Characterizing and predicting the interaction of proteins with nanoparticles

Nanoparticles are being used or implemented in a wide array of applications including health care, cosmetics, automotive, food, beverage, coatings, consumer electronics, and coatings. Despite this widespread use, we are unable to predict how nanoparticles will interact with biological systems. This is important for both nanotoxicity resulting from human exposure to nanomaterials and the development of new nano-based biotechnologies. The first step in the interaction of nanoparticles with biological systems is often with blood, for biomedical applications, or lung fluid, when nanoparticles are inhaled. In both cases, the nanoparticles are exposed to proteins that form a "corona" by adsorbing on the nanoparticle surface. The subsequent cellular response is determined by this protein corona rather than the bare nanoparticle.Our goal is to develop a predictive capability for protein-nanoparticle interactions. This requires lab automation, large datasets, machine learning, and mechanistic studies. We first developed and validated a semi-automated approach to generate, purify, and characterize protein coronas using a liquid handling robot and low-cost proteomics. Using this semi-automated approach, we characterized the formation of protein coronas with increasing incubation time and serum concentration. Incubation time was found to be an important parameter for corona composition and concentration at high incubation concentrations but yielded only a small effect at low serum incubation concentrations. To better understand how the protein corona affects biological responses, we investigated the response of macrophage cells to titanium dioxide nanoparticles as a function of the protein corona. As in our previous work with serum proteins, we measured the concentration and composition of murine lung fluid proteins adsorbed on the surface of titanium dioxide nanoparticles. We found that a low concentration of lung fluid corona, relative to fetal bovine serum and bovine serum albumin coronas, led to an increased expression of cytokines (interleukin 6 (IL-6), tumor necrosis factor-alpha (TNF-α), and macrophage inflammatory protein 2 (MIP-2)), indicating an inflammation response. This underscores the importance of understanding how the composition and concentration of the protein corona governs organism responses to nanoparticle exposures. Our validated high-throughput lab automation work allowed us to synthesize a library of magnetic nanoparticles and characterize their resulting protein coronas. The resulting nanoparticle dataset has 12 unique NP surfaces, seven surface charges, two core sizes, and two core materials. We used this dataset to generate and characterize, via proteomics, a variety of protein coronas varying incubation concentration and purification methods. We used the resulting proteomic dataset in conjunction with a database of protein physicochemical properties to build a machine learning model that identifies the most important nanoparticle and protein properties for protein corona formation. The model was tested using external datasets and found that it can predict human serum and lung fluid coronas on varying nanoparticle surfaces. Overall, this combination of lab automation, machine learning, and mechanistic analysis suggests that a generalizable understanding of the protein corona formation and its effects is forthcoming.