Emily Roashan1, Neil Benson1, Daniel Jones2, Sylwia Marshall2
1Quantlmed Ltd, 2invoX Pharma
Introduction/Objectives: Only a fraction of cancer patients respond to monotherapies targeting the PD-(L)1 pathway (1). FS222 is a bispecific tetravalent monoclonal antibody, designed to promote conditional CD137 agonism in areas of PD-L1 expression (e.g., tumours) whilst also inhibiting PD-1/PD-L1 interactions. Here we describe a mechanistic model to understand the potential complexity of FS222’s mode of action, support decision-making for optimising dose and regimen, enable comparisons with alternative drug approaches, and inform future clinical trial designs. Methods: Data from a first-in-human (FIH) study in advanced solid tumor patients (2) were analysed using NONMEM, Matlab, and the Matlab SimBiology toolbox. PK data were available for 79 patients and PD data were available for 74 patients after exclusions (3). A two-compartment mammillary mixed-effects (ME) population PK model was developed to obtain parameters for individual PK. These results were used to inform a subsequent mechanistic ME PKPD model for the analysis of soluble plasma biomarkers sPD-L1 and sCD137 (soluble PD-L1 and soluble CD137, respectively) assuming random binding of FS222 to targets. Additionally, an indirect-effect Emax model downstream of ternary binding complex formation (FS222 simultaneously bound to PD-L1 and CD137), was used to characterise the relationship between FS222 exposure and Ki67+ CD8+ T-cell proliferation. Results: PK data were best described by a two-compartment mammillary ME model and soluble biomarker data for PD-L1 and CD137 were sequentially fit using a mechanistic ME model to estimate binding parameters. The estimated equilibrium dissociation constant (KD) from patient groups for CD137 was, within error, identical to the calculated in vitro value. The KD value for PD-L1 was ca. 10-fold higher than the in vitro value. This discrepancy could be uncertainty in parameter estimation or reflect a genuine biological difference between the in vitro and clinically determined values. No clear evidence of avidity type effects was observed. Observed dose-dependent increases in sPD-L1 and sCD137 were most consistent with increased shedding and decreased biomarker clearance upon FS222 binding. The results support a mechanism requiring FS222 binding to PD-L1 receptors and proximity to tumour cells to trigger sCD137 release, whereas binding to PD-L1 receptors alone is sufficient for increased sPD-L1 release. These observations are concordant with reports of sCD137 upregulation in patients treated with urelumab (4). In addition, there are some reports of increases in sPD-L1 after immune check point inhibitor treatment in patients (5). Assuming equal affinity of FS222 binding to soluble and receptor bound targets, the model was used to simulate intra-tumoural trimer concentrations and estimate receptor occupancy (RO). Model simulations predicted sustained receptor occupancy greater than 50% for CD137 across the dosing interval at doses above 1.5 mg/kg Q4W, and >50% RO for PD-L1 for half the dose period at doses greater than 1 mg/kg Q4W. Furthermore, the model suggested that hypothetical ‘three body’ type competition, which can lead to ‘bell-shaped’ responses under certain conditions, is limited for FS222 compared to drugs lacking the tetravalent binding mode of action. Ki67+ CD8+ T-cell proliferation data were well characterized by an indirect-effect Emax model downstream of ternary binding complex formation. A clear PKPD relationship was apparent, consistent with the proliferation effect initiating as the target receptors are engaged. Conclusions: The aim of the model work was to analyse population clinical data, but also to generate mechanistic hypotheses regarding e.g. optimal dose regimens and the relative nature and extent of any ‘bell shaped’ pharmacology. As such, the developed model incorporated mechanistic complexity beyond a more purely empirical approach. The model effectively described clinical data for FS222, enabling simulation of receptor occupancy and PD biomarker responses. The analysis highlighted the limited potential for bell-shaped pharmacology due to the tetravalent binding mechanism of FS222. Ki67+ CD8+ T-cell proliferation emerged as a robust downstream biomarker indicating receptor engagement, suggesting its potential as a ‘Pillar 2’ biomarker of pharmacology and is consistent with expectations of success in disease outcome. This model provides valuable insights for exploring optimal doses and regimens, informing future clinical trial design, and optimising the properties of future bispecific monoclonal antibody drugs.
[1] Sharma et al., 2017; DOI: 10.1016/j.cell.2017.01.017 [2] Elena Garralda et al. JCO 42, 2505-2505(2024). DOI:10.1200/JCO.2024.42.16_suppl.2505 [3] Jones et al, November 2024, 12(Suppl 2):A766-A766 [4] Glez-Vaz et al, J. Immunother Cancer. 2022 Mar;10(3):e003532.( doi: 10.1136/jitc-2021-003532). [5] Oh et al, Sci Rep. 2021; 11: 19712.
Reference: PAGE 33 (2025) Abstr 11646 [www.page-meeting.org/?abstract=11646]
Poster: Drug/Disease Modelling - Oncology