De-novo Protein Design by Tertiary Motif Mining
The ability to design functional proteins on demand would transform health research, enabling robust development of molecular probes, therapeutics, and signaling pathway regulators. Computational protein design (CPD) offers the potential to fill this need, but despite considerable progress, major limitations remain. The goal of this proposal is to advance a novel CPD method based on the degeneracy of protein structure at the level of local tertiary structural motifs (TERMs). This method, dubbed dTERMen, is different in most aspects from existing techniques. It works by decomposing the target template it into its constituent TERMs and using close structural representatives of each TERM in the Protein Data Bank to quantify the sequence-structure compatibility rules of the target. We have now applied dTERMen to a variety of design challenges, with experimental validation. This proposal focuses on de-novo protein design—i.e., a task in which both the structural template and a sequence for stabilizing it are designed from scratch. A systematic demonstration of dTERMen on this challenging task, in comparison with the state of the art, will firmly establish the merits of the method, enabling the broader biomedical and protein-science communities to take advantage of its versatility and robustness. We already have exciting developments in this direction. Collaborating with the David Baker lab at University of Washington has enabled us to experimentally characterize ~10,000 de-novo dTERMen design variants targeting a broad variety of structural templates, which included diverse all-beta, all-alpha, and mixed-topology structures. Using a high-throughput experimental platform, we showed that ~17% of our designs were stably expressed and folded. Excitingly, there were successful variants for every single designed topology! De-novo design at this scale and diversity has not been attempted before, and our results indicate an unprecedented level of robustness for dTERMen. Crucially, however, the experimental platform characterizes only the overall stability of designed sequences but does not report on whether the folded state matches the design target. Thus, in Aim 1, we propose to structurally characterize a representative subset of the design variants, sampled from a range of structural topologies. Furthermore, comparison with existing methodologies will be critical for establishing the relative merits of dTERMen. Thus, Aim 2 proposes to perform a similar design-to-experiment cycle using Rosetta Design, which represents the current state of the art in CPD.
Engineering Immunogens to Focus the Immune Response on Broadly-Neutralizing Dengue Epitopes
Dengue virus causes hundreds of millions of human infections each year. There are four co-circulating dengue serotypes; primary infection by a single serotype results in febrile illness and subsequent lifelong immunity to that serotype, while secondary infections by heterotypic serotypes can lead to severe disease and even death. Severe dengue disease is likely caused by cross-reactive antibodies elicited during primary infection that can bind multiple serotypes but not neutralize them, instead facilitating viral entry into Fcγ receptor positive cells and thereby yielding “antibody-dependent enhancement“ (ADE) of infection. We propose to use structure-guided protein engineering to develop novel immunogens that focus the immune response so as to elicit broadly protective antibody responses avoiding ADE. In particular, we target two critical epitopes, the E glycoprotein domain III (DIII) and the E dimer interface (EDE), which are sites of binding by broadly neutralizing antibodies (bNAbs).
Target Validation Against Bacterial Pathogens
Many bacterial pathogens of humans have cbb3 oxidases terminal oxidases with high affinities for O2 that enable microbes to compete for O2 in the host, particularly in the microoxic settings associated with infections. In Pseudomonas aueroginosa (Pa), cbb3 oxidases are essential for fitness of the biofilm communities, and they are highly expressed by bacteria during chronic infections in cystic fibrosis. Because cbb3 oxidases, and the proteins that they interact with, are unique to bacteria and have several distinct features from those of mammalian oxidases, these enzymes are attractive drug targets for potential antibacterial therapies. Development of such drugs requires mechanistic understanding of these enzymes. We have identified two critical elements of the Pa cbb3-dependent respiratory pathway: the delivery of O2 under microoxic conditions by the O2 carrier hemerythrin (Hr) and the delivery of electrons for cbb3 catalysis by the electron carrier cytochrome c4 (cyt c4). The goals of this pilot project are to get key preliminary data on the mechanisms by which Hr and cyt c4 interact with cbb3 oxidases and to explore the possibility of targeting them in drug design.
Apoptotic in Triple-negative Breast and Ovarian Cancers
Triple-negative breast cancer (TNBC) and ovarian cancer (OVCA) are aggressive malignancies that share molecular abnormalities but have limited treatment options. To identify therapeutic strategies based on molecular profiles that may be applicable to subgroups of both TNBC and OVCA, we performed a combined analysis of gene expression profiles and identified two major subgroups of mixed TNBC/OVCA cell lines (Mesenchymal-like and Basal-like). Drug sensitivity analysis revealed that Mesenchymal-like TNBC/OVCA cells are exquisitely sensitive to inhibitors of Hsp90, while Basal-like TNBC/OVCA cells are resistant. Dissecting the pathways underlying these different sensitivities will offer strategies to fully exploit therapeutic vulnerabilities and identify novel drug targets for subgroups of TNBC/OVCA. We found that Hsp90 inhibition decreases activation of the MEK/ERK oncogenic pathway in both Mesenchymal-like and Basal-like cells, but only Mesenchymal-like cells upregulate the pro-apoptotic proteins Bim (an ERK substrate) and PUMA (inducible by inhibition of ERK), and undergo apoptosis. We hypothesize that Hsp90 inhibition disrupts MEK/ERK signaling to A) repress Bim phosphorylation, inducing Bim stabilization, and/or B) induce FOXO-mediated transcription of Bim and PUMA, thereby inducing apoptosis in Mesenchymal-like but not Basal-like TNBC and OVCA. We will test our hypothesis through the following Specific Aims: 1) determine the level(s) of processing at which Hsp90/MEK/ERK inhibition upregulates Bim and PUMA to induce apoptosis in Mesenchymal-like TNBC and OVCA; 2) determine the response of Basal-like TNBC and OVCA cells to inhibition of Hsp90/MEK/ERK at the levels of Bim/PUMA expression, apoptotic priming, and cell fate. Understanding the mechanism by which Hsp90 inhibition induces apoptosis in Mesenchymal-like TNBC and OVCA would support clinical testing of Hsp90 inhibitors in this molecularly defined subgroup, and offer strategies to enhance the anti-cancer effects of Hsp90 inhibition. Understanding the disconnect between MEK/ERK and Bim/PUMA and cancer subgroup-specific differences in processing of pro-apoptotic proteins will offer novel therapeutic strategies for the Basal-like subgroup of TNBC/OVCA.