A fluctuation-response relationship in the bacterial chemotaxis system
The Fluctuation-Dissipation Theorem describes the relationship between noise and response in physical systems at thermodynamic equilibrium. Based on our model, we predicted that we could infer cellular response to a stimulus from the behavioral variability in nonstimulated cells. To experimentally establish the existence of a fluctuation-response relationship in the chemotaxis system of E. coli, we used a linear approximation as a general framework for monitoring pre- and post-stimulus switching behavior of individual bacterial motors. This study highlights that under certain conditions, the fundamental fluctuation-response relationship is extensible to living cells not at thermodynamic equilibrium.
This project was the work of former graduate student Heungwon Park and was published in Nature.
Sensitivity of chemotactic response to the cellular functioning state
Using a capillary assay, we studied how the chemotactic response depends on the functioning state determined by the relative chemotactic protein expression level. We found that the chemotactic response was not robust. Instead, it was fine-tuned to various functioning states. Cellular response to the small external stimulus was highest around the average functioning state of wild-type cells.
This project has been published in Current Microbiology.
Non-mutative source of behavioral variability
In our work, we have employed the chemotaxis network responsible for the motion of E. coli bacteria as a model system for the general study of signal transduction networks. We asked whether there were specific molecular events that could cause behavioral variability in an individual cell. By developing a single-cell approach we were able to characterize the design principles of the signal transduction network in bacteria (chemotaxis). We found that the inherent randomness (noise) in the chemical reactions taking place in the intracellular signal processing was a "tuneable" source of behavioral variability. Moreover, the signal transduction network determining cellular behavior was found to be tuned to maximize its sensitivity. We used a combination of computer simulations and genetic experiments to identify the key molecular components that allow the cell to adjust signal transduction sensitivity. We found as a consequence of the network design that maximum sensitivity would be also accompanied with maximum behavioral variability, Korobkova et al., Nature (2004). We believe that such interplay between sensitivity and variability in signal transduction is shared by many other signaling systems.
The role of network design in evolution
As modern biology unveils the architecture of a growing number of large intracellular networks, their intricate topology raises a general question: Is there a specific network topology that confers some kind of evolutionary advantage? In this project, we ask how networks with distinct topologies can evolve towards a pre-established target phenotype (i.e. a specific dynamical biology) through a process of random mutations and selection.
We use Boolean networks as model systems for studying the dynamical properties and the evolutionary potential of large intricate intracellular pathways. For example, simple Boolean rules permit the modeling of real biological systems such as the cell-cycle of Yeast or the expression pattern of the segment polarity genes in Drosophila embryos. We focus on the relationship between the topology and the network's ability to evolve toward a stable target phenotype. We compare the evolutionary paths of networks with scale-free and homogeneous random topologies.
In random networks the distribution of connections is homogeneous and the average number of connections per node is the relevant topological parameter to characterize the network architecture. By contrast, the topology of scale-free networks is highly heterogeneous and the number of connections per node is power-law distributed, Aldana et al., PNAS (2003). Under this condition, the average connectivity is not physically relevant to describe heterogeneous systems, and the only significant parameter is the scale-free exponent. Experimental data shows that many biological networks exhibit the scale-free topology. The characterization of a relationship between topology and dynamical properties of large /biological/ networks primarily motivated our work. But we feel that in view of the significance of scale-free networks in other fields such as sociology and economics our results could be also interesting beyond the scope of biology. This research was published in Nature Physics.
In-silico evolution of small signal transduction networks
To explore the properties of small signal transduction networks, we want to understand how easy it is to start from known signalling modules, such as de/phosphorylation cycles and phosphate transfers, and connect them to recreate a known network that exhibits an adaptive response. We developed a coarse grained model for the evolution biological networks. We used representations of the cell that self-replicate and randomly mutate, corresponding to in silico rounds of evolution and selection. We selected for optimal adaptive behavior. We explored the different dynamical behaviors and network designs that emerge from such a evolutionary model.
This project is the work of former postdoc Panos Oikonomou.
Mar promoter activity in single E. coli cells
At the core of multiple antibiotic resistance (mar) network is the mar operon which acts as a master switch for over 100 genes. The dynamics of the mar operon likely plays a crucial role in the cell's regulation of antibiotic resistance. Because this resistance is reversible, it is difficult to study at the population level. In a single cell at steady-state, fluctuations in gene expression can arise from global factors affecting the entire cell, such as stage in the cell cycle, cell growth, and exposure to environmental chemicals, or from intrinsic factors at specific DNA loci. When intrinsic factors lead to cell-to-cell variability in gene expression, it is important to determine whether these epigenetic changes are inherited through generations. We are using single cell studies to trace the temporal evolution of mar expression across genealogical trees. We are also developing simple coarse-grain models that can capture the essential dynamical behaviors obtained from these single cell experiments.
In E. coli, the mar operon is responsible for the multiple antibiotic resistance phenotype. This operon regulates the efflux pump AcrAB-TolC, which is the main determinant of multi drug resistance. The genetics of mar has been studied solely at the population level. To understand cellular variability of resistance to antibiotics, we studied the dynamics of the mar operon at the single cell level. Using Venus-YFP, we monitor in real time the activity of the mar promoter in single cells grown in linear micro-colonies. We determined the activity of the mar promoter for different values of the inducer salicylate. In response to a steady level of inducer, mar promoter activity was widely heterogeneous between daughter cells. Heterogeneity in promoter activity varied with inducer concentration. We observed that different promoter activity levels are maintained and inherited for several generations.
This project is the work of former postdocs Calin Guet and Panos Oikonomou and former grad student Luke Bruneaux, in collaboration with Max Aldana at UNAM.
Transcriptional network regulated by global regulator MgrA in S. aureus
Recently, a multi-drug resistant strain of the human pathogen Staphylococcus aureus has emerged in hospitals and surrounding communities. To combat this resistance in methicillin-resistant S. aureus (MRSA), we must understand the mechanisms regulating virulence and drug resistance in S. aureus. Interestingly, in S. aureus, the mechanisms regulating these antagonistic behaviors are correlated. In S. aureus, the DNA-binding protein MgrA, which is in the same family as E. coli multi-drug resistance repressor MarR, is a global regulator and controls 355 genes in S. aureus. For example, MgrA represses norA, which encodes multi-drug efflux pumps, and sarV, which encodes an autolysis regulator. To understand how MgrA regulates these two antagonistic function in S. aureus, we are investigating MgrA activity in different environmental conditions. To investigate the regulatory network of virulence and drug resistance genes in S. aureus, we are examining how promoter activity changes in response to antibiotics.
This project is the work of Satoshi Nishida and Kevin Wood in collaboration with Chuan He's lab at the University of Chicago.
Real-time RNA profiling in single, living cells
Using a noninvasive fluorescence correlation spectroscopy (FCS) technique, we measured the number of RNA transcripts within a single, living E. coli cell. Work focused on the measurement of mRNA transcripts has been published in Nucleic Acids Research and work focused on measurement of non-coding RNAs has been published in PNAS.
Summary: Nearly half a century ago, the discovery of messenger RNA as an "unstable intermediate" established RNA instability as a key dynamical property in the molecular organization of life. The balance between the kinetics of synthesis and the degradation of RNA transcripts exhibits a primitive example of molecular adaptation of gene expression.This molecular adaptation also allows rapid responses to regulatory and environmental signals. But the difficulty in studying RNA dynamics lies in the real-time dissection of each dynamical process that a specific RNA transcript has to undergo. Expression levels of a specific RNA transcript extracted from ensemble measurements come generally from cells that exist in slightly different states of growth and that exhibit also different behavior due to the variations of their internal biochemical parameters. Due to the transient nature of RNA transcripts, it has been technically challenging to monitor in real-time how the response to an environment signal is directly reflected in the transcription activity of a given gene within an individual cell. We monitored the number of transcripts within a single, living E. coli cell using a series of two plasmid constructs. One plasmid (pZE31-dsRed-ms2x2) encodes RFP. Following the stop codon in the RFP sequence are two MS2 viral coat protein binding sites. The second plasmid (pZS12-MS2-GFP) encodes the fusion protein GFP-MS2. This fusion protein binds to the transcript from pZE31-dsRed-ms2x2. Because free MS2-GFP and MS2-GFP bound to the transcript diffuse at different rates, we can calculate the concentration of transcripts within the cell. Additionally, we measured the RFP concentration within the cell to confirm that transcript levels are correlated with protein expression levels.
Experimental FCS setup:
Using fluorescence correlation spectroscopy (FCS), we monitored in real-time and within a single bacterium of E. coli the dynamics of synthesis and the degradation of a specific RNA transcript. The goal of this study is to characterize the relationship between the dynamics and the design of simple transcriptional networks at the single cell level.
Real-time concentration profiles of specific RNA transcripts:
AgentCell: multi-scale simulation of signal transduction in E. coli
We developed a software framework to study the design principles of signal transduction in living organisms. We named this software AgentCell, because it uses agent-based technology to model important properties of the intracellular E. Coli chemotaxis network and the coordinated motion of bacterial cells in direct response to environmental stimuli. For further detail about this software, please see our paper in Bioinformatics.
Summary:The chemotaxis network for E. coli serves as a model system because it is well characterized and experimentally accessible. The dual level simulation models the dynamics of the signal transduction networks within single cells at one (micro-) level, and simulates the movement of the cells through a medium at the other (macro-) level. Repast is the agent-based simulation framework used for the macro-level simulation; it treats each of the cells as an individual agent. Cells (agents) interact dynamically with the environment (for example, by sensing concentrations of attractants) and with other cells. An existing molecular simulation, StochSim (D. Bray Lab), is used to model the chemotaxis signal transduction network at the micro-level. Each molecule is modeled individually as an object and each molecular reaction is modeled within a cell as it occurs over time. The likelihood of a molecular reaction occurring is proportional to the rate constants of the kinetic reaction equations. This requires millions of reactions to be simulated per second of simulated time for each cell. The higher-level Repast simulation performs several functions. Repast defines a class for the chemotaxis network simulation, and each StochSim run is represented as an instance of that class. The Repast simulation defines several other classes in addition to this to control, coordinates the cell simulations and translates the simulated chemical reactions into cell motion, either run or tumble. The motion state of the cell is translated into movement over time and space in terms of cell location and cell orientation. Finally, the Repast simulation models the motion of the cell as it moves through the representation of the larger space. The spatial representation assures adherence to spatial constraints such as boundaries, and potentially is a basis for modeling cell congestion. The Repast simulation synchronizes the flow of time across and within cells and also seamlessly coordinates the cell movement space and the molecular interaction space within cells. This project is the work of former postdoctoral fellow Thierry Emonet (Yale), in collaboration with Charles Macal and Mike North from Argonne National lab.