2025 Theses Doctoral
Harnessing protein engineering for the study of antiviral drug resistance and the development of therapeutics targeting neurodegenerative disease
Proteases play an indispensable role in medicine, serving both as drug targets and as therapeuticagents. Protease inhibitors are a key component of our antiviral arsenal, and are widely used to combat human immunodeficiency virus (HIV), hepatitis C virus (HCV), and, most recently, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Nirmatrelvir, an inhibitor of the 3-chymotrypsin like (3CL) protease essential for SARS-CoV-2 replication, was granted emergency use authorization in late 2021, formulated with ritonavir and sold under the brand name Paxlovid. Since then, Paxlovid use has become widespread, raising the possibility that nirmatrelvir resistance could emerge in circulating SARS-CoV-2. It is therefore important to understand how the 3CL protease might mutate to lose nirmatrelvir sensitivity so that circulating variants can be monitored for these mutations and future generations of inhibitors can be designed to prevent cross-resistance.
The most widely used therapeutic proteases are the botulinum neurotoxins (BoNTs),which are effective in the treatment of a wide array of movement, pain, and autonomic disorders. These toxins exert their therapeutic effect by cleaving members of the SNARE protein family inside neuronal cytosol, preventing neurotransmitter release. Much work has been dedicated to engineering members of the BoNT family to extend their therapeutic utility, including altering receptor tropism, extending half-life, and modifying protease specificity. While significant progress has been made in each of these areas, the extent to which these proteases can be reprogrammed to target the proteins that cause human disease remains unexplored.
In this dissertation, we employ diverse methods to study and direct protease evolutionand to facilitate protein engineering more broadly. To investigate the emergence of nirmatrelvir resistance, we passage SARS-CoV-2 in increasing concentrations of the drug and sequence the 3CL protease gene over time in resistant lineages. We then validate the observed 3CL protease mutations by incorporating them into recombinant SARS-CoV-2 and testing the nirmatrelvir sensitivity of the resulting viruses. We find that the development of nirmatrelvir resistance typically begins with the acquisition of a precursor mutation such as T21I, P252L, or T304I, which confers a low level of resistance and enables strong resistance conferring mutations, such as E166V, to emerge without imposing a significant fitness cost.
To explore the programmability of the BoNT proteases, we engineer type E (BoNT/E) to cleave proteins involved in neurodegenerative disease. To accomplish this, we employ targeted mutation based on a structural model as well as the continuous evolution platform known as OrthoRep, selecting for variants that cleave the desired substrates with a circuit that links protease activity to the growth of a Saccharomyces cerevisiae (yeast) cellular chassis. We first use this approach to target ATXN3, the protein whose aggregation causes type 3 spinocerebellar ataxia (SCA3). We then profile the substrate specificities of the BoNT/E variants that emerged during our ATXN3 engineering, identifying patterns that can inform the selection of new targets. Based on one of these patterns, we generate a new BoNT/E variant capable of cleaving TDP-43, a protein implicated in amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD).
Finally, to facilitate future protein engineering campaigns, we develop a machine learning pipeline that uses deep mutational scanning (DMS) data to model a protein’s fitness landscape and predict optimized variants with user-defined constraints. We name this pipeline OptiProt, and we demonstrate its utility by applying it to the human chaperone DNAJB6, which rescues cellular toxicity in a yeast model of ALS. In this context, we probe OptiProt’s engineering capabilities with a range of challenges. We first predict hyper-functional variants with up to 50 mutations. We then restore activity to variants harboring one of two known loss-offunction mutations. Finally, to demonstrate the OptiProt’s amenability to complex engineering constraints, we restore activity to a variant harboring a deleterious mutation while simultaneously mutating a set of highly conserved residues.
Altogether, the work presented in this dissertation highlights the value of laboratory protease evolution in two clinically relevant applications and provides a tool that can facilitate future work in this field.
Subjects
Files
This item is currently under embargo. It will be available starting 2027-01-14.
More About This Work
- Academic Units
- Cellular, Molecular and Biomedical Studies
- Thesis Advisors
- Chavez, Alejandro
- Degree
- Ph.D., Columbia University
- Published Here
- January 15, 2025