Optimizations of in vitro model systems and protein designs for drug delivery applications

Abstract

For the treatment and detection of diseases, protein-based drugs have emerged as researchers attempt to improve affinity and selectivity when compared to traditional small molecule drugs. Among other obstacles, delivery remains a key challenge associated with protein-based drugs as conventional administration either orally or systemically necessitates traversing multiple barriers prior to arriving at their sites of action. While small molecule drugs can often take advantage of membrane permeability, most protein drugs require more sophisticated delivery mechanisms to traverse cellular barriers. Current strategies have focused on hijacking endogenous cellular mechanisms, like transcytosis, to increase uptake across barriers. To accurately predict how these mechanisms will manifest in protein drug delivery applications and to assess the potential of drug candidates in general, we need improved model systems of relevant human epithelial barriers. Furthermore, there is a need for exploration of transcytosis-engaging domains to aid in protein drug delivery. The work described in this thesis aims to improve stem cell-derived epithelial barrier systems and develop enhanced transcytosis-engaging proteins to better control and predict drug performance in drug discovery applications. First, we demonstrate development of a placental model from human induced pluripotent stem cells (hiPSCs) by treatment with two small molecules: retinoic acid and CHIR 99021. The placental barrier is of general interest in the drug delivery community given the limited available information about drug permeability and trafficking properties of the human placenta. Though other hiPSC-derived placental models have been developed, they require more factors and longer duration of differentiation to obtain relevant cell types. Our model is simple and rapid which makes it easy to use and so could have applications for disease modeling or predicting the behavior of drug compounds on and across the placenta. We also optimized a differentiation strategy to generate populations of intestinal epithelial cells from hiPSCs which has broad interest given the prevalence of oral drug delivery. By focusing on efficiency and purity, we developed a platform to screen multiple factors simultaneously based on their ability to promote expression of intestinal enterocyte markers. Because the most commonly used stem cell-derived intestinal models are organoids, which contain mixed populations of cells, we predict this strategy will lead to efficient differentiations towards specific subtypes of intestinal epithelial cells which could be used to assess the potential for systemic delivery of drug candidates taken orally. Finally, we ventured to develop novel protein reagents that have the potential to improve drug delivery across a barrier of interest. The strategy was to create a non-surface-binding transcytosis domain that when fused to a cell-surface binder would allow for barrier-specific hand-off in the endosome. An attractive initial candidate for the transcytosis domain is the signaling hormone leptin, which is naturally produced by adipocytes. Leptin binds to leptin receptor, which mediates transcytosis of the hormone across multiple barriers, including the intestine, placenta, and brain. To improve selective delivery, we used directed evolution to discover variants of leptin that show reduced binding and transcytosis at neutral pH while retaining binding and trafficking behavior at acidic pH (pH 6.5). These modifications reduce the neutral pH binding of leptin to its receptor on cell surfaces to minimize binding to non-target barriers. Future work aimed at fusing these leptin variants to a barrier-specific binder could potentially achieve hand-off in the lower pH endosome which would enable barrier-specific transcytosis. Thus, leptin variants with pH-sensitive binding represent a unique tool in applications of barrier-specific drug delivery. Advancements in human epithelial barrier cell models and in barrier-specific drug delivery will enable better predictions of drug interactions with cell barriers with the goal of increasing protein drug delivery to target cells and reducing drug off-target effects.