We are interested in how genetic circuits, composed of interacting genes and proteins, enable individual cells to make decisions, oscillate, and communicate with one another. To learn about these issues, we develop and use several experimental techniques:

• We build our own synthetic genetic circuits and study their functions.  These synthetic circuits are simpler counterparts to the complex circuits one finds in nature.  This approach is often called “synthetic biology.” 
• We make time-lapse movies to quantitatively observe dynamics of natural and synthetic genetic circuits in individual cells.  These experiments take advantage of multiple fluorescent proteins to observe several parts of a circuit simultaneously in the same cell.
• We study variability within cell populations, and try to understand how genetic circuits generate variability, through intrinsic noise, use such variability (for differentiation), or operate reliably in spite of variability. 

Projects in the lab also make extensive use of relatively simple theoretical models of genetic circuits.

Examples of recent projects:

• Cell fate decision-making: How do cells make random decisions about whether and when to differentiate? And once they do decide, how do they ensure that differentiation proceeds in an orderly fashion? Both questions hinge on the ability of genetic circuits to manage fluctuations, or noise, within their own components. Noise enables genetically identical cells to make different decisions in the same environment, effectively "rolling the dice." Recently, we have begun to use Bacillus subtilis as a model organism to address these questions. In B. subtilis, competence is a transient differentiated state in which cells can take up extracellular DNA. The decision to become competent is probabilistic and occurs in at most 10-20% of cells. Using time-lapse fluorescence microscopy movies, we analyzed the dynamics of the genetic circuit controlling competence at the single-cell level (Süel et al, 2006). Our results suggest that entry into competence and subsequent exit from it are controlled together by a core module of three genes which generate noise-excitable dynamics in a cell-autonomous fashion. (An excitable system, such as a neuron, is one in which a small perturbation can generate a well-defined response, such as an action potential). These results show that cells have evolved a dynamical mechanism which allows them to regulate the probability of competence much as a neural system can control the firing rate of action potentials.

• Tuning and re-wiring of differentiation circuits: How does the behavior of a genetic circuit depend on its architecture, quantitative parameter values, and noise?  We study this question using the example of competence differentiation in B. subtilis (Süel et al, 2007).  By re-wiring the circuit, systematically perturbing the basal expression levels of key genes, and reducing global noise levels, we have been able to identify new principles underlying the operation of this cell fate decision system.   For example, we found that the system reliably maintains the ability to transiently differentiate across a wide range of parameter values.  At the same time, it exhibits tunability, by which the probability and duration of differentiation events can be quantitatively and independently adjusted using gene expression levels.  Under other expression levels, cells even exhibit qualitatively different behaviors such as oscillation.  These results show that the system possesses evolutionary plasticity.  The wild-type and potential behaviors can be understood together in the context of a stochastic model of the underlying circuitry. Other techniques like re-wiring of this circuit, showed that the precision of differentiation events can be increased, while in vivo noise reduction provides strong evidence that differentiation actually depends on fluctuations in this system.  Together these results show how a cell fate decision system can be understood quantitatively at the single-cell level, and provide a framework for tackling analogous phenomena at other levels of biological organization.

• Synthetic biology: One example of this approach is the Repressilator, a synthetic oscillatory network constructed in the bacteria Escherichia coli (Elowitz & Leibler, 2000). The Repressilator is designed to cause oscillations in the level of gene expression over time in individual cells. It consists of a negative feedback loop of three transcriptional repressors. When combined with a green fluorescent reporter gene, the Repressilator causes growing E. coli cells to flash periodically, or twinkle, demonstrating that oscillations can be genetically programmed. Interestingly, these programmed oscillations are far less regular than those of natural cellular clocks, such as the circadian clock that operates in many organisms. We are interested in how natural biological clocks behave so reliably, and conversely, in understanding what, if anything, limits the accuracy of synthetic genetic clocks.  Another example is a recent study of the functions generated by a library of 'random' genetic circuits (Guet et al, 2002).
  

• Noise and gene regulation: A second example is our recent studies of stochasticity, or "noise," in gene regulation: (Elowitz et al, 2002) and (Rosenfeld et al, 2005). Because cells are small and contain few copies of certain molecules, stochastic fluctuations in biochemical reactions are expected to be significant, and may in fact be the origin of much cell-cell variability. We developed an experimental technique that enables detection of gene expression noise in vivo, using two differently colored fluorescent protein genes under the control of identical regulatory sequences in the same cell (see figure). In this image, noise causes individual cells to appear reddish or greenish, rather than yellow, which is the color they would be without noise (yellow is equal parts red and green). This approach should contribute to a quantitative understanding of how genetic elements function in the intracellular milieu. In (Rosenfeld et al, 2005) we use time-lapse movies to understand the biochemistry of gene regulation at the single cell level. In this study, we found that extrinsic noise can have a very slow correlation time -- that is, a long memory. Fluctuations persist for timescales on the order of the cell cycle time, placing fundamental limits on the accuracy of gene regulation.