While T cell-based immunotherapies using cells expressing antigen-specific receptors (CARs or TCRs) have produced promising clinical responses, current approaches are limited to autologous T cells due to the risk of graft-versus-host disease (GvHD) from allogeneic T cells through endogenous TCR expression and rejection through MHC incompatibility. Human pluripotent stem cells (PSCs) have the potential to address these challenges as they are an infinitely self-renewing source of hematopoietic cells, and are amenable to gene editing approaches to address alloreactivity; however, the two main barriers to using allogeneic T cells represent two critical components required for inducing positive selection in developing T cells. Mature, na�ve T cell differentiation is dependent on signaling through TCR-MHC interactions to induce positive selection, and removal of either would effectively block this process. This dissertation explores gene editing and T cell differentiation strategies that can circumvent the basic requirements for inducing positive selection to generate antigen-specific, Class I MHC- null, cytotoxic T cells using the in vitro “Artificial Thymic Organoid” system developed in our lab.

Chapter 2 focuses on preventing alloreactive TCRs development by ablating the recombination activating genes (RAG1 and RAG2), generating RAG1-/-RAG2-/- double knockout (DKO) PSCs, responsible for endogenous TCR recombination. To support positive selection, edited PSCs were transduced to express a fully rearranged, Class I MHC-restricted 1G4 TCR (DKO+TCR), and it was determined that positive selection hinged on endogenous expression of the 1G4 TCR’s cognate MHC.

Chapter 3 centers around artificially inducing positive selection in the absence of both endogenous TCRs and MHCs. To prevent T cell rejection, DKO+TCR PSCs were edited to knockout beta-2-microglobulin (B2M), a critical subunit of Class I MHC heterodimers, to generate RAG1-/-RAG2-/-B2M-/- triple knockout PSCs with the 1G4 TCR (TKO+TCR). In order to induce positive selection of antigen-specific, Class I MHC-null, mature, na�ve T cells from TKO+TCR PSCs, the stromal component of the ATO system was engineered to provide the 1G4 TCR’s cognate MHC. Functional and transcriptional validation of TKO+TCR engineered T cells revealed a similar phenotype and cytokine release profile to unedited T cells, complete restriction of endogenous TCRs, and improved antigen-specific cytotoxicity in vivo.

Interlayered soil deposits with alternating and discontinuous beds of sands, silts, and clays underlie and interact with infrastructure worldwide. However, common simplified methods have been shown to typically over-predict earthquake ground deformations at interlayered sites, due to limitations in site characterization tools and methods, triggering and strain correlations, and analysis approaches. The most common of these simplified methods are one-dimensional liquefaction vulnerability indices (LVIs) that integrate strains and factors of safety against liquefaction triggering using empirical correlations with geotechnical site investigation data [e.g., cone penetration tests (CPTs) and borings]. Recent improvements of computational models for evaluating seismic soil responses have shown great value in resolving over-prediction biases.

This dissertation first provides a framework for site-specific CPT-based fines content (FC) correlations, which has posed significant uncertainty for both LVIs and nonlinear dynamic analysis (NDA) evaluations of liquefaction effects in interlayered deposits. Next, a system-level workflow that integrates detailed subsurface modeling with two-dimensional (2D) NDAs is utilized to evaluate the spatial extent and magnitude of ground deformation at three well-documented post-earthquake case history sites. These sites produced sediment ejecta, lateral spreading, and an extensional graben, within unique geologic environments in New Zealand, Taiwan, and California, respectively. Alternate geostatistical approaches for modeling subsurface spatial variability conditioned on available CPT data and geological details were considered, including deterministic, transition probability-based indicator simulation, kriging, and sequential Gaussian simulation (SGS) methods. The NDAs were performed with the Fast Lagrangian Analysis of Continua (FLAC) finite-difference program, with site-specific calibrations of the PM4Sand and PM4Silt constitutive models. The factors and neglected mechanisms that most contributed to prediction biases at each of these sites are explored, and improvements as well as limitations of the NDA workflow relative to LVIs and other simplified methods are discussed. This dissertation concludes with a holistic discussion of insights gained from a larger set of post-earthquake NDA case studies, and future research needs for evaluating interlayered deposits in practice.