Improving Antibody-drug Conjugate Tumor Distribution and Efficacy Using Single-Cell Imaging and Multiscale Modeling

Abstract

Antibody-drug conjugates (ADCs) are a targeted cancer therapy combining the tumor cell specificity of antibodies with small-molecule chemotherapy. Despite the widespread use of ADC therapeutics, they exhibit a heterogeneous, perivascular distribution in tumors, often leaving significant portions of the tumor untargeted. Furthermore, the relationship between the heterogeneous distribution of ADCs in tumors and their overall efficacy is poorly understood and therefore can be underappreciated. In this thesis, I develop experimental techniques to quantify ADC distribution in tumors using near-infrared (NIR) fluorophores, construct a computational model to simulate antibody distribution at several length scales, and show, for the first time, that the antibody distribution in the tumor plays an important role in the efficacy of ADCs. To better characterize the multiscale distribution of ADCs, I first measure the residualization properties of common NIR dyes, identifying both non-residualizing and residualizing dyes. Next, I show that fluorescent dye structure and dye-to-protein ratio can be optimized for labeling antibodies with NIR fluorophores to prevent the dye from impacting antibody pharmacokinetics. I then develop a novel dual label, ratio-imaging technique to quantify antibody distribution and metabolism in vivo with unprecedented single cell resolution. Using this technique, I show the clinical dose of 3.6 mg/kg the distribution of T-DM1 is heterogeneous in high HER2 expressing tumors, only targeting 10% of tumor cells. Examining the absolute uptake of ADC in targeted cells shows that they actually receive more ADC than necessary to kill the cell, despite most of the tumor not receiving any ADC. In the second part of my thesis, I develop a multiscale modeling framework combining a physiologically-based pharmacokinetic (PBPK) and Krogh Cylinder tissue model to predict both the systemic and tumoral distribution of antibodies. Using this model, I predict, and verify experimentally, that coadministration of trastuzumab with T-DM1 at 3:1 and 8:1 ratios drives a constant dose of T-DM1 deeper into the tumor. Using this dosing strategy, the total number of cells targeted increases albeit with a lower average number of ADC molecules per cell. These results are consistent across a number of antibodies, targets, and payloads, indicating the model can be used to predict ADC distribution in other tumor models. Finally, I test the efficacy of coadministration of trastuzumab with T-DM1 in a trastuzumab resistant xenograft mouse model. T-DM1 therapy alone showed a significant improvement in efficacy and survival, as expected, while trastuzumab alone had no impact. Counterintuitively, coadministration of trastuzumab, which has no efficacy in vivo and is antagonistic to T-DM1 in vitro, actually acts synergistically with T-DM1 in vivo. Coadministration of trastuzumab at 3:1 and 8:1 trastuzumab to T-DM1 dosing levels show a statistically significant improvement in survival over T-DM1 alone. These results are the first to show that the tumoral distribution of ADCs plays a major role in their overall efficacy. Overall, this dissertation provides unique tools to study antibody and ADC distribution and metabolism, quantitative computational tools to simulate in vivo distribution, and concrete guidance on how to improve efficacy of ADC therapeutics. Although I show the importance of the antibody distribution in the tumor for efficacy, additional imaging with other ADC systems, lower and more heterogeneous antigen expressing tumors, and the antibody distribution in clinical samples will further improve our understanding of the relationship between distribution and efficacy.