Designing Dynamic and Modular Biomolecules and Assays to Interrogate and Control Protein Fate


Cancer has long been the second-leading cause of death in the United States and represents the leading cause of death in midlife (age 40-60). While the prognosis for many cancers has vastly improved over the last thirty years, many cancers remain elusive due to the late-onset of symptoms, the specific organ systems they affect, the primary sites of metastasis, and, of course, the type of tumor (e.g. solid v. blood) and the subsequent oft-immunosuppressive tumor microenvironment. Biological therapeutics (i.e. biologics) have revolutionized the way we treat cancer due to their inherent ability to successfully target overexpressed antigens – often, proteins expressed on the surface of cancer cells – while minimally affecting healthy cells. The most common biologic is the immunoglobulin G (IgG) monoclonal antibody (mAb), a Y-shaped protein secreted by plasma B cells of the adaptive immune system. However, there is an inherent inability to easily optimize the structure of an IgG for maximal efficacy, and this lack of programmability can contribute to issues biologics often face such as low tumor penetration, nonspecific immunogenic responses, rapid clearance, and high dosage requirements. To modulate the structure-function of biologics to improve cancer treatment, mitigating the dosage of non-discriminatory traditional chemotherapy in the process, our lab has developed a protein assembly platform technology known as ‘megamolecules’ (Chapter 1) which uses rapid, specific, and irreversible enzyme-inhibitor reaction chemistries to covalently bring fusion proteins together. The megamolecule approach provides atomic-level precision over the synthesis of protein scaffolds, and these scaffolds can modulate inherent properties of biologics such as binding specificities, affinities, orientations, and stoichiometries with relative ease. In Chapter 2, this next-generation, modular assembly strategy was utilized to develop a library of therapeutics towards breast cancer research, building off our lab’s initial demonstration of synthesizing, characterizing, and utilizing megamolecules to create mimics of the mAb trastuzumab. While trastuzumab – often in combination with the mAb pertuzumab – has shown moderate success in the clinic for HER2+ breast cancer patients, immune tolerance typically results, leading to a transiently efficacious drug. Thus, there is sufficient room to improve upon this well-researched mAb. I used megamolecules to investigate how HER2-targeting scaffolds can be modulated to interrogate biologic properties such as binding affinity, avidity, net internalization rate of the megamolecule-receptor complex, and downstream inhibition of cell proliferation. Increasing the binding valency of our megamolecule scaffolds from 2 to 3 only modestly improved binding affinity and had no effect on increasing megamolecule-HER2 endocytic rate nor the inhibition of cell proliferation. Creating bispecific (biparatopic) scaffolds that targeted two different epitopes on HER2 was the only way to significantly increase net internalization rate by cross-linking domains I and IV on the HER2 extracellular domain. Interestingly, scaffolds that only presented the trastuzumab Fab domains were the only candidates that showed significant inhibition of proliferation. Here, even adding an extra nanobody towards domain I within scaffolds that had two trastuzumab Fabs completely abrogated the inhibition of cell proliferation seen with scaffolds that had two trastuzumab Fabs alone. Next, Chapter 3 explores the utility of the megamolecule platform as a proof-of-concept reversible protein switch. Here, we utilized synthetic chemistry to build terpyridine-terminated small molecules that irreversibly reacted with one of our megamolecule enzymes, cutinase. Once incorporated into a megamolecule scaffold, two terpyridine groups could reversibly coordinate upon addition of bivalent transition metals (e.g. Ni2+, Co2+, Zn2+). Strategically positioning each terpyridine group at opposing ends of a linear megamolecule scaffold allowed for quaternary-scale domain cyclization, which could be quantitatively discerned through Förster Resonance Energy Transfer (FRET). Ultimately, I demonstrated that terpyridine coordination – and therefore, FRET signal – was dependent on addition of specific divalent transition metals, which could be reversibly sequestered by addition of excess ethylenediaminetetraacetic acid (EDTA). The specific FRET response was unique to the length of each sensor as well as the individual metal ion; the data strongly correlated with long-standing literature of terpyridine-metal and EDTA-metal coordination kinetics. Longer scaffolds had faster coordination kinetics (i.e. kon) towards the bidentate complex, which, again, were unique to each individual metal. Coarse-grain modeling and small-angle X-ray scattering (SAXS) showed good agreement with experimental results, suggesting that the megamolecule platform’s flexibility for synthesis of various protein scaffolds could be utilized within a protein sensor framework. While the treatments for cancer are varied and complex, appropriate diagnosis and favorable prognoses rely on early and accurate detection. In Chapter 4, I utilized surface chemistry techniques to pattern single cells into specific shapes that, when stained for their actin cytoskeleton, could discriminate between cancer and non-cancer cells with a feature-extraction machine learning algorithm. High-resolution (60X) confocal microscopy imaging against the actin cytoskeleton without any patterning was sufficient to discriminate between two cell populations in the instances where phenotypes were quite distinct, which ran against our initial hypothesis of always requiring shape normalization a priori. In fact, patterning cells into shapes for algorithmic discrimination was only effective when cell lines had similar, overlapping phenotypes. This work demonstrates a compelling proof-of-concept incorporation of high-resolution confocal microscopy into quantitative machine learning workflows. In Chapter 5, I present a co-author project from earlier in my PhD, which provided a necessary breadth to my five years at Northwestern. This published work interrogated phosphatase activity and specificity from various cancer and non-cancer cell lysate utilizing our platform technology known as SAMDI. Here, high-throughput, modular peptide arrays were treated with cell lysate, and we were able to demonstrate that phosphatase activity and specificity were conserved across cell lines, cancer states, and species. Furthermore, phosphatases in the lysate were universally more active towards phosphorylated threonine than serine on our peptide arrays, which may contribute to the reported differences in phosphorylation seen across the phosphoproteome. This work is important because most research in the field focuses on activity and specificity of kinases. In Chapter 6, I shortly reflect on my PhD, the major conclusions of my work, and discuss potential research projects for future students.

Alternate Identifier
Date created
Resource type
Rights statement