Escherichia coli metabolism under dynamic conditions: The tales of substrate hunting

Dynamic environmental conditions govern microbial metabolism and affect cellular growth. Many applications in biotechnology require cultivating microorganisms in large-scale bioreactors. These environments are commonly characterized by physicochemical gradients, due to imperfect mixing and have been the cause of reduced performance of cell factories in industry. Changes in substrate and gas concentrations, pH and temperature are some example of the generated gradients.  The aim of this thesis is to unravel and understand the effects of repetitive substrate fluctuations on the cellular behaviour of Escherichia coli K12 MG1655, using experimental and modelling approaches. Chapter 1 is a general introduction to biotechnology and its applications, with a focus on upstream bioprocesses. In addition, the role of the bacterium Escherichia coli as a model organism, as well as a working horse of biotechnology, is discussed. In Chapter 2, the quantitative experimental and kinetic modelling approaches, currently used for studying microbial metabolism under dynamic conditions, are summarized and discussed. Current challenges and future perspectives finalize this chapter.  In the experimental Chapter 3, a block-wise feeding regime was applied to an aerobic E.coli culture, with the aim to grow cells under substrate (glucose) gradients, following a reference chemostat (steady-state) growth. This regime was called “fast feast-famine”, as cells experienced periods of substrate excess, limitation and depletion in a time-scale of seconds. The regime was characterized by repetitive cycles of 20 s feeding and 380 s without feeding. The perturbations were applied for at least 8 generations, allowing the cells to adapt to the dynamic environment (highly reproducible cellular response). The specific substrate and oxygen consumption (average) rates increased during the feast-famine regime, compared to the reference steady state cultivation. The increased rates at same (average) growth rate led to a reduced biomass yield (30% lower), while there was no significant by-product formation. Such observation suggests the emergence of energy spilling reactions. With the increase in extracellular substrate concentration, the cells rapidly increased their uptake rate. Within 10 seconds after the beginning of the feeding, the glucose uptake rate was higher (4.68 μmol/gCDW/s) than reported during batch growth (3.3 μmol/gCDW/s). The high uptake led to an accumulation of several intracellular metabolites, during the feast phase, accounting for up to 34% of the carbon supplied. Although the intracellular metabolite concentrations changed rapidly, the cellular energy charge remained homeostatic, suggesting a well-controlled balance between ATP producing and ATP consuming reactions.  The importance of combining experimental perturbation studies and kinetic modelling, in order to reveal metabolic strategies for coping with dynamic conditions is highlighted in the following Chapter 4. In Chapter 4, a published kinetic model for central carbon metabolism by Peskov K, et al. was used to investigate if the experimental observations from Chapter 3 could be reproduced with a model originating from steady-state calibration. Only after parameter optimization, with significant changes, could the data be reproduced, highlighting significant alterations in the enzymatic kinetics of glycolysis during feast-famine, compared to steady-state growth. Post transcriptional modifications were assumed to explain the sudden decrease in the substrate uptake rate, observed while glucose was still in excess. To reflect such change in the modelling approach, the feast-famine cycle was split into two phases and the experimental uptake rate was used as fixed input. Nevertheless, this was not yet sufficient to fully reproduce the experimental observations. The time course of the glycolytic intermediates could only be reproduced when introducing glycogen synthesis and assimilation in the model. Here, glycogen acted as a storage pool, providing carbon and energy to reinitiate growth during famine conditions. Furthermore, ATP spilling reactions were needed to reproduce the observed adenylate energy homeostasis. Additionally, a continuous draining of ATP supported the hypothesis of increased maintenance during the feast-famine regime. In Chapter 5, multi-omics approaches, i.e. shotgun cellular proteomics and 13C-labelled metabolomics were used for untargeted analysis and generation of new hypotheses on cellular regulatory mechanisms, when cells were subjected to fluctuations in substrate availability. The extracellular dynamics were expected to trigger global stress responses, in line with the observed reduced biomass yield. Surprisingly, this was not the case – stress related proteins did not alter from steady-state to feast-famine conditions. On the other hand, the cellular proteome adjusted for specific functional categories, including biosynthesis and translation processes (ribosomes). This increase can be explained by either increased protein production to support the rapid growth changes, during the short time of substrate availability, or ribosome stalling due to amino acid limitation during the famine phase. During substrate-limited growth (constant feeding) cells have an overcapacity of metabolic enzymes (involved in central carbon pathways), which is used under nutrient up-shift to handle rapid increase in metabolic fluxes. The down-regulation of several enzymes in glycolysis, TCA cycle and pentose phosphate pathway, as well as, transporter proteins, revealed that cells respond more to the substrate excess period than the starvation period during the block-wise feeding regime. This is also in accordance with the observed down-regulation of the glyoxylate-shunt enzymes. Moreover, the increased levels of polyphosphate kinase indicated the use of a polyphosphate pool as a putative buffer for energy homeostasis. Glycogen production and degradation was verified by the proteomic and 13C tracing analysis and is suggested to contribute to the ATP spilling (biomass yield losses), along with the increased protein turnover, which was identified by an increased section of the cellular proteasome. The generated insights of the whole thesis are summarized in Chapter 6. Additionally, open questions are discussed. The future challenges include scale-down experiments, research on the effects of dynamics on production hosts, the use of mutant strains for validation experiments and data integration toward multi-scale modeling.