Basic cellular functions, from gene regulation to cellular differentiation, are carried out by circuits of interacting genes and proteins. Increasingly, information about the identities of required genes, and their mutual interactions (i.e., who is doing what to whom), is becoming available. However, it often remains mysterious why a particular circuit architecture has been selected for a given process. My lab addresses this general problem using three key approaches: First, we use time-lapse microscopy and image analysis to monitor the dynamics of genetic circuits in individual cells quantitatively. Second, to understand what circuit designs are sufficient for generating particular cellular behaviors, we engineer and analyze synthetic genetic circuits (synthetic biology). Finally, we use mathematical models to analyze the dynamics of actual and potential genetic circuit architectures. We apply these approaches in bacteria, yeast, and mammalian cell culture, model systems that enable precise manipulation and measurement of the dynamics of cellular gene circuits.
In choosing genetic circuits to study, we focus especially on cases in which cell-cell variability plays an important role. Cells can be strikingly heterogeneous, even when they are genetically identical, and even when they are grown in a homogeneous environment. Cells exhibit stochastic fluctuations, or "noise," in the concentration of proteins. More dramatically, seemingly equivalent cells differentiate into diverse fates in response to the same cues. Nongenetic variability appears to be a key attribute of cells in both microbial and multicellular organisms, and to play a key role in the operation of gene circuits that underlie cellular functions. My lab focuses on fundamental questions related to this issue: How do genetic circuits generate and modulate variability in particular cellular components? Conversely, how does variability enable specific genetic circuits to function reliably and to generate probabilistic behaviors? Finally, what role does variability play in gene circuit evolution? These questions are critical for understanding the structure, function, and evolution of genetic circuits in both microbes and multicellular organisms. While difficult to address using traditional biochemical techniques that average over cell populations, they can be explored at the level of individual cells.
Gene Regulation at the Single-Cell Level
Consider, for example, the process of gene regulation in bacteria. The in vivo relationship between the concentration of a transcription factor in a cell and the activity of its target promoter is of central importance for understanding and modeling genetic circuits. However, it has been unclear how stochasticity in gene expression and other sources of noise impact this process at the single-cell level. To address this problem, we used quantitative time-lapse microscopy and image analysis to track transcription factor concentration and target promoter activity simultaneously in individual cells. The resulting movies provided quantitative measurements of the origin, amplitude, and timescale of noise, both in the expression of the target gene (intrinsic noise) and in the regulation process itself (extrinsic noise). These movies alone enabled us to determine effective biochemical parameters of the system in molecular units. More critically, they revealed two different timescales in gene regulation: Intrinsic noise (stochasticity) in gene expression fluctuates rapidly (and can be "averaged out" by cells over longer timescales). However, extrinsic noise, which results from the combined effects of fluctuations in many cellular components, such as ribosomes and polymerases, varies on the slower cell cycle timescale. Because they do not rapidly average out, these slow extrinsic fluctuations fundamentally limit the accuracy of gene regulation. We recently showed that the resulting model of gene regulation can predict the behavior of simple synthetic gene circuits that we constructed in the lab. We are now using related techniques to understand the characteristics of signaling systems in mammalian cells.
Transient, Probabilistic Differentiation
Some differentiation processes appear to be probabilistic, occurring spontaneously in some cells but not others. What accounts for the decision of an individual cell to differentiate, or not? What gene circuit architectures enable cells to regulate the probabilities of differentiation? To address these issues, we turned to Bacillus subtilis, a model bacterium that can differentiate into distinct cell types. In isogenic populations of B. subtilis, it was shown by David Dubnau (Public Health Research Institute Center) and others that a small fraction of cells spontaneously differentiate into the alternative cell fate of genetic competence, in which they can take up extracellular DNA. After some time, cells revert to their original vegetative state. Thus, an essential aspect of competence is that it is both probabilistic and transient. We used genetic perturbations and time-lapse movies to identify a core gene circuit that explains this behavior and to show that the key principle underlying the design of the system is excitability.
The competence control circuit consists of a fast autoregulatory positive-feedback loop and a slower post-translational negative-feedback loop that together give the circuit the property of excitability. An excitable system, such as a neuron, is one in which a small perturbation, or noise, can generate a larger stereotyped transient response. In a neuron, this stereotyped response takes the form of a fast action potential. In B. subtilis, it consists of a much longer (many hours) episode of transient differentiation. Mathematically, the two systems are similar, despite their very different molecular components and timescales. In the case of competence, noise-induced excitability can explain both the probabilistic and the transient characteristics of competence in a unified way.
Tuning Transient Differentiation
The analysis of this transient differentiation system raised a basic question: What is the advantage to the cell of an excitable gene circuit compared to alternative designs? One possible answer is to facilitate independent tuning of the frequency and duration of differentiation episodes, characteristics that could need to vary differently in different environments. In addition, the excitable circuit architecture can also give rise to a repertoire of behaviors besides excitability, including oscillations and bistability. Systematic perturbation of the gene circuit away from its wild-type parameter values revealed these alternative behaviors. In fact, the circuit can even be made more precise: "Rewiring" it (altering its regulatory connections) by adding an additional negative-feedback loop reduces variability in the duration of differentiation events. Competence provides a wonderful paradigm for understanding how a relatively simple gene circuit can exhibit a repertoire of behaviors accessible with relatively minor genetic perturbations. Our work on genetic competence is a collaboration with Gürol Süel (University of Texas Southwestern Medical Center at Dallas) and Jordi Garcia-Ojalvo (Escola Tècnica Superior d'Enginyeries Industrial i Aeronàutica de Terrassa, Spain)
Frequency Modulation and Coordination of Gene Regulation
Recently, we have begun exploring the role of circuit dynamics in eukaryotic signal transduction and gene regulation. Cellular responses to external signals involve two steps. First, signals must be represented in the cell in the concentrations, states, and dynamics of transcription factors. Second, these active transcription factors in turn regulate the expression of genes. We recently completed an initial study of signal encoding and gene regulation during calcium stress response in yeast. This work built upon the discovery, by the groups of Martha Cyert (Stanford University) and Kyle Cunningham (Johns Hopkins University), of the calcium response system in yeast. In this system, the calcineurin-responsive zinc finger transcription factor Crz1 is dephosphorylated by calcineurin in response to calcium. When dephosphorylated, it transits to the nucleus, where it can activate more than a hundred target genes.
We recorded movies of individual yeast cells in which Crz1 was fused to a fluorescent protein. These movies revealed that a step change in extracellular calcium levels causes rapid stochastic "bursts" of Crz1 nuclear localization, rather than a steady shift in the (otherwise static) fraction of Crz1 molecules in the nucleus. These bursts continued throughout the movie, typically about 10 hours. Analysis of bursting at different levels of calcium showed that calcium controls the frequency, but not the duration, of these bursts. In other words, the cell encodes calcium levels using a frequency modulation (FM) system.
What advantage does FM encoding provide the cell? A basic problem for cells is to coordinate multiple target genes so that they are expressed in fixed proportions across a wide range of expression levels. In bacteria, operons can perform this function for small groups of genes. How might eukaryotic cells, which lack operons, achieve coordination? FM regulation enables Crz1p to coordinate the expression of its many target genes in fixed proportions across many levels of activity. This is because calcium effectively regulates the fraction of time Crz1p is active (nuclear localized) rather than the fraction of Crz1p molecules in the nucleus, as it would in a more conventional (amplitude modulation [AM]) system. Consequently, all target genes are expressed in proportion to nuclear localization burst frequency. As frequency increases (at higher calcium levels), all target genes increase their expression by the same factor.
FM coordinates target genes even when their individual promoters differ in affinity and cooperativity with which they bind to transcription factors. Because nuclear localization dynamics are unsynchronized, and appear stochastic, this basic principle of gene regulation could only be identified using movies. We suspect that this type of regulation may play critical roles in regulation in other biological systems and contexts, from bacteria to multicellular organisms.
Future Directions
These examples show some of the ways in which variability is directly integrated into the architecture and function of genetic circuits. While we expand the use of these approaches in microbial systems, we have also embarked upon a new effort to address conceptually related problems in mammalian differentiation and development using cell culture. At the same time, in other projects we are using synthetic biology approaches in a "build-to-understand" approach: we construct novel gene circuits that implement key cellular functions, and use these circuits to understand the minimal requirements for basic cellular functions.
