Transcription factors (TFs) are important messengers in the information cascade necessary for cells to respond to their environment. In both Saccharomyces cerevisiae and mammalian cells, the limited set of TFs must use a variety of strategies to convey complex information about the cellular environment to the many genes that they regulate downstream. These strategies include TF identity, concentration, and combinations of TFs. A subset of TFs also modulate their subcellular localization with distinct dynamic patterns in response to environmental stimuli; this dynamic information may also be used to regulate downstream genes.

Previous studies have attempted to elucidate the effect of spatiotemporal dynamics on downstream genes by using environmental inputs to induce localization changes. Using these stimuli, these studies have shown that TF spatiotemporal dynamics likely transmit information which is then decoded by downstream genes. An important limitation of these works is that the inputs used modulate not only TF spatiotemporal dynamics, but also other factors, like TF concentration and activation of other TFs and transcriptional regulators. In this study, we develop an optogenetic tool, termed CLASP, which sidesteps these limitations to precisely quantify the effect of TF spatiotemporal dynamics on downstream genes.

Specifically, we used CLASP in S. cerevisiae to control the localization of multiple TFs, and find that canonical downstream genes activate more efficiently in response to constant TF nuclear localization than to short pulses of TF localization. In contrast to these data, we focus on Crz1, a pulsatile TF, and find that many of its downstream genes activate more efficiently in response to short, pulsed TF inputs. We then use computational modeling to interrogate the promoter architecture which could yield either outcome, and find that simple promoters respond similarly to many of the downstream genes of Crz1, with a higher output in response to pulsed TF inputs. Surprisingly, a more complex model is needed to explain the phenotype seen for one Crz1 gene and the genes profiled for other TFs, where the gene is more efficiently induced by constant inputs.

In the second study, we extend CLASP to mammalian cells, and show that it is also capable of regulating localization of multiple TFs across multiple cell lines. We focus on a single TF, RelA, in mouse fibroblasts, and perform RNA-seq to measure the effect of its spatiotemporal dynamics on all genes. Additionally, we measure the effect of TNF-alpha, a common environmental stimulus used to modulate RelA spatiotemporal dynamics. First, we find that TNF-alpha regulates many genes, even when RelA is not expressed in cells, and that this regulation is extremely different from that caused by RelA-CLASP. Secondly, we find that RelA-CLASP activates many downstream genes, and that these genes can respond differently to pulsed and constant inputs. Critically, this is a novel demonstration of the ability of RelA to regulate downstream genes using only translocation, without post-translational modifications or activation of other transcriptional regulators. Using a simple model of gene expression to simulate a subset of genes in response to constant inputs, we qualitatively predict the response to pulsed inputs for some genes. However, more complex models are needed to explain others.

Together, these studies provide direct demonstration across eukaryotes of the importance of spatiotemporal TF dynamics in regulating gene expression. Additionally, these studies demonstrate an integrated approach of engineering novel tools, measuring dynamic gene expression, and modeling of promoter activation to elucidate how genes respond to complex inputs.

Cells routinely deploy a spectrum of specialized counter-measures to deal with stress. However, a near constant feature of response to stress is a down regulation or arrest of the cell cycle. It has been widely assumed that this cell cycle modulation facilitates a faster cellular adaptation, a hypothesis that has been posited but still awaits strong support. Here, we used the \textit{S. cerevisiae} hyperosmotic shock response as a model system to pursue the role of cell cycle arrest in stress adaptation. We found that decoupling the stress response program from the cell cycle led to an adaptive program that was accelerated compared to the wild type strain. Furthermore, we uncovered that stress-induced cell cycle arrest restricted the utilization of metabolic resources, whereas unabated cell cycle progression diverted a surplus into the stress adaptation response. To probe why a wild type cell would ration its metabolic resources for an ostensibly suboptimal response, we identified an adaptive tradeoff that makes this rationing of resources advantageous under repeated stress. Given the universality of cell cycle regulation in stress responses, we suggest that this study might present a new paradigm in which the coordinated action of stress responses, cell cycle and metabolism help cells navigate the tradeoffs between rapid stress adaptation and long-term robustness.

Cellular decision making is accomplished by complex networks, the structure of which has traditionally been inferred from mean gene expression data. In addition to mean data, detailed quantitative measures of distributions across a population can be obtained using techniques such as flow cytometry that measure expression in single cells. The resulting distributions, which reflect a population's variability or noise, constitute a potentially rich source of information for network reconstruction. A significant portion of molecular noise in a biological process is propagated from the upstream regulators. This propagated component provides additional information about causal network connections. Here, we devise a procedure in which we exploit equations for dynamic noise propagation in a network under non-steady state conditions to distinguish between alternate regulatory relationships in a network. We test our approach in silico using data obtained from stochastic simulations as well as in vivo using experimental data collected from synthetic circuits constructed in yeast.

Transcription factors are key mediators of environmental signals. In \textit{S. cerevisiae}, cells may change the concentration, phosphorylation state, binding partner, or temporal dynamics of transcription factors to mediate a response to upstream signals. Dynamic localization of transcription factors to and from the nucleus have been observed for a number of stress-responsive transcription factors in yeast. Two key questions arise from this observation. First, what are the upstream determinants of these nuclear localization events? And second, what are the downstream target gene interpretations?

To probe the upstream determinants of nuclear localization, the activity of PKA, an upstream regulator of the transcription factor Msn2, was perturbed using a combination of genetic mutations and optogenetics. Optogenetic activation with bacterial Photoactivatible Adenylyl Cyclase (bPAC) established a causal connection of PKA activity on Msn2 dynamic nuclear localization. Genetic deletions of PKA network components illustrated that changes to the upstream signaling network can alter Msn2 localization dynamics. The computational analyses revealed that a negative feedback on PKA could explain the observed transient dynamics of Msn2 nuclear localization, and that different PKA network components could tune those Msn2 localization dynamics.

To probe the second question of the downstream target gene interpretations of nuclear localization, another optogenetic tool was optimized and used to directly control transcription factor nuclear localization. CLASP (Controllable Light Activated Shuttling and Plasma membrane sequestration), was engineered to enable precise, modular, and reversible control of TF localization using a combination of two optimized LOV2 optogenetic constructs that sequestered the TF to the plasma membrane in the dark and delivered the TF to the nucleus in the light. CLASP achieved minute-level resolution, reversible localization of many TF cargos, large dynamic range, and tunable target gene expression. Dynamic control of the nuclear localization of Crz1, a naturally pulsatile TF, using CLASP revealed that some Crz1 target genes respond more efficiently to pulsatile TF inputs than to continuous inputs, while others exhibited the opposite behavior. Computational modeling showed that efficient gene expression in response to short pulsing required fast promoter activation and slow inactivation, and that the opposite phenotype can ensued from a multi-stage promoter activation, where a transition in the first stage was thresholded. These data directly demonstrate differential interpretation of TF pulsing dynamics by different genes, and provide plausible models that could achieve these phenotypes.

In summary, this work explored dynamic signal encoding and decoding of transcription factor nuclear localization using optogenetics in \textit{S. cerevisiae}. Control of PKA activity and deletion of PKA network components showed encoding of dynamic transcription factor nuclear localization by the PKA network. Arbitrary control of transcription factor nuclear localization identified differential decoding of transcription factor dynamics by target genes.

Biological tissues are made up of many individual cells that perform different tasks in concert to carry out sophisticated and essential processes. Knowing the behavior of each cellular subpopulation in a tissue can therefore provide fundamental insight to how a tissue's normal functions are mis-regulated in disease. The recent advent of single cell RNA-sequencing has enabled us to measure the gene expression profiles (a proxy of cellular behavior) of the individual cells that constitute a heterogeneous tissue.

Single cell RNA-seq has already been used to identify new cell types in heterogeneous tissues, however its further application is limited by the need for analyses designed specifically for this new data type. Because single cell RNA-seq samples gene expression profiles from the underlying tissue sample, the result is distribution of global gene expression states. Analyses developed for handing individual gene expression profiles from bulk RNA-sequencing are not applicable to this new data type, nor are they capable of handling the increased measurement noise or tissue heterogeneity.

In this thesis we present new statistical analyses for analyzing gene expression distributions within tissue samples and comparing them across tissue samples. These analyses share a common approach, to exploit a natural property of transcriptional systems to reduce the complexity of this data. Through a survey of over 500 datasets, we find that global gene expression profiles can be accurately represented as a linear combination of a relatively small number of gene expression ``programs''.

In order to represent heterogeneous tissues, infer a statistical model of the underlying distribution of single cell gene expression states in a tissue. Using low dimensional representations, we can now use Gaussian mixture models to fit a distribution to each cellular subpopulation within a tissue. This unbiased model introduces a natural sense of distance between heterogeneous tissue samples, can be used to identify patient specific signatures, monitor disease progression through treatment, and classify disease state from just 15 single cell transcriptomes. Throughout these applications, our model also reveals biological insights including shifts in myeloid cell abundances, MHC I downregualtion for immune evasion, and a clonally expanded tumor population.

The ability to self-regulate in feedback is what allows cells to operate robustly despite the uncertainties of both a changing external environment and their own internal biochemistry. While feedbacks underlie many dramatic cellular behaviors including cell fate decisions and chemotaxis, the vast majority of feedback is responsible for more subtle changes such as quantitatively shaping the time scales and strength of a signaling response. In these cases, the traditional genetic perturbations such as gene knockout or overexpression that are commonly used to identify feedback regulators are ill-suited for investigating their function, in part because feedback is inherently dynamic. The aforementioned static perturbations cannot test hypothesis with respect to when and how much feedback is needed for normal cellular function. This is further compounded by the system level nature of feedback as these genetic perturbations to a feedback regulator can be obscured by compensatory changes occurring in the system the feedback regulates.

In this work, we address these issues by developing a new method, closed loop optogenetic control (CLOC), for systematically determining the temporal requirements of feedback regulators of a signaling pathway. This method, which uses control theory to extend the classical genetic compensation framework to the study of feedback, relies on the ability to precisely and in real time control the activity of a feedback regulator. This is accomplished through the use of a custom built hardware and software infrastructure that allows for \textit{in silico} control of an optogenetically activated version of the feedback regulator. The platform monitors in real time the output of a pathway deleted for a feedback regulator and automatically calculates and delivers the appropriate light input needed activate the optogenetic feedback regulator and compensate for the effects of the feedback deletion. The time varying optogenetic input needed to rescue wild type signaling dynamics serves as proxy for defining the temporal requirements of a feedback regulator in the context of native pathway signaling. In chapter 2, previous work related to in silico control of intracellular processes is reviewed. In chapter 3, I introduce the idea of studying signaling pathways using non-traditional, quantitative genetic perturbations by investigating the effect of graded expression of pathway regulators on yeast mating pathway signaling. In chapter 4 I develop CLOC as a method and use it to study three negative feedback regulators of the yeast pheromone mating pathway: SST2, MSG5, and GPA1. Surprisingly, CLOC revealed distinct dynamic requirements for the expression of each of the three feedback regulators.

The endoplasmic reticulum (ER) is the compartment in eukaryotic cells in which secreted and membrane-spanning proteins are folded, modified and assembled. These proteins are the means by which cells sense and respond to their environment and neighboring cells. Thus it is vital that these proteins fold into their appropriate structures so they can function properly. When cells are exposed to various environmental stresses, mutations or differentiation cues, the ER protein folding machinery can become overwhelmed, a condition known as ER stress. During ER stress unfolded and misfolded proteins accumulate in the ER, eliciting an intracellular signaling pathway called the unfolded protein response (UPR) that functions as a feedback loop to restore homeostasis to protein folding in the ER.

The UPR is an ancient pathway found in all organisms with an ER. The most conserved component of the UPR is Ire1, an ER-resident transmembrane protein that contains a domain in the ER lumen that senses unfolded proteins and effector kinase and RNase domains in the cytosol that initiate the response. In most organisms misfolded proteins in the ER activate Ire1 to initiate a nonconventional mRNA splicing reaction. Splicing results in the production of a transcription factor that induces UPR target genes to increase the folding capacity of the ER and thereby relieve the stress.

In the work compiled in this thesis, we find that maintaining homeostasis in the ER requires not only that Ire1 activates specifically, but also that Ire1 deactivates efficiently. Activation and deactivation of the UPR are regulated at the level of Ire1 oligomerization on both sides of the ER membrane. Binding to unfolded proteins drives oligomerization while de-oligomerization is aided by binding to the chaperone protein BiP in the lumen of the ER (Chapter 2) and Ire1's kinase activity in the cytoplasm (Chapter 3). Unmitigated UPR signaling imposes a fitness cost on cells (Chapter 4) underscoring the need for such deactivation mechanisms. Similar deactivation mechanisms may play underappreciated roles in controlling activity in many signaling pathways and stress responses.

Complex emergent biological behavior requires coordinated activity among an entire genome’s biomolecules. Cells respond to changes in environmental conditions by activating signaling pathways and gene expression programs to maintain homeostasis. Here, we first present a synthetic tool that can either dynamically regulate the expression levels of two genes independently or regulate the log mean and variance of a single gene’s expression. We then present the use of similar tools to perturb and produce a large dataset of perturbed, whole genome, gene expression measurements to explore the relationship between environmental stresses, kinase signaling, and global gene expression in S. Cerevisiae. With these data, we reconstructed canonical stress pathways, identified examples of crosstalk among pathways, and implicated numerous kinases in novel environment-specific roles. Specifically, we investigated how individual kinases tuned the magnitude of induction of the environmental stress response (ESR) in environment-specific ways. Our findings suggest that the ESR integrates inputs from multiple sensory kinases to modulate gene expression and growth control.

Great strides have been made in understanding the evolution of gene regulation by focusing on the budding yeast. The budding yeast have similar lifestyles across a wide timespan of evolution and, like their cardinal member the model single-celled eukaryote, are relatively easy to culture in the lab. The phylum also has a long and growing list of members whose genomes have been fully sequenced. One important direction of study that took advantage of the status of budding yeast as a model phylum was the comparative study of gene expression data across species in response to a variety of stresses. A key result from those investigations was that there was a major shift in the set of genes activated under a broad set of stress conditions, termed the Environmental Stress Response, that coincided with an ancient interspecies hybridization, known as the yeast Whole Genome Hybridization (WGH), which doubled the number of chromosomes in an ancestor of S. cerevisiae. Although that result provided the field with a rich dataset and robust framework for identifying evolutionary shifts in gene programs, as well as some specific intriguing examples of shifts of entire programs of similarly regulated genes, The specific nature of the genetic rewirings that gave rose to these changes have yet to be illuminated.

We took advantage of unique tools for analyzing the PKA pathway, namely the ability to monitor pathway activity through nuclear localization of the transcription factor Msn2, as well as our ability to construct a mutant strain in which PKA could be inhibited directly to compare signaling and transcriptional responses in this pathway across two species that straddle the phylogenetic divide of the WGH, S. cerevisiae, and K. lactis.

In Chapter 2, by tracking nuclear localization of the stress responsive transcription factor Msn2 (or SC.Msn2) and its ortholog in K. lactis (KL.Msn2). We observed that, while S. cerevisiae senses a switch from glucose to sorbitol as a stress, K. lactis does not. We also saw that Nuclear localization of KL.Msn2 in S. cerevisiae was responsive to osmotic shock, but it was not responsive to PKA inhibition. Furthermore, replacing the nuclear localization sequence of KL.Msn2 with that of S. cerevisiae, resulted in responsiveness to PKA inhibition. By creating a K. lactis strain in which PKA could be directly inhibited we were able to see that PKA inhibition does cause KL.Msn2 nuclear localization in its native environment in a K. lactis cell. These results showed that nuclear localization of Msn2 in response to PKA inhibition is preserved in S. cerevisiae and K. lactis, but that the conditions that cause PKA inhibition between species are not preserved, and the specific link between PKA inhibition and Msn2 nuclear localization is altered between species.

In Chapter 3, by comparing the transcriptional profiles of S. cerevisiae strains in which PKA was inhibited, we identified substantial set of paralogs from the yeast WGH, termed ohnologs that were differentially expressed under PKA inhibition. We determined using similar data collected in K. lactis and with data from stress-based gene expression studies in multiple budding yeast species that a high-basal level with no regulation by PKA is likely to be the ancestral phenotype for most of these differentially expressed ohnologs. We then studied the emergence of their responsiveness to PKA inhibition by analyzing patterns of the appearance of Msn2/4 binding sites in the promoter sequences of a range of recently sequenced budding yeast species. These analyses revealed that regulation by PKA arose in some ohnolog pairs in one of the two parental lineages prior to the WGH, leading to differential expression afterwards. In other examples, regulation by PKA and differential expression appears to have arisen following the WGH. This result provides a clear illustration of the unique opportunities presented by a WGH event for generating functional divergence by bringing together two parental lineages with separately evolved regulation into one species. Importantly, it poses the idea that functional divergence of two paralogs can be facilitated through such regulatory divergence, which can persist even when functional differences are erased by gene conversion