Date of Award

Fall 2004

Project Type


Program or Major

Engineering: Materials Science

Degree Name

Doctor of Philosophy

First Advisor

Yvon Durant


Intranasal drug delivery has been a topic of increasing interest for a decade as a convenient and reliable method for the systemic administration of drugs. The low bioavailability of simple formulation of protein drugs, such as insulin, can be greatly improved by using permeation enhancers. We studied the effect of cyclopentadecanolide (CPE-215RTM) as a permeation enhancer in protein release through lipid bilayer membranes. We successfully designed a novel in-vitro membrane permeability model using liposomes and performed a series of transmembrane protein release experiments. These were carried out under a wide range of conditions in the presence of different permeation enhancer combinations. The experimental results showed that CPE-215RTM is an effective membrane permeation enhancer for proteins and a phase transfer agent, for example, cyclodextrins, can further enhance the effect of CPE-215RTM.

Besides the release experiments, studies on insulin solution properties (self-diffusion and self-association states), the interaction between insulin and liposome and the interaction between CPE-215RTM and liposomes were carried out. Based on the mechanistic study and release data, we hypothesized that CPE-215RTM can form transient "pores" in the lipid bilayer that dissolve when CPE-215RTM distributes homogeneously within the bilayer and restore the barrier function of the lipid bilayer. We performed several experiments that corroborate our hypothesis.

A mathematical model was developed based on our hypothesized release mechanism. A semi-empirical nonlinear equation involving four parameters effectively fits the protein release profiles. The quality of the data fit with this model is good supporting evidence for the validity of our mechanistic model. Finally we used a neural network approach to correlate the different release condition parameters and the four semi-empirical fitting parameters based on our limited data sets. Reasonable neural networks were formed for the three major parameters of the mathematical model and provided acceptable prediction results.