(T1330-02-10) A PKPD Model to Predict the Therapeutic Index of Bispecific Antibody Targeting MerTK in Preclinical Mouse Model

Purpose: Inhibiting MerTK on macrophages is a promising therapeutic strategy for augmenting anti-tumor immunity[1]; however, blocking MerTK activity in retinal pigment epithelial cells (RPEs) results in retinal toxicity in mice[2]. To mitigate this toxicity, multiple bispecific therapeutic antibodies (bsAbs) targeting MerTK and PD-L1 are developed to reduce drug binding to MerTK on RPEs, since PD-L1 is expressed on macrophages but not RPEs. To aid the rationale design of this drug, we developed and utilized a bispecific PKPD model to understand the relationship between dose/pharmacokinetics (PK)/toxicity/efficacy of bsAbs in mice. We then used our model to predict PK, efficacious dose, toxic dose, and therapeutic index (TI) of anti-MerTK/PD-L1 bsAbs. These model predictions were in line with experimental observations with these bsAbs.

Methods: Two variants of anti-MerTK/PD-L1 bsAbs were produced, one with potent (anti-MerTK WT/PD-L1 bsAb) and one with attenuated (anti-MerTK Atten/PD-L1 bsAb) MerTK binding. PK, anti-tumor efficacy, and toxicity of anti-MerTK/PD-L1 bsAbs were compared to that of anti-MerTK monospecific antibody (msAb). Serum PK was measured via ELISA in mice following a single-bolus intravenous injection of the drug. Anti-tumor efficacy studies were performed using a MC38 mouse tumor model. Toxicity evaluation was performed in healthy mice following twice weekly (BIW) injections of anti-MerTK msAb and bsAbs. Serum toxicokinetics (TK) measurement was obtained via ELISA. A bispecific TMDD (target-mediated drug disposition) model was developed (Fig. 1a). Model parameters were obtained by fitting the model to the serum PK data. Serum, retinal, and tumor PK of anti-MerTK bsAbs in mice treated with varying doses of the drugs were simulated using the model, assuming tumor:serum partition of 5% and retina:serum partition of 0.25%[3]. Finally, to predict maximal tolerable dose (MTD) and minimal efficacious dose (MED), these retinal or tumor PKs were compared to the measured concentrations of the drugs that would induce 20% or 90% of MerTK receptor occupancy in RPEs (EC20_RPE) or macrophages (EC90_MΦ), respectively. Therapeutic index was calculated by dividing MTD by MED. Similarly, a monospecific TMDD model was used to predict the PK and TI of anti-MerTK msAb.

Results: We developed a bispecific TMDD model to describe the binding of the bsAbs to MerTK and PD-L1, as well as clearance and distribution of bsAbs. We fitted our model to the serum PK of both anti-MerTK bsAbs, which demonstrated TMDD behavior (Fig. 1b). This model predicted the experimental serum TK data of both anti-MerTK bsAbs (Fig. 1c). Similarly, the monospecific TMDD model was fitted against the experimental PK data of anti-MerTK msAb and successfully predicted the experimental TK data of this msAb. Using the developed models, we simulated serum, retinal, and tumor PK of mice receiving BIW dosing of 1~60 mg/kg anti-MerTK msAb or bsAbs. For anti-MerTK Atten/PD-L1 bsAb, we found that ≤30 mg/kg dosing produced a retinal drug concentration that is lower than EC20_RPE, while ≥3.5 mg/kg dosing predicted tumor drug concentration that is higher than EC90_MΦ (Fig. 2). Based on these simulated results, we predicted MTD of 30 mg/kg, MED of 3.5 mg/kg, and TI of 8.6 for anti-MerTK Atten/PD-L1 bsAb. A TI of 2.9 (MTD=10 mg/kg, MED=3.5 mg/kg) and TI of 1 (MTD=2.5 mg/kg, MED=2.5 mg/kg) were predicted for anti-MerTK WT/PD-L1 bsAb and anti-MerTK msAb, respectively (Fig. 3). These results suggest a widest therapeutic window for anti-MerTK Atten/PD-L1 bsAb. This conclusion matches well with experimental observations, as in vivo efficacy and toxicity studies showed that anti-MerTK Atten/PD-L1 bsAb was better tolerated and produced equivalent anti-tumor efficacy compared to the other two antibodies.

Conclusion: We have developed bispecific TMDD PKPD models to predict PK, MED, MTD, and TI of therapeutic bsAb targeting both MerTK and PD-L1. We believe this bispecific modeling framework could aid the rationale design of other bsAbs that aim to enhance on-target/on-tumor efficacy and reduce on-target/off-tumor toxicity.

References: [1] Zhou et al., Immunity, 2020; [2] Hamm et al., Arch.Toxicol., 2002; [3] Shivva et al., Front.Pharmacol., 2021.

Fig. 1. (A) Bispecific PKPD model for Anti-MerTK/PD-L1 bsAbs. (B) Experimental observation and model fit of the anti-MerTK Atten/PD-L1 bsAb PK data. (C) Experimental observation and predicted values of the anti-MerTK Atten/PD-L1 bsAb TK data.



Fig. 2. Model prediction of serum (A), retinal (B), and tumor (C) PK of anti-MerTK Atten/PD-L1 bsAb in mice following various dosing regimens. Toxic and efficacious concentration ranges were derived from MerTK receptor occupancy data obtained from RPEs and macrophages, respectively.



Fig. 3 Predicted MTD, MED, and TI of anti-MerTK antibodies for tumor:serum partition of 5% and retina:serum partition of 0.25%.