Bacteria live in environments featuring various complex chemical concentration gradients. The formation of these gradients is a combined result of mass transfer and bacterial activities, creating a challenge in isolating the effects of specific gradients on bacterial activities. Since the size of a single microbe is 1 to 10 μm, the spatial profiles of the surrounding chemical concentration gradients are at a resolution similar to the size of a single microbe. Moreover, time-dependent bacterial activities could lead to temporal variation in these chemical concentration gradients. Both spatial chemical heterogeneity and temporal chemical variation are important in shaping bacterial activities. The critical roles of gradients on bacterial behavior poses a significant challenge: the coexistence of numerous gradients simultaneously complicates discerning the effects of a specific gradient on a particular bacterial activity. To address this challenge, I established a workflow using electrochemistry to create electrochemical gradients to simulate specific chemical gradients in bacterial environments. Additionally, I leveraged electrode design to control the generated gradients and minimize side effects where electrochemistry could introduce to the bacterial environment. In my first research project (Chapter 2), I fabricated a flow device equipped with a microwire electrode to generate and control oxygen (O2) and hydrogen peroxide (H2O2) gradient to mimic bacterial gradients. The first highlight of this research is that I generated electrochemical H2O2 gradients through an electrocatalyst design to simulate spatial distributions of reactive oxygen species in bacterial environments. By coating the electrodes with Au catalysts to facilitate the two-electron oxygen reduction reaction (2e-ORR), I achieved a spatial profile of H2O2 perpendicular to the electrode surface. The second highlight of this project is that I enabled spatial and temporal control of both electrochemical gradients via electrode morphology design and applying time-dependent external voltages, respectively. This spatiotemporal control strategy achieved spatial control at a precision within 10 μm and temporal control on a second scale, comparable to the spatiotemporal variation of bacterial environments. Moreover, the concentration ranges of O2 (0 to 250 μM) and H2O2 (0 to 40 μM) in the electrochemical gradient, aligned with the conditions of microbiological research. To advance this research, I, together with co-authors Jingyu Wang and Ben Hoar, developed a machine-learning-assisted design strategy for desired O2 and H2O2 gradients, offering a rapid and efficient solution for electrode optimization. I created another novel electrochemical flow device for my second research project (Chapter 3). I introduced two main electrode design strategies to enhance compatibility and convenience in gradient-bacteria research. Firstly, I deposited a layer of porous copolymer on the electrode surface to increase the oxygen reduction reaction (ORR) selectivity and create a “clean” O2 gradient for bacterial research. This innovation was particularly critical in mitigating the interference of external voltages with bacterial redox systems, thus increasing the integrity of experimental results. Based on size exclusion, the coated polymer maintained the ORR current while rejecting the access of larger bacterial redox molecules to the electrode surface. Using a redox molecule produced by Pseudomonas aeruginosa, pyocyanin, as an example, the coated polymer enabled the ORR current to remain around 70% compared to that on the current on a bare electrode while suppressing almost 100% of the redox current of pyocyanin. Secondly, I adopted planar electrode arrays instead of the vertical electrode arrays used in Chapter 2. The planar geometry enabled spatial O2 concentration profile gradient generation both perpendicular and parallel to the electrode surface. This innovation proved instrumental as it facilitated microscopic microbial research while also allowing the replication of gradients within three-dimensional bacterial structures. In summary, this porous polymer coating enabled the creation of a “clean” O2 gradient, allowing investigation of the effect of O2 on microbial behavior. In Chapter 4, I proposed a potential application using electrochemical gradients in microbiological research. I describe utilizing electrochemical O2 gradients to investigate the effect of O2 variation on bacterial adenosine triphosphate concentration ([ATP]) at the single-cell level. I reviewed the published literature on the relationship between O2 concentration and microbial ATP activity. Most previous research is focused on the average effect of O2 on adenosine energy charge (AEC) and [ATP] in a bulk bacteria culture. As such, the effect of O2 on single cells is missing. I proposed an experiment plan to utilize the electrochemical O2 gradient platform to study the effects of O2 on single microbes and explored the feasibility of this plan. I raised several potential concerns or questions that could arise from an electrochemistry-microbiology hybrid system. In summary, my thesis comprehensively explores electrochemical concentration gradients and their implications for microbial behavior. By employing innovative electrode design strategies, I explored the feasibility of using electrochemical concentration gradients for nuanced investigations into the intricate dynamics of bacterial ecosystems. Looking ahead, I believe the fusion of electrochemistry and microbiology holds promise for advancing our understanding of bacterial behavior, especially at the single-cell level and beyond.

DNA nanotechnology as a versatile tool has been growing rapidly in the past three decades, with applications across different disciplines such as drug delivery, imaging, and even computing. Its versatility stems from the programmable nature of DNA and the ability to self-assemble through the A-T and C-G base pairing by hydrogen bonds. These features make DNA nanotechnology the ideal tool for mimicking and studying biological processes that are hard to otherwise model. In this work, the process we focus on is Clathrin-Mediated Endocytosis (CME). CME is one of the major mechanisms for cell entry. This pathway internalizes Influenza A, vesicular stomatitis virus, and many others. Mimicking this process will help us further understand it, and also find ways to utilize it in the field of drug delivery.The process of CME can be described in 3 major steps. The first step is array formation. Clathrin can cluster on cell surface in the form of triskelion, to form small patches of an array. The second step involves the recruitment of adaptor proteins, that transforms the array into three-dimensional lattices to encapsulate the foreign agent. The third step is the transportation of the vehicle inside the cell, along with the disassembly of the vehicle to release the agent. Specifically for each stage, we designed mimicking mechanisms using DNA nano-structures. For the array stage, we designed a three-point-star motif to mimic the triskelion structure, and functionalized it with cholesterol to integrate the array to the cell membrane. Characterization through liquid atomic force microscopy (AFM) showed clear hexagonal pattern, and in vitro cell experiment also showed the integration of the arrays to cell membranes. For the transition stage, we designed a reversible 2D-3D transition mechanism that allowed the 2D arrays to transform into 3D particles with the addition of a particular stimuli, that is a DNA single strand. The transformability and reversibility were confirmed through polyacrylamide gel electrophoresis (PAGE), dynamic light scattering (DLS) and AFM. For the delivery stage, we designed several 3D DNA structures for better drug delivery efficiency, and the designs were also characterized by PAGE and AFM. The successful designs for all three stages led us closer to understanding the structural transformation of clathrin triskelion during CME. The 2D-3D transition mechanism also has the potential to be used in other systems, such as stimuli-controlled drug release and DNA computing using single strand DNA.

A platform is established to embed functional proteins onto a DNA nanogel. These proteins are embedded through self-assembly mechanisms between the DNA and the proteins. The DNA nanogel is composed of single stranded DNAs, which forms a 3-dimensional nanogel structure by using a G-Quadruplex secondary structure as the crosslinker. The DNA nanogel has multiple overhanging 20 adenosine sequences, which can be used to modify the nanogels by simple Watson-Crick base pairing. In this thesis, we have demonstrated that the DNA nanogel has embedded the unmodified proteins while still preserving the structure and function of the proteins. As a proof of concept, we have confirmed the embedding with mobility shift assay of native agarose gel electrophoresis, imaged embedded proteins onto DNA nanogel with Atomic Force Microscopy(AFM), characterized the size and zeta potential with Dynamic Light Scattering(DLS). In addition, we have proved that embedded proteins preserved their function by showing the nanogel-IgG system can be conjugated to the surface proteins of E.Coli. To develop our system as a Nanomedicine platform, we have modified our nanogel to be used as targeted drug delivery vehicle for KB cells. Doxorubicin was loaded onto the DNA nanogel with embedded Immunoglobulin G(IgG)s. The nanogel-IgG system showed higher cytotoxicity to KB cells than controls due to the ability of the IgG to function as a targeting ligand. This result demonstrates the self-assembled system can function as a possible next generation drug delivery vehicle.

This dissertation is devoted to understanding household saving rates of the two largest countries in the world - China and the United States. The first two chapters explain why the Chinese elderly save at extraordinarily high rates and the third chapter explains why the U.S. personal saving rate has been falling since 1980's.

Chapter 1 explores the potential explanation of the high saving rates of the Chinese elderly. The high saving rate of China has attracted global attention. Furthermore, the saving rates of the Chinese elderly are especially high. Understanding why the elderly in China save at high rates is important for two reasons: (1) it partially explains the high aggregate saving rate in China, and (2) the fact that the elderly save more than the middle-aged contradicts the predictions of the life-cycle model. In this chapter, I present evidence that pension income is the primary explanation for the high saving rates of elderly Chinese households. I provide this evidence in two steps. First, I document three stylized facts that are consistent with this hypothesis: (1) saving rates are higher in years with higher pensions, (2) saving rates are higher for those with more generous pension plans, and (3) policy reforms that exogenously increase pensions also increase saving rates. However, a higher pension income on its own cannot explain the entire pattern because a household can simultaneously adjust its consumption. Therefore, in the second step, I demonstrate that concerns regarding future medical expenditures and bequest motives can explain why households do not increase their consumption commensurate with increases in pension income.

In Chapter 2, I build and estimate a dynamic life cycle model for two purposes. The first is to quantify the effect of pension income. The second purpose is to carry out counterfactual policy simulations. The model is a standard life-cycle model with three main components. First, pension income is properly modeled to capture the increase observed in the data. The second part of the model is about uncertainty. In the model, I cover income uncertainty, health status and medical expenditures as the main source of uncertainty for the elderly. Finally, individuals have bequest motives. I estimate the model using the method of simulated moments. The estimation results show that it is possible to match the data with reasonable parameters. It is noteworthy that the estimated degree of relative risk aversion for the Chinese elderly is similar to that of U.S. population in other studies. This implies when factors including pension income, medical expenditures and bequest motives are properly taken into account, it is not necessary to assume Chinese elderly to be highly risk averse to explain their high saving rates. With the model, I am able to carry out various policy simulations. The most interesting simulation is if the Chinese pension and economy growth rate becomes similar to those in the United States, the saving rates of the Chinese elderly will fall to the level of the U.S.

Chapter 3 is a joint work with Maurizio Mazzocco and Bela Szemely. In this chapter we provide evidence that most of the decline in the U.S. personal saving rate from 9 percent in the early eighties to 2 percent in 2007 can be explained by the steep increase in health expenditure experienced by the U.S. economy during the same period. The most convincing evidence is provided using the FDA approval of new drugs as a source of exogenous variation in medical expenses. Employing this source of variation, we find that a $1$ percentage point increase in health expenditure generates a decline in the U.S. saving rate that is between $0.58$ and $0.67$ percentage points. Using this result, we calculate that the rise in health expenditure explains about 83 percent of the drop in the U.S. saving rate. To evaluate whether households changed their consumption decisions to mitigate the effect of higher medical expenses, we develop a stylized model of household's and government's decisions. Using the model jointly with our empirical results, we find that the households' response to the rise in health expenses was negligible. This is why the saving rate dropped by a significant amount. Finally, with the objective of better understanding why households did not respond, we provide evidence on how the increase in medical expenditure was funded. We find that it was paid almost exclusively by an increase in government debt, a reduction in other government expenses, and an increase in employer contributions to health funds. The main implication of these findings is that the households were barely affected by the rise in health expenditure. The households' negligible response was, therefore, rational.

In the field of DNA Nanotechnology, a field that utilizes synthetically designed DNA strands and their inherent complimentary nature to create two and three dimensional structures, there are many approaches to creating structures to act as vehicles for drug delivery and platforms for diagnostics. However, this field holds an expense due to the number of oligonucleotide strands that are required in order to form the structures. These structures are also limited by the amount of therapeutic load they can carry, the density of targeting modalities that can be attached to the surface, and the variability of other proteins or small molecules that we may wish to include. Our approach to this problem was to utilize a natural, bacterial DNA amplification process known as Rolling Circle Amplification, or RCA. With RCA, we can amplify a sequence into a long stranded DNA sequence carrying the complementary sequence repeating multiple times sequentially on one strand. Using these repetitive sequences, we can create structures by binding them to themselves in certain manners. Any unbound sequences can be used to bind to, and capture, many different modalities and carry a large variety of materials or therapeutics to a designated target. They can also be of great use in capturing large quantities of an analyte in detection motifs and modalities. My RCA structures are capable of binding large quantities of materials, thusly delivering them to targeted cell populations through the use of binding many ligands on the surface. These structures are also capable of undergoing morphological changes in order to facilitate delivery or act as a diagnostic modality in the presence of the target analyte or environment. Thusly, these structures, created by amplifying a single DNA strand, as opposed to utilizing many, have a great advantage over current DNA Nanotechnology techniques for creating delivery and diagnostic vehicles due to their great flexibility in binding capabilities and the decreased cost for synthesis.

Specific synaptic connections underlie the ability of our nervous systems to perform the complex computations that account for our daily perception and behavior. How these connections arise during development is a central question in neuroscience. Due to the cellular complexity of the central nervous system (CNS) and the small size of synapses, it is difficult to efficiently visualize synapses of identified neurons in vivo, and this has become a major obstacle to studying mechanisms of synapse formation and synaptic specificity. Synapses are traditionally visualized with Electron Microscopy (EM), which can generate comprehensive and accurate synaptic connectivity maps. However, EM techniques are time consuming and labor intensive, making it difficult to study the dynamic process of synaptic development and identify molecular pathways involved in synapse formation using this method. Therefore, my thesis research has focused on developing techniques facilitating synapse visualization in vivo with light microscopy. I first adapted a technique originally developed in C elegans called GFP reconstitution across synaptic partners (GRASP) to the nervous system of Drosophila melanogaster. I showed that adapted GRASP effectively detected synapses between known synaptic partner neurons within the fly visual system. It also successfully mapped neuronal connections within the neural circuit that underlies fly mating behavior. Due to some caveats in the design, in certain cell types GRASP failed to distinguish synapses from general cell-cell contacts. Inspired by GRASP and the prospect of addressing its limitations, I designed a new method called Synaptic Tagging with Recombination (STaR), which labels endogenous presynaptic and postsynaptic proteins in a cell-type-specific fashion. I modified genomic loci encoding synaptic proteins within bacterial artificial chromosomes such that these proteins, expressed at endogenous levels and with normal spatiotemporal patterns, were labeled in an inducible fashion in specific neurons through targeted expression of site-specific recombinases. Within the fly visual system, the number and distribution of synapses labeled with STaR correlate with EM studies. Using two different recombination systems, presynaptic and postsynaptic specializations of synaptic pairs can be co-labeled, and synapses between specific partners can be identified by assessing the apposition of these specializations. With STaR, I characterized synaptic development in photoreceptor neurons and uncovered a novel transformation phase of growth cones to synaptic terminals. This has led to the generation of gene expression profiles before, during and after the transformation phase and an in vivo RNAi screen to identify genes regulating photoreceptor synaptogenesis using the STaR markers. Furthermore, combining STaR with two-photon microscopy allowed visualization of synapse formation in live animals. In principle, STaR can be adapted to the mammalian nervous system. Both GRASP and STaR will facilitate our future investigation of key questions in synapse biology.

Modeling consumer choice plays a central role in modern business operations and demand prediction. Precedented approaches assume consumers are rational, i.e., assume that consumer choice models follow rational choice theory and are based on the random utility maximization (RUM) principle. However, abundant evidence from marketing science, psychology, and behavioral economics has shown that consumer choice is not always consistent with the RUM assumption.

In this thesis, we introduce a nonparametric and data-driven choice model that is capable of representing any consumer choice, including those that are outside the RUM class. We theoretically characterize the model complexity and propose two practical estimation procedures to learn the model from data. Using real-world transaction data, we demonstrate the out-of-sample prediction ability of the proposed model and extract business insights.

We further transform the proposed model into effective prescriptions. We consider a mixed-integer optimization approach to find the optimal assortment that maximizes expected revenue under the proposed model. We introduce three formulations, analyze the necessary conditions for integrality, and solve them at a large scale by applying Benders decomposition method. Using synthetically generated data, we demonstrate the tractability of our approach and its edge over heuristic approaches from the literature.

Finally, we generalize the estimation procedure of the proposed model as a general solution method for solving large-scale linear optimization problems. The proposed solution method is a randomized algorithm that first samples a set of columns and then solves the linear program that only consists of sampled columns. We theoretically characterize the algorithm's convergence property and apply it to a wide range of applications, including Markov decision processes, covering and packing problems, portfolio optimization, and choice model estimation.