Quantitative Analysis of Regulatory Networks
Summary: Erin O'Shea's lab focuses on (1) understanding how gene regulatory networks encode and decode information to control gene expression and (2) investigating the function and mechanism of oscillation of a three-protein circadian clock. In a new project, she is studying the mechanism of drug action, making use of a human cell line amenable to loss-of-function genetic approaches.
Quantitative Studies of Transcriptional Control
Our work in the area of transcriptional control is motivated by a desire to understand the relationship between transcription factor activity and gene expression output—the gene regulation function. We anticipate that a quantitative understanding of gene expression output will require integration of many inputs: transcription factor levels and binding constants, competition and cooperation between transcription factors, transcription factor dynamics, network structure, and nucleosome occupancy (using approaches such as genome-wide analysis and evolutionary comparisons). We use a small-scale approach in which we focus on investigating the transcription factor activity–gene expression output relationship in single genes. We also employ a large-scale approach in which we ask what information is needed to predict transcription factor binding and function in the genome.
Genome-wide studies of transcription networks and their evolution. For most transcription factors, high-affinity binding sites occur far more frequently in the genome than instances of binding or regulation by these factors. What determines where transcription factors bind, and what influences whether such binding is functional? We used genome-wide approaches to investigate how chromatin, transcription factor competition, and cooperativity influence the binding and regulatory landscape of the budding yeast phosphate-responsive transcription factor Pho4. We demonstrated a role for competition between Pho4 and Cbf1, two transcription factors belonging to the same family but regulating distinct cellular responses. Competition suppresses "leaky" gene activation in conditions where the phosphate-responsive signaling pathway is turned off, and also prevents Pho4 from activating expression of genes outside the phosphate regulon when the pathway is activated. Cooperative interactions with the transcription factor Pho2 ensure that Pho4 can overcome competition from Cbf1, but only when the phosphate-responsive signaling pathway is activated. Taking into account nucleosomes, competition, and cooperativity, we can predict on a genome-wide level the sites to which Pho4 binds and the genes it regulates.
Building on this work, we are exploring how the balancing forces of competition and cooperativity arose in the Ascomycota fungi. Through a comprehensive study of a panel of yeast species, we will deduce if these mechanisms of transcriptional control are ancestral traits. In doing so, we will uncover how such control mechanisms arose in these eukaryotes. We will use an approach analogous to our study of cooperation and competition in Saccharomyces cerevisiae and extend it to the study of yeast species selected from different clades within the phylogenetic tree. Our preliminary observations suggest that Pho2 plays a much less important role in phosphate-responsive gene expression in a related, pathogenic yeast species.
Encoding and decoding information in transcription factor dynamics. The dynamics of activation of signaling pathways can have an important influence on physiological outcomes. Many transcription factors undergo dynamic changes in subcellular localization in response to alterations in signaling, and for some factors these dynamics influence gene expression. The budding yeast general stress-responsive transcription factor Msn2 is activated and translocates to the nucleus in response to stress. We have shown that different types and amounts of stress elicit different dynamics of Msn2 activation, modulating the amplitude, duration, and/or frequency of nuclear translocation. We also demonstrated that Msn2 target genes respond differently to different dynamics of Msn2 nuclear localization. Our work suggests that multiple environmental signals can be encoded in a single transcription factor and then decoded differently by different promoters to generate stimulus-specific gene expression patterns.
Our current work focuses on understanding how dynamic information is encoded and decoded by transcription factors and by promoters. We are investigating how transcription factors process dynamic information from upstream protein kinases to learn what behaviors are possible—signal tracking, filtering, and/or integrating—and what factors determine the signal-processing outcome. For this work we also use Msn2 and its regulatory protein kinase PKA as a model system. We are also studying how promoters process information encoded in the dynamics of transcription factor activation to identify parameters that determine differences in promoter response to dynamic transcription factor activation.
Reprogramming transcription in response to stress. Cells undergo rapid and extensive changes in transcriptional programs in response to stress. We study the osmotic stress response in budding yeast as a paradigm for understanding these large-scale changes. When challenged with osmotic shock, S. cerevisiae induces hundreds of genes, despite a concurrent reduction in overall transcriptional capacity. The stress-responsive MAP kinase Hog1 activates expression of specific genes through interactions with chromatin-remodeling enzymes, transcription factors, and RNA polymerase II (Pol II). We have shown that RNA Pol II undergoes a large-scale redistribution from housekeeping to stress genes; this redistribution requires Hog1. Decreased RNA Pol II occupancy is the default outcome for highly expressed genes upon stress, and stress-responsive genes bypass this default through the selective recruitment of Hog1 and RNA Pol II by the stress-responsive transcription factors Sko1 and Hot1. The combination of reduced global transcription with a gene-specific override mechanism allows cells to rapidly switch their transcriptional program in response to stress. Our working model is that Hog1 modifies and/or associates with RNA Pol II upon stress, preventing RNA Pol II from being deployed to housekeeping genes and enabling its recruitment to stress-response genes. Our current work is focused on evaluating this model.
The Cyanobacterial Circadian Clock
Circadian clocks are endogenous molecular oscillators that help organisms coordinate their physiology with the environmental light-dark cycle. Oscillations in circadian clocks are thought to derive from a feedback mechanism involving transcription and translation. In contrast, the cyanobacterial circadian clock is a post-translational oscillator that can be reconstituted in vitro with three purified proteins (KaiA, KaiB, KaiC) and ATP. Moreover, the oscillator is phosphorylation-based: the phosphorylation state of the KaiC protein serves as a readout of clock function in vivo and in vitro, as it oscillates with circadian periodicity.
The mechanism of timekeeping by a three-protein circadian clock. We demonstrated that oscillations in the in vitro system arise from an ordered cycle of phosphorylation of two amino acids on KaiC and feedback generated by interactions of one of the phosphorylated forms of KaiC with KaiA and KaiB. We measured the rates of interconversion of different phosphorylated forms of KaiC and used these measurements to parameterize a simple mathematical model. Strikingly, this model was able to reproduce oscillations with the expected period, suggesting that it captures the basic mechanisms underlying oscillations.
One remarkable feature common to all circadian oscillators is temperature compensation: the 24-hour period is nearly invariant to temperature over a wide physiological range. Through a series of biochemical experiments and computational modeling, we are investigating the mechanism underlying temperature compensation in the cyanobacterial circadian clock.
Although the Kai proteins of the cyanobacterial circadian clock are sufficient to generate oscillations in vitro, an additional layer of complexity exists in the organism: the post-translational oscillator is coupled to a transcription-translation feedback system. We are investigating the role of the transcription-translation feedback and the post-translational oscillator in vivo by analyzing circadian oscillations in single cyanobacterial cells grown in a microfluidics device and monitored using fluorescence microscopy.
Oscillator input: How is information about the environment communicated to the oscillator to produce a phase shift? A feature common to all circadian clocks is the ability to alter the phase of oscillations to match those of the environment. Cyanobacteria are obligate phototrophs and, in vivo, the phase of the clock can be shifted by a dark pulse. We have shown that changes in light levels trigger changes in the ADP/ATP ratio in the cell, which in turn directly affects the kinase activity of KaiC to cause a phase shift.
We are currently characterizing the mechanism of the phase shift in single cyanobacterial cells, focusing on a quantitative description of the phase-shift process. We seek to understand the parameters that influence whether the clock filters out a change in light intensity or responds to it. These experiments will exploit our ability to grow and monitor lineages of single cyanobacterial cells in microfluidics devices.
Clock output: How does the oscillator control circadian gene expression? The cyanobacterial circadian clock coordinates cell physiology with the environmental light-day cycle, in part by controlling gene expression. Genetic studies have identified several proteins that are required for circadian gene expression. Our results support the model that the core oscillator controls chromosome compaction/supercoiling, which in turn influences the transcription of a large number of genes. A two-component regulatory system, SasA-RpaA, interacts with the core oscillator and is required for circadian gene expression. We are studying the function and regulation of RpaA, a putative DNA-binding response regulator. To test the hypothesis that RpaA binds to DNA and regulates genes important for circadian gene expression, we have identified genomic targets of RpaA, which we are evaluating for their role in supercoiling and gene expression. Additionally, genes have different transcriptional responses to changes in supercoiling: for some, the transcription rate increases when the chromosome becomes more negatively supercoiled, and for others the transcription rate decreases. We are investigating the basis for these different responses by analyzing regulatory regions and sigma factors to identify the sequence elements and binding proteins that characterize a response. This latter objective requires reasonably complete knowledge of operon structure, which we have recently determined experimentally through a combination of RNA sequencing and microarray analysis.
Although the best-characterized clock output is the control of gene expression, it is likely that the clock controls other physiological processes. To gain insight into regulators of the clock and processes regulated by it, we have initiated an unbiased biochemical study to identify proteins that interact with known clock components. It is our hope that this work will reveal novel connections between the clock and other biological processes.
Mechanism of Drug Action
Although ~1,000 drugs have been approved for use in humans, our understanding of the mechanism of drug action and the determinants of drug sensitivity is incomplete. To gain insight into these issues, we are using a genetic approach pioneered by Thijn Brummelkamp (Whitehead Institute) and his colleagues that enables loss-of-function genetic screening in a human cell line. Since this cell line has a largely haploid karyotype, we can use retroviral gene-trap technology to generate loss-of-function perturbations. We then select for mutants resistant to the cytotoxic effects of drugs and use high-throughput sequencing to identify the genes that are inactivated. Our current efforts are focused on screens with statins, bisphosphonates, biguanides, and chemotherapeutics.
As of August 27, 2012