PII-213 - APPLICATION OF MECHANISTIC TRANSLATIONAL PK/RO/PA MODELING TO PREDICT FIRST-IN-HUMAN DOSE FOR HPN536, A T-CELL ENGAGER TARGETING MESOTHELIN.

O. Demin Jr, G. Kolesova, D. Shchelokov; InSysBio CY.

Background: Mesothelin (MSLN) is a glycophosphatidylinositol-linked tumor antigen overexpressed in a variety of malignancies, including ovarian, pancreatic, lung cancer. HPN536 is a 53-kDa, trispecific, T-cell-activating protein-based construct, which binds to MSLN-expressing tumor cells, CD3 on T cells, and to serum albumin. The aim of our work was to predict minimal recommended starting dose (MRSD) for HPN536 using mechanistic translational pharmacokinetic (PK), receptor occupancy (RO), pharmacological activity (PA) modeling.

Methods: In vitro model was developed to describe published data on T-cell dependent cytotoxicity, cytokine secretion (IFNg, TNFa), and T-cell activation (% of CD25+ cells) in the presence of various HPN536 concentrations. PK/RO/PA model was developed to fit PK in cynomolgus monkey and translate it to human using standard allometric scaling exponents without fitting. The model for cancer patients includes distribution of TCE into the tumor (ovarian cancer was considered) and PA (cytotoxicity, T-cells activation and cytokine secretion) based on EC50 values identified in the in vitro model. PA depends on a number of timers of HPN536 bound with CD3 and MSLN in immunological synapse between T-cell and cancer cell. Minimal anticipated biological effect level (MABEL) approach was also used to estimate MRSD.

Results: MRSD calculated by MABEL approach (1 – 23 ng/kg) and starting dose used in phase 1 clinical trial of HPN536 (6 ng/kg) were significantly lower than MRSD predicted by the model (170 – 3150 ng/kg). Model prediction was validated by the fact that maximal tolerated dose in HPN536 first-in-human trial was not reached even at dose 3600 ng/kg. If binding of soluble MSLN with HPN536 was turned off in the model, predicted MRSD range was lower (110 – 2000 ng/kg).

Conclusion: PK/RO/PA model provided more accurate prediction of MRSD for HPN536 taking into account specific features of the candidate and target, as well as concentration of drug in site of action and complexity of mechanism of action (formation of trimers in immunological synapse). Application of mechanistic modeling to predict MRSD instead or in addition to MABEL approach might help to reduce the number of dose escalation steps exposing patients to sub-therapeutic doses in first-in-human trials.