The University Digital Conservancy is home to open access articles, institutional documents, dissertations, datasets, university produced publications, campus newspapers, podcasts & more. Learn about the UDC.
Prostate cancer (PC) is one of the most diagnosed malignancies and a leading cause of cancer deaths in US men. Though primary and localized PC can be well-managed by current interventions, a fraction of PC will progress to castration-resistant prostate cancer (CRPC), a lethal subtype that displays universal resistance to standard-of-care (SOC) therapies. Given the current lack of efficacious treatments, survival among CRPC patients remains poor. Therefore, there is an urgent need to develop new treatment strategies to combat advanced PC. Traditional drug development pipelines, however, remain costly and time-consuming. Recently, there has been a rapid increase in available cancer genomic, phenotypic, and high-throughput drug screening data; this has enabled the invention of computational approaches to quickly repurpose existing compounds and consequently shorten the cycle of designing new therapeutics. Nonetheless, there is still a lack of integration of efficient computational methods for discovering new treatment opportunities for CRPC. Thus, it remains imperative to establish efficient in-silico drug repurposing pipelines for CRPC that leverage computational models. Such methods will enable the rapid discovery of treatment options to improve CRPC prognoses and patient outcomes and have the potential to be applied to other diseases to advance our knowledge towards better patient care. To this end, this dissertation leverages computational approaches to cast light on drug resistance mechanisms and facilitate rapid drug repositioning for advanced PC. Chapter 1 systematically reviews recent advances in computer-aided drug discovery strategies for PCs. Chapter 2 develops a computational pipeline to screen for new drugs against resistance to androgen-deprivation therapies (ADTs) in CRPC and nominates COL-3 which shows higher efficacy in ADT-resistant models in vitro. Chapter 3 designs an analytical pipeline to select effective treatments against neuroendocrine prostate cancer (NEPC)—a detrimental CRPC subtype with extremely limited treatment options—and proposed nicotinamide phosphoribosyltransferase (NAMPT) inhibitors as drug candidates. A novel biomarker discovery approach is also developed to select robust key genes strongly associated with response to NAMPT inhibitors. Chapter 4 and Chapter 5 tackle intratumoral heterogeneity which is often linked to therapy resistance in many cancers. Chapter 4 discusses recent approaches for predicting drug response at an individual-cell level to depict variations in therapeutic vulnerability within tumors. Finally, Chapter 5 develops a new computational algorithm to infer cellular drug response and showcases drug discovery applications for diverse heterogeneous tumors, including taxane resistant CRPC. Collectively, this dissertation develops and implements computational frameworks to quickly identify efficacious drug candidates for advanced PC patients. Proposed drugs are also validated experimentally using appropriate models in vitro and warrant further investigations. Once clinically validated, these drugs can be used to tailor PC patient care and help curb the current high mortality rate in these advanced diseases. Meanwhile, this work also presents methodological contributions toward single-cell drug sensitivity predictions and applications of drug discovery for heterogeneous tumors. In addition, the proposed computational methods may be adapted to enable efficient drug screens in many other diseases.
Brain tumors are the leading cause of cancer-related death in children and efficacious treatment remains a critical unmet need. The blood-brain barrier (BBB) is a major hurdle for effective delivery of treatments for tumors in the central nervous system (CNS). While the paracellular transport of large, hydrophilic molecules is largely limited by tight junctions, efflux transporter systems are a key element of the BBB that can limit the penetration of lipophilic drugs. Histone deacetylase inhibitors (HDACIs) have been widely explored for their application in oncology, including the field of neuro-oncology. Despite their high in vitro potency and CNS-penetration-favorable physicochemical properties, the in vivo efficacy of HDACIs has been poor for CNS tumor treatment. This lack of in vitro-in vivo correlation may be in part attributed to poor CNS distribution. In this thesis project, we investigated the CNS distribution of three potent HDACIs, panobinostat, vorinostat, and quisinostat, following systemic administration. We characterized the systemic pharmacokinetics and CNS distributional kinetics of these compounds in wild-type and transgenic mice lacking p-glycoprotein (P-gp) and/or breast cancer resistance protein (Bcrp), two major efflux transporters expressed at the BBB. The in vitro stability studies show that all three hydroxamic acid-based HDACIs are enzymatically metabolized in mouse plasma, highlighting the need for careful sample handling to have accurate measurements of in vivo drug concentrations. In vivo experiments in the different mouse genotypes show that the CNS distribution of panobinostat and quisinostat is moderately limited by P-gp, but not Bcrp. Although the CNS penetration of vorinostat was not restricted by P-gp and Bcrp, its small unbound CNS tissue-to-plasma partition coefficients suggest that other efflux transporters could be involved. In addition, our results show that a tolerable dosing regimen of panobinostat would not result in adequate CNS exposure of unbound panobinostat in patients. In summary, our data show that the lack of adequate exposure of the active moieties can be a major reason for the lack of efficacy of these HDACIs in the CNS when systemically delivered. This result indicates that alternative approaches to improve delivery (e.g., convection-enhanced delivery or focused ultrasound) should be considered.
The standard setting for rate-induced tipping involves fixing a particular parameterized family of smooth forcing functions and identifying a critical value of the rate parameter. In contrast, we consider a broad collection of all possible forcing functions, continuous but not necessarily smooth, and seek a general property possessed by those which effect tipping behavior. We focus on rigidly shifting asymptotically autonomous scalar systems x ̇ = f(x + λ(t)) and identify a nonsmooth choice of forcing function λ(t) which is an optimal tipping strategy in the sense that it utilizes the least possible maximum speed. Under a co-moving change of coordinates, the problem of finding this optimal λ(t) becomes dual to the problem of finding an additive control function that achieves basin escape with minimum fuel. We show the optimizer is a bang-bang control. The outcome is a lower bound on the speed |λ ̇(t)| that must be attained at least once in order to induce tipping. Its value depends on the total arclength ∞s∞|λ ̇(t)| dt of forcing, and may be interpreted as a safe threshold rate associated to each given arclength, such that if the speed of forcing remains everywhere slower than this, tipping cannot occur. The bound is tight in the sense that there exists a forcing function which induces tipping, possesses the required arclength, and never exceeds the threshold speed. Further, the threshold speed is a strictly decreasing function of arclength, thus capturing the abstract trade off between how fast and how far of a minimal disturbance characterizes tipping. While our results assume a scalar setting, the prospect of generalizing to n-dimensions is discussed and formulated as a conjecture. The control-theoretic construction used in deriving the above inspires a new theory of resilience, which is a slight modification of the intensity kof attraction framework of McGehee and Meyer. This is a family of resilience values parameterized by a number representing the allowable L1 norm of perturbations; in the limit as the integral-constraining parameter grows unbounded, these values approach the intensity of attraction. This integral-constrained intensity of attraction has the advantage of increased descriptiveness under scenarios where limited total resources are available for perturbing the system. We suggest it to be the natural choice for quantifying the resilience of a rigidly shifting system to externally-forced tipping.
Bacteria can be engineered as vehicles for context-dependent protein production, enabling the diagnosis and treatment of numerous diseases. However, translation to clinical applications still requires concurrent efforts to enhance tunability, specificity, efficacy, and safety. Our work contributes to three such areas. First, we performed model-guided engineering of DNA sequences with predictable site-specific recombination (SSR) rates, demonstrating that recombinase attachment sites with predictable SSR rates could be used to achieve kinetic control in gene circuit design. This high-throughput, data-driven method enhanced our understanding of recombinase function and expanded the synthetic biology toolbox, and it can enable rational tuning of the response dynamics of bacterial diagnostics and the pharmacokinetic profile of bacterial therapeutics. Second, we engineered Escherichia coli to sense extracellular proteins. A cell-surface sensor is crucial to enable broad diagnostic and therapeutic applications of bacteria; however, Gram-negative bacteria such as E. coli can only detect small molecules that cross the cell envelope. We proposed a novel strategy that leverages protein-specific attenuation of maltodextrin uptake via engineered LamB porins to modulate intracellular maltose signaling, thereby linking the extracellular protein concentration to the cell response. We then demonstrated its modularity to detect different target proteins and its tunability to alter the dose-response curves. Endowing E. coli with this ability to sense extracellular proteins would enable detection of clinically relevant proteins both in vitro and in vivo. Finally, E. coli strain Nissle 1917 (EcN) is a versatile probiotic that is safe for human consumption; however, its use in diagnostic and therapeutic oral delivery applications is hindered by the low pH and high concentration of bile salts in the gastrointestinal tract that reduce EcN viability. We performed adaptive laboratory evolution on EcN to select for mutated populations with greater tolerance to pH and bile and we identified specific mutations that contribute to their enhanced viability in the presence of these chemical stressors. The developed strains may not only enhance the bioavailability of orally administered EcN but also modulate their interactions with intestinal cells, suggesting potential applications to improve gut health. In summary, through biomolecular engineering, genetic circuit design, mechanistic and data-driven modeling, and strain development, our research has contributed to improved understanding and engineering of bacterial properties that should accelerate their translation to diagnostic and therapeutic applications.
