Dynamic Models Of Escherichia Coli Metabolism: An Accurate Description Of Our Pathetic Thinking
Open access
Autor(in)
Datum
2024Typ
- Doctoral Thesis
ETH Bibliographie
yes
Altmetrics
Abstract
The introduction of this thesis consists of two parts. In the first of these we will introduce systems biology as a field of study by placing it in the historical context of the overarching discipline of biology. A quick stroll through this history offers us a much needed and often absent perspective on the field as part of the larger body of scientific philosophy that it is part of. One might argue such positioning is paramount in any systems science whose existence predicates on the essentiality of contextualization. In this excursion we will swiftly move from the Greeks to the 20th century, touching on historical figures and events to give an impression of how the philosophy of biological sciences developed throughout. Then follows an examination of the scientific method, the development of which took center stage in the scientific debate during the 20th century, where we review different models of scientific inquiry and introduce the question of how to distinguish science from non-science, known as the demarcation problem. The section concludes with the introduction of the question of what systems biology is assumed to be, according to self-proclaimed founders of the field and their predecessors, and what promises accompanied its advent.
The second section of the introduction concerns itself with cellular metabolism. Here the relevant background on different aspects relating to the specific system studied in this work is provided to the reader. First a general perspective on central carbon metabolism and the different regulatory mechanisms that control it is laid out. Finally, an overview of the study of regulation of central carbon metabolism in the model species Escherichia coli is presented in which we introduce the current state of the field by summarizing contemporary work. This section concludes with an outline of the experimental and computational challenges that scientists in the field are faced with.
In chapter 2 we focus on the prediction of allosteric regulation in Escherichia coli metabolism by single reaction modeling. This model structure identification approach uses the available metabolomics, fluxomics and proteomics steady-state data over different series of nutrient-limited growth conditions, and then asks whether a generalized reversible enzyme kinetics rate law without regulation is sufficient to describe the observations, or whether the inclusion of allosteric regulation significantly improves this ability. In this study we take a reductionistic approach in that we study single reactions in isolation. Herein we sacrifice the ability to probe into how a regulatory interaction might affect emergent behavior such as network dynamics in favor of scalability that the simplistic nature of this approach offers. We systematically assess the regulatory potential of 284 metabolites as allosteric interactors of 84 metabolic reactions. In order to lower the number of false positives among the predicted top ranking interactions we include additional lines of evidence, such as the reported presence of the regulatory interaction in another organism or its detection in physical interaction studies. We then selected the top ranking interactions for follow-up study using enzyme assays in order to test whether the predicted regulators modulate enzyme activity in vitro. In these validation experiments we find evidence for the existence of 11 novel metabolite-protein interactions with potential physiological relevance.
In chapter 3 we concern ourselves with the prediction of allosteric regulation in the tricarboxilic acid cycle of Escherichia coli by using a system of coupled differential equations. This approach combines steady-state data on the metabolome, fluxome and proteome with that of thermodynamic and kinetic parameter estimates to derive model priors over the initial conditions. As in chapter 2, we opt for generalized reversible enzyme kinetics rate laws, but in contrast to it, we study the system as an interconnected network of metabolic reactions. Two additional noteworthy distinctions are the fact that we work with absolute quantification data on all the aforementioned -omics levels, and that we use time series data of the observed metabolome dynamics after a carbon source perturbation, rather than a series of steady states. Using this reaction network we exploit existing dependencies, such as those among thermodynamic parameters, but also those between kinetic parameters, biochemical species and their thermodynamics as described by the Haldane relationship. Our primary goal is to answer the question of whether, based on the available data, we can derive a model that can explain the transient dynamics observed after a carbon source switch, and if not, to systematically identify key regulatory mechanisms that help coordinate this dynamic adaptation. Unexpectedly however, we encountered formidable challenges during this endeavor trying to reproduce the work of predecessors. For that reason the majority of our effort was spent addressing this issue in order to obtain a framework that produces models and predictions that are reproducible, reliable and reusable. Venturing back to the tricarboxylic acid cycle, we find that part of the observed dynamics cannot be sufficiently explained. By assembling an ensemble of models with different regulatory interaction topologies we assess which protein-metabolite interactions help to increase the explanatory potential of the model in describing the observed dynamics most, and derive a set of predictions from these results.
Finally, we summarize the key findings of this thesis. The first achievement is that we predicted and found in vitro evidence for the existence of up to 11 metabolite-protein interactions from our single reaction modeling study using steady-state data and in vitro enzyme assays. However, the absence of baseline expectations poses a problem that highlights inadequacies in scientific rigor. The second achievement is that we have produced a modeling pipeline that generates models and predictions that are reproducible, models that are both reliable, in that the automated construction process minimizes the risk of errors, and reusable, since components are properly annotated, and robust, in that the predictions obtained are less variable than those in previous work. Finally, we show that the transient metabolome dynamics that we observe cannot be sufficiently explained by combining the available data sources, and generate predictions on which allosteric interactions might exist that help shape the response. We conclude with a discussion on the current state of the field, the scientific merit of the endeavor, and bring the whole to fruition with a series of recommendations for future experiments. Mehr anzeigen
Persistenter Link
https://doi.org/10.3929/ethz-b-000668803Publikationsstatus
publishedExterne Links
Printexemplar via ETH-Bibliothek suchen
Beteiligte
Referent: Sauer, Uwe
Referent: Teusink, Bas
Referent: Stelling, Jörg
Referent: Claassen, Manfred
Verlag
ETH ZurichOrganisationseinheit
03713 - Sauer, Uwe / Sauer, Uwe
ETH Bibliographie
yes
Altmetrics