Final Report Summary - OPRA (OPTIMALITY PRINCIPLES IN RESPONSES TO ANTIBIOTICS) Summary description of the project objectivesBacteria regulate gene expression in response to antibiotic stress. While many details of this response have been thoroughly mapped, key questions remain open. The long-term objectives of this project are to address the questions: How variable is the dynamical response to antibiotic stress between cells? Is the regulation of genes optimized to increase survival and growth? And how can temporal antibiotic administration patterns be designed to maximize growth inhibition and killing of pathogenic bacteria? In this project, we use a combined experimental-theoretical approach to tackle these questions, based on quantitative high-throughput experiments with computational data analysis and theoretical descriptions of key cellular response systems and their regulation. We use Escherichia coli as our main model system since it offers unsurpassed potential for precise measurement and manipulation of gene expression in living cells. The specific aims of the project are:Aim 1. Measure the dynamical transcriptional response to antibiotics at high time resolution.Aim 2. Analyze transcriptional response data using statistical data analysis approaches and develop theoretical descriptions of gene regulation responses.Aim 3. Synthetically manipulate gene expression to identify optimal and non-optimal regulation in response to antibiotics.Work performed in this projectWe have completed the key components of all aims of this project. Specifically, for aim 1, we used a genome-wide library of E. coli fluorescent transcriptional reporters to measure the global gene regulatory response to the sudden addition of well-characterized antibiotics such as trimethoprim (a folic acid synthesis inhibitor), tetracycline, and chloramphenicol (translation inhibitors). These measurements were performed at high time resolution using a customized liquid handling robot with an integrated microplate reader. Importantly, in our protocol, cultures are diluted several times during each experiment; this is done to keep the cultures in exponential growth and prevent entry into stationary phase which would interfere with the specific dynamic response to the drug we are interested in. We developed custom Matlab code to analyze these data and extract quantitative measures for the dynamic response of each gene. These measures include the maximum fold-change in expression and the response time which we defined as the time it takes to achieve the half maximum response. These data thus provide information about the global dynamic response to drugs at an unprecedented level of detail.We further developed an assay in which we study the response of selected genes in single cells. This assay uses a microfluidics chamber in which bacteria are trapped. We can then suddenly switch the environment from pure growth medium to antibiotic-containing medium. Time-lapse fluorescence imaging of the cells growing in the chamber is performed throughout these experiments. We have adapted and developed Matlab code to automatically segment and analyze the movies obtained in this way. This approach enables us to follow the response of selected genes in individual cells and thus quantify the cell-cell variability of the responses. Importantly, in this assay, we quantify not only the variability in expression level but also the variability in the response timing. A crucial aspect of this experiment is the construction of strains that carry fluorescent reporters integrated into their chromosome to avoid potential artifacts that can occur when such reporters are on plasmids. We have developed several efficient techniques to generate such constructs. In particular, these enable the construction of strains that carry fluorescent reporters with different colors for two different genes. We constructed several such strains with reporters for selected combinations of genes and used them to reveal if the dynamics of gene expression changes is coupled. To identify optimal and non-optimal gene regulation (aim 3), we further developed a separate high-throughput protocol for precise growth rate measurements. We used this technique to measure the growth rate of all strains of a genome-wide gene deletion library in the presence of different antibiotics. In addition, we developed an efficient protocol for constructing E. coli strains in which the expression level of any gene can be controlled using an inducible promoter. Specifically, these strains have a chromosomally inserted copy of the gene of interest that is under the control of a lac promoter which we induce with IPTG. To achieve complete repression in the absence of IPTG and continuously inducible expression levels, these strains further carry a plasmid with the lac repressor gene; we found that the latter needs to be highly expressed to properly control expression. We have focused on essential genes including initiation factor 2 (IF2) and folA. We have further used existing plasmids from the ASKA library, which contain all genes in the E. coli genome under an inducible lac promoter, to study the effects of overexpression on growth rate.Main resultsAnalysis of the experimental data from aim 1 revealed that different genes respond on vastly different time scales ranging from minutes to hours. We identified a clear temporal hierarchy in which certain genes respond to the stress almost immediately while others only respond with a significant time delay. We have further used statistical data analysis techniques (aim 2), specifically principal component analysis (PCA), to identify the dominant response modes of the global gene expression response to antibiotic stress measured in aim 1. This analysis revealed that for typical antibiotics, two principal components explain a large fraction of the variance in the data. Here, the first component captures a permanent shift of expression levels to a new steady state expression level. Intriguingly, the second principal component describes a pulse-like response in which gene expression levels depart from their original steady state, transiently increase to a high level, and then quickly return nearly to the original level. Performing gene ontology enrichment analysis on the genes that are dominated by each of these principal components, we identified that many of the genes showing the pulse-like behavior play a role in the acid stress response. This suggested that the intracellular pH drops as a result of antibiotic addition which could act as a trigger for a considerable part of the observed global stress response. We performed various follow-up experiments to test this idea and were able to confirm that pH can indeed drop in response to certain antibiotics. The data acquired in this project allows us to quantify effective kinetic parameters that characterize gene regulation in response to antibiotic treatments and to formulate new hypotheses for how bacteria adapt to antibiotics. These results will lead to a faithful theoretical description of bacterial gene regulation and physiological changes in response to antibiotic stress. It will enable us to reveal if a lack of coordination between different cellular systems is a typical cause of non-optimal gene regulation in the presence of antibiotics. Our results will further enable us to predict temporal antibiotic treatment strategies that are optimized to circumvent bacterial defense responses. Our results further provide the foundation for gaining a quantitative picture of the extent to which bacterial gene regulation in response to antibiotics is optimized. This will directly confront the often implicitly made assumption that microbes respond to changes in their environment in a nearly optimal way.
