The green alga Chlamydomonas reinhardtii is a good biological model to study fundamental biological questions ranging from motility, photosynthesis to metabolism. As a photosynthetic cell, Chlamydomonas thrives in minimal media, carbon dioxide and light, but it can also grow heterotrophically on acetate. Unlike yeast or E.coli, any metabolite detected in the intra- or extracellular space is therefore a product of its enzymatic machinery, since only inorganic nutrients are needed. In addition, cell division of Chlamydomonas can be synchronized by light periods, so at any given time >95% of all cells in a culture are identical. Chlamydomonas batch cultures enable pursuing multiple experiments with limited efforts in small time frames, yet high analytical accuracy and precision.
We have utilized Chlamydomonas to study its metabolic responses under nitrogen depletion time courses using four different levels of N-supply and four different time points of stress response, each with eight independent replicates. Metabolite profiles were analyzed using improved methods that had median technical errors of 16% CV on 80 identified compounds using 5x10^6 cells per sample. Data were processed using an automated database approach and data were analyzed in a first pass using classic univariate and multivariate statistics. The dominant effect on metabolic variance was found to be depending on cell cycles states which are controlled by putrescine metabolism in this organism. In order to further the underlying understanding of additional metabolic perturbations by the physiological stress treatment, data were mean centered to the standard N-condition controls and analyzed for biochemical changes. Surprisingly, large and non-uniform differences could be observed for both dose and time of N-depletion on a wide number of metabolites. Data analysis was complemented by analyzing metabolic networks using Likelynet, a Bayesian likelihood method that is geared towards unbiased detection and verification of linear relationships in metabolic datasets, taking into account the technical error estimates for each variable (Kose F. et al, BMC Bioinformatics 2007, 8:162 http://www.biomedcentral.com/1471-2105/8/162 ).
Work done in collaboration with Do Yup Lee, Jan Budzcies, and Frank Kose.
Human biology functions through (largely unknown and) bewildering sets of interacting molecules. Even the parts list is incomplete - while we know that roughly 23,000 human proteins are expressed, each protein can be changed by alternative splicing, variable post-translational adducts, and protein processing. Perhaps the number of different human proteins is a couple hundred thousand, but no one really knows. The situation for RNA is worse - it is likely that in human cells every possible RNA is expressed at a low level, adding another 60,000,000 non-overlapping RNA 100-mers to the mix. If one loves the idea of a pre-biotic RNA World, one would be foolish to discount the possible functions available today in this deep human RNA sequence space. Bacteria have smaller genomes and parts lists, and have had more "genetics" - between faster generation times and higher mutation rates, bacteria are engineered today through a more complete Darwinian process than has been possible for mammals.
Thinking this way raises a question for metabolic engineering - should one design or evolve novel biochemical systems? Clunky mammals and less-clunky bacteria suggest that evolution strategies should be tried.
The deepest combinatorial selection paradigm (SELEX) has provided an astonishing set of reagents (aptamers), and at the same time provided lessons for how to manage evolution/selective strategies. Even though I'd like to lecture about the human plasma proteome (the main interest at SomaLogic), I will focus my talk on the possible value of in vitro and in vivo selections aimed at the discovery of novel biochemical systems.
Metabolic engineering has been defined as "the improvement of cellular activities by the manipulation of enzymatic, transport, and regulatory functions of the cell with the use of recombinant DNA technologies." The elucidation of the genome sequences for many microbes, fungi, animals and plants has provided a number of unique tools to tackle the challenge of engineering metabolic pathways as part of interdisciplinary efforts that integrate biology and chemistry with engineering and mathematics. However, fundamental issues remain, such as the adverse social reaction to the utilization of genetically modified organisms (GMO), the difficulties associated with predicting the effect of genetic manipulations on the metabolome, and the problems associated with targeting metabolites to the desired cellular or sub-cellular locations. These and other issues will be described, putting metabolic engineering within the perspective of alternatives (such as chemical synthesis) with the goal to provide a perspective of the opportunities and future of the field for interdisciplinary interactions.
In a first part of my talk I will briefly review relevant methods for metabolic flux analysis. This includes metabolite balancing and flux analysis using labelled substrates with necessary experimental and computational methods. I will finish my talk with two recent case studies:
An increasing number of small RNAs (sRNA) have been shown to regulate critical pathways in prokaryotes and eukaryotes. In bacteria, sRNA regulation is predominantly involved in coordinating intricate stress responses. The mechanisms by which sRNA modulate expression of its targets are diverse. In common to most is the possibility that the level of functional sRNA may be altered via its interaction with its targets. Aiming to understand the unique role played by sRNAs, we study quantitatively two classes of bacterial sRNAs in Escherichia coli using a combination of experimental and theoretical approaches. Our results demonstrate that sRNA provides a novel mode of gene regulation, with characteristics distinct from those of protein-mediated gene regulation. These include a threshold-linear response with a tuneable threshold, a robust noise resistance characteristic, and a built-in capability for hierarchical cross talk. Knowledge of these special features of sRNA-mediated regulation is crucial towards understanding the subtle functions that sRNAs play in coordinating various stress-relief pathways, and can help guide the design of synthetic genetic circuits with properties difficult to attain with protein regulators alone.
Malaria infects 300-500 million people and causes 1-2 million deaths each year, primarily children in Africa and Asia. More than half of the deaths occur among the poorest 20% of the world's population. One of the principal obstacles to addressing this global health threat is a lack of effective, affordable drugs. The chloroquine-based drugs that were used widely in the past have lost effectiveness because the Plasmodium parasite, which causes malaria, has become resistant to them. The faster-acting, more effective artemisinin-based drugs - as currently produced from plant sources - are too expensive for large-scale use in the countries where they are needed most.
We have metabolically engineered E. coli to produce high levels of mono-, sesqui-, and diterpenes, most notably the sesquiterpene precursor to artemisinin, amorphadiene. The result of these studies is an E. coli host capable of producing 1,000,000-fold higher levels of amorphadiene than the strains and expression systems that had been available previously. The engineered strain contains a heterologous mevalonate-based terpene biosynthetic pathway and an amorphadiene cyclase gene resynthesized with the E. coli codon usage. Recently, we cloned the final steps in the artemisinin biosynthetic pathway and engineered yeast to produce artemisinic acid at high levels. The development of this technology will eventually reduce the cost of artemisinin-based combination therapies significantly below their current price.
The cellular environment is abuzz with noise. Generated by random molecular events, cellular noise not only results in random fluctuations within individual cells but it is also a source of phenotypic variability among clonal cellular populations. In some instances fluctuations are suppressed downstream through an intricate dynamical network that acts to filter the noise. Yet in other instances, noise induced fluctuations are exploited to the cell's advantage. Intriguing mechanisms that rely on noise include stochastic switches, coherence resonance in oscillators, and stochastic focusing. While mathematical models of genetic networks often represents gene expression and regulation as deterministic processes with continuous variables, the stochastic nature of cellular noise necessitates an approach that models these variables as discrete and stochastic. In this framework, probability densities of the system states evolve according to a (usually infinite dimensional) Chemical Master Equation (CME). Until recently, sample trajectories have been computed almost exclusively with Kinetic Monte Carlo methods, such as Gillespie's Stochastic Simulation Algorithm. In this talk we present a new direct approach for computing the relevant statistics, which involves the projection of the solution of the CME onto finite subsets. We illustrate the algorithm underlying our Finite State Projection approach and introduce a variety of systems theory based modifications and enhancements that enable large reductions and increased efficiency with little to no loss in accuracy. Model reduction techniques based on linear perturbation theory allow for the systematic projection of multiple time scale dynamics onto a slowly varying manifold of much smaller dimension. The proposed projection approach is illustrated on few important models of genetic regulatory networks.
Advances in structural and molecular biology have spurred the proliferation of protein engineering technologies, allowing fundamental questions about protein evolution to become approachable. The questions themselves, in turn, can be the drivers for the development of new tools. I will describe how my interests in protein evolution shaped the development of structure-based combinatorial protein engineering (SCOPE), a tool for connecting evolutionary endpoints in local and global sequence space. I will first discuss the inception of SCOPE as a homology-independent recombination method and its application to create multiple-crossover libraries from distantly-related DNA polymerases, and then describe adapting the technique for combinatorial mutagenesis to recapitulate the more recent functional divergence of closely-related terpene cyclases. Throughout my discussions, I will highlight probabilistic considerations of library design, detail key experimental results and what they tell us about the evolution of specialized metabolism and implications for metabolic engineering.
Bacteria are tiny chemical factories that can sense their environment and react to stimuli according to a pre-programmed response. These responses are controlled by genetic networks. By constructing synthetic versions of these genetic networks, we can program bacteria to behave in completely novel and useful ways. In this talk, we present our work on designing bacterial edge detectors: bacteria that use cell-cell communication and parallel computation to sense where borders between lighted and dark regions exist and report the edges. We use a mathematical model combining statistical mechanics with partial differential equations to determine which alterations to the synthetic gene network will improve the bacterial edge detection.
Work done in collaboration with J. Tabor and C.A. Voigt.
Metabolic networks have a bow-tie architecture: a wide diversity of nutrients is disassembled into a few molecular currencies, which are then reassembled into a large variety of other molecules. At the "knot" of this bowtie lie cycles whereby a set of reactions transfer a molecular group (moiety) from various donor metabolites to a common carrier (metabolic currency) from which another set of reactions transfers the moiety to various accepting metabolites. These circuits couple moiety supply to demand. Their role, and the performance criteria they should fulfill, are akin to those of a power-supply unit in an electronic circuit.
The first part of this talk will explore the design principles enabling this general class of metabolic circuits to operate effectively as "moiety-supply units". The second part of the talk will examine quantitative aspects of the design of concrete biological realizations of these circuits.
Recent progress in developing frameworks for the construction of RNA devices is enabling rapid advances in cellular engineering applications. These devices provide scalable platforms for the construction of molecular communication and control systems for reporting on, responding to, and controlling intracellular components in living systems. Research that has demonstrated the modularity, portability, and specificity inherent in these molecules for cellular control will be highlighted and its implications for synthetic and systems biology research will be discussed. In addition, new tools that translate sequence information to device function to enable the forward design and optimization of new devices will be discussed. The flexibility of the specified framework enables these molecules to be integrated as systems that perform higher-level signal processing based on molecular computation strategies. The application of these molecular devices to studying cellular systems through non-invasive /in vivo /monitoring of biomolecule levels and to regulating cellular behavior, in particular in the control and optimization of biosynthesis, will be discussed.
Small noncoding RNAs are genes for which RNA, rather than protein is the functional end product. In bacteria, many small RNA genes (sRNAs) appear to act as post-transcriptional regulators by basepairing with target messenger RNAs. In this talk, we will look at computational and experimental approaches to characterize these sRNA genes in bacteria. First, we will describe high-throughput approaches for identifying sRNA genes in a bacterial genome. In particular, we will consider a probabilistic model that combines heterogeneous data sources (including primary sequence data, comparative genomics information, and microarray expression data) for the purpose of predicting sRNA genes throughout a genome. We will then investigate methods, both computational and experimental, for characterizing regulatory targets of sRNA action. Finally, we will explore how these high-throughput approaches are used to elucidate the roles of specific sRNA genes and the pathways in which the genes are involved.
Stoichiometric approaches have been tremendously successful as mathematical models in metabolic engineering. Their linearity permits an unparalleled repertoire of mathematical and computational tools, and the combination of stoichiometric models with experimental data has yielded valuable insights into flux distributions under different conditions. However, as we strive to understand the details of control and regulation in vivo at a deeper level, refined models are needed, and these must be nonlinear. While simulations with nonlinear metabolic models are no longer a significant computational hurdle, the estimation of suitable parameter values continues to be a major challenge. In this presentation I will review current approaches to metabolic parameter estimation, especially for time series data, and demonstrate why it is important to obtain fast solutions on standard computers. As an example for many aspects of my presentation, I will use the regulation of glucose utilization in Lactococcus lactis, for which we have high-precision in vivo data describing the dynamics of intracellular metabolite pools.