Paleja Bhairav1, Seshasai Pallikonda Chakravarthy1, Pratik Mondal1, Goutam Nair1, Mrittika Roy1, Swagatika Sahoo1, Dinesh Bedathuru2
1Vantage Research Pvt Ltd, 2Vantage Research Inc
Background: CD3 bispecific antibodies (CD3-BsAbs) engage T cells to selectively recognize and eliminate cancer cells. However, their potent immune activation can also trigger cytokine release syndrome (CRS), primarily due to on-target, off-tumor toxicity. Achieving an optimal therapeutic window for CD3-BsAbs requires a quantitative understanding of their mechanism of action (MoA), particularly the formation of CD3-BsAb–HLA-G trimers in both tumor and healthy tissues expressing the target antigen. Mechanistic modeling, guided by experimental data, enables the characterization of key pharmacodynamic drivers—most notably, CD3 and HLA-G receptor occupancy—which are critical determinants of both antitumor efficacy and CRS risk. Objectives: The CD3 bispecifics quantitative systems pharmacology (QSP) platform model is designed to identify key physiological drivers of both efficacy and adverse events, with the goal of optimizing clinical dosing while minimizing the risk of cytokine release syndrome (CRS). In previous work, we applied this platform to support translational analysis for Xaluritamig, a STEAP-targeted T-cell dependent bispecific (TDB) for prostate cancer, recommending an alternative dosing regimen and incremental dose strategy [1]. In the current study, we extend the application of this platform to integrate the totality of evidence on both tumor killing and cytokine release, leveraging in vitro data across multiple cell lines and effector-to-target (E:T) ratios, xenograft preclinical data, and clinical data for JNJ-78306358. This work will focus on: (1) to derive a clinically efficacious dose based on mouse tumor growth inhibition (TGI) studies, and (2) to refine dose priming strategies through mechanistic modeling of CRS. Methods: We developed a quantitative systems pharmacology (QSP) platform model for JNJ-78306358, a CD3 × HLA-G bispecific T-cell engager under investigation for the treatment of ovarian, renal, and colon cancers. Building on the foundational work by Hosseini et al. (2020) [2], the model characterizes drug interactions with CD3 and tumor-associated antigens (TAA), incorporating detailed representations of T-cell activation, margination, and trafficking dynamics. The model predicts the fraction of activated T cells and drug pharmacokinetics within both the tumor microenvironment and healthy tissues—factors identified as key drivers of antitumor efficacy—while also quantifying on-target, off-tumor engagement that contributes to cytokine release. The original study emphasized the importance of activated T-cell dynamics and drug distribution in determining both therapeutic benefit and off-tumor toxicity, offering valuable insights into the clinical translation of CD3 bispecifics in Non-Hodgkin’s lymphoma [2]. Building on the foundational framework established by Hosseini et al. [2], we adapted the model for HLA-G–expressing advanced solid tumors to optimize efficacy and refine dose priming strategies for CRS mitigation. The platform was calibrated using in vitro cytotoxicity and xenograft mouse efficacy data for JNJ-78306358 [3], and subsequently translated to determine the first-in-human (FIH) starting dose. Early clinical cytokine data from initial patient cohorts were then integrated into the model for further calibration, enabling refinement of priming and step-up dosing regimens tailored to mitigate CRS risk while maintaining antitumor activity. Results: The QSP model successfully integrated in vitro and xenograft mouse data to enable efficacy translation and incorporated early clinical data to inform step-up dosing strategies in Phase 1 trials. The model-derived MABEL-based first-in-human (FIH) dose was retrospectively validated against the clinical trial’s starting dose, demonstrating strong concordance. Specifically, the FIH clinical starting dose for JNJ-78306358 was 46 µg [3], while the model-predicted MABEL-based dose was 110 µg – both within the same order of magnitude, supporting the model’s translational relevance and predictive accuracy. The model also successfully captured the totality of evidence across in vitro, xenograft, and clinical datasets – including cytokine excursions observed in 2-step and 3-step dosing regimens in cohorts 4-7 of the JNJ-78306358 trial [3] – with reasonable variability, demonstrating its robustness and applicability across translational stages. Conclusions: The model enables analysis of antigen expression levels, binding affinities, and baseline T-cell concentrations on anti-tumor efficacy and CRS incidence. This platform approach supports early clinical drug development by guiding first-in-human (FIH) starting doses and dose escalation strategies.
[1] Goutam Nair, Dinesh Bedathuru, Bhairav Paleja, Urvashi Nakul, Seshasai Pallikonda Chakravarthy, 2024, Development of QSP platform model for predicting clinical efficacy and CRS incidence of T-Cell Engagers (TCEs) targeting CD3xHLA-G in advanced solid tumors, American Conference on Pharmacometrics 15 (ACoP15). [2] Hosseini I, Gadkar K, Stefanich E, Li CC, Sun LL, Chu YW, Ramanujan S. 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 Aug 28;6(1):28. doi: 10.1038/s41540-020-00145-7. PMID: 32859946; PMCID: PMC7455723. [3] Geva R, Vieito M, Ramon J, Perets R, Pedregal M, Corral E, Doger B, Calvo E, Bardina J, Garralda E, Brown RJ, Greger JG, Wu S, Steinbach D, Yao TS, Cao Y, Lauring J, Chaudhary R, Patel J, Patel B, Moreno V. Safety and clinical activity of JNJ-78306358, a human leukocyte antigen-G (HLA-G) x CD3 bispecific antibody, for the treatment of advanced stage solid tumors. Cancer Immunol Immunother. 2024 Aug 6;73(10):205. doi: 10.1007/s00262-024-03790-7. PMID: 39105878; PMCID: PMC11303617.
Reference: PAGE 33 (2025) Abstr 11609 [www.page-meeting.org/?abstract=11609]
Poster: Drug/Disease Modelling - Oncology