days). Additionally, the cytokine release after the second TCB dose increased with the dosing time interval, up to the first dose levels when the two doses were 28 days apart.
Methods:
A literature review was conducted to identify the literature cytokine and tumor growth longitudinal data as well as existing mathematical models in this context. The models, including the ones from Hosseini et al [3], Chen et al. [4] as well as Jiang et al. [5] were explored and a mechanistic model describing cytokine release and efficacy upon repeated dosing of TCBs was proposed, harmonizing previous models and integrating current state-of-art. The model was developed based on cytokine data published by Li et al. [2]. The data was digitized directly from Figure 1.A which includes cytokine levels in-vivo in the context of repeated TCB administration for different dosing intervals (from 1 to 28 days). The model development was conducted in R version 4.1.3 using the RxODE package.
Results:
In the proposed model, the diminution in cytokine release upon repeating dosing was associated with a desensitization of T-cells (previously not considered). The model contains three T-cell populations with the following sequence of (1) baseline T-cells that will be activated upon TCB exposure leading to (2) the activated T-cells that are able to produce cytokine and kill tumor cells. The activated T-cells rapidly shift into a (3) desensitized T-cell status, which correspond to T-cells that retain their ability to kill tumor cells without releasing cytokines. Cytokine release was modeled as a function of both the drug concentration and the number of activated T-cells. The transition from the resting baseline T-cells to the activated state was modeled with five transit compartments to account for the delay between drug administration and cytokine peak. The replenishment of the baseline T-cell pool was parameterized to enable a return to first dose cytokine levels upon second dosing at around 28 days. Tumor cell killing was modeled as a function of the drug concentration and the effector-to-target (E:T) ratio, so that below a certain E:T threshold, no killing could be induced. In addition, an autocrine loop for T-cell proliferation has been included, supposed to be driven by IL-2, in order to induce activated T-cells proliferation enabling reaching the needed E:T ratio upon TCB activation to trigger killing. Nevertheless, little data were available to estimate parameters (E:T ratio for maximum efficacy mainly) from this part of the model.
Conclusions:
The proposed PK/PD mechanistic model has been successfully developed accounting for the cytokine release profile and target elimination as a consequence of the dynamic of the T-cell phenotype following TCB administration. Simulations were confronted with literature data observing adequate behavior of the model. Indeed, it was able to describe the loss of cytokine release without a loss of efficacy upon the second administration. This hypothesis should be further investigated with more experimental data, involving mRNA sequencing, immuno-phenotyping or other biomarker exploration. In this regard, the developed model could be used to guide the design of new experiments and identify relevant biomarkers. Finally, the model could be further refined to display the loss of efficacy that is usually observed over time [6] by including a fourth T-cell state. This type of model could be used in a drug development setting to optimize the dosing protocol (including step-up dosing) of any TCBs to maximize the patient’s benefit while minimizing the risk of non manageable CRS.
References:
[1] Saber et al., An FDA oncology analysis of CD3 bispecific constructs and first-in-human dose selection, Regulatory Toxicology and Pharmacology (2017)
[2] Li et al., CD3 bispecific antibody–induced cytokine release is dispensable for cytotoxic T cell activity, Sci. Transl. Med. 11(2019)
[3] Hosseini et al., Mitigating the risk of cytokine release syndrome in a Phase I trial of CD20/CD3 bispecific antibody mosunetuzumab in NHL: impact of translational system modeling. NPJ Syst Biol Appl. (2020)
[4] Chen et al., A Modeling Framework to Characterize Cytokine Release upon T-Cell-Engaging Bispecific Antibody Treatment: Methodology and Opportunities. Clin Transl Sci. (2019)
[5] Jiang et al., Development of a minimal physiologically-based pharmacokinetic/pharmacodynamic model to characterize target cell depletion and cytokine release for T cell-redirecting bispecific agents in humans. Eur J Pharm Sci. (2020)
[6] Sánchez et al., Preclinical InVivo Data Integrated in a Modeling Network Informs a Refined Clinical Strategy for a CD3 T-Cell Bispecific in Combination with Anti-PD-L1, The AAPS Journal (2022)