Metabolic flux analysis of recombinant protein overproduction in Escherichia coli

Genome-scale models of metabolism and expression predict the metabolic burden of recombinant protein expression

The production of recombinant proteins in a host using synthetic constructs such as plasmids comes at the cost of detrimental effects such as reduced growth, energetic inefficiencies, and other stress responses, collectively known as metabolic burden. Increasing the number of copies of the foreign gene increases the metabolic load but increases the expression of the foreign protein. Thus, there is a trade-off between biomass and product yield in response to changes in heterologous gene copy number. This work proposes a computational method, rETFL (recombinant Expression and Thermodynamic Flux), for analyzing and predicting the responses of recombinant organisms to the introduction of synthetic constructs. rETFL is an extension to the ETFL formulations designed to reconstruct models of metabolism and expression (ME-models). We have illustrated the capabilities of the method in four studies to (i) capture the growth reduction in plasmid-containing E. coli and recombinant protein production; (ii) explore the trade-off between biomass and product yield as plasmid copy number is varied; (iii) predict the emergence of overflow metabolism in recombinant E. coli in agreement with experimental data; and (iv) investigate the individual pathways and enzymes affected by the presence of the plasmid. We anticipate that rETFL will serve as a comprehensive platform for integrating available omics data for recombinant organisms and making context-specific predictions that can help optimize recombinant expression systems for biopharmaceutical production and gene therapy.

Dynamic flux balance analysis of high cell density fed-batch culture of Escherichia coli BL21 (DE3) with mass spectrometry-based spent media analysis

Dynamic flux balance analysis (FBA) allows estimation of intracellular reaction rates using organism-specific genome-scale metabolic models (GSMM) and by assuming instantaneous pseudo-steady states for processes that are inherently dynamic. This technique is well-suited for industrial bioprocesses employing complex media characterized by a hierarchy of substrate uptake and product secretion. However, knowledge of exchange rates of many components of the media would be required to obtain meaningful results. Here, we performed spent media analysis using mass spectrometry coupled with liquid and gas chromatography for a fed-batch, high-cell density cultivation of Escherichia coli BL21(DE3) expressing a recombinant protein. Time course measurements thus obtained for 246 metabolites were converted to instantaneous exchange rates. These were then used as constraints for dynamic FBA using a previously reported GSMM, thus providing insights into how the flux map evolves through the process. Changes in tri-carboxylic acid cycle fluxes correlated with the increased demand for energy during recombinant protein production. The results show how amino acids act as hubs for the synthesis of other cellular metabolites. Our results provide a deeper understanding of an industrial bioprocess and will have implications in further optimizing the process.

Silk fibroin production in Escherichia coli is limited by a positive feedback loop between metabolic burden and toxicity stress

To investigate the metabolic elasticity and production bottlenecks for recombinant silk proteins in Escherichia coli, we performed a comprehensive characterization of one elastin-like peptide strain (ELP) and two silk protein strains (A5 4mer, A5 16mer). Our approach included ¹³C metabolic flux analysis, genome-scale modeling, transcription analysis, and ¹³C-assisted media optimization experiments. Three engineered strains maintained their central flux network during growth, while measurable metabolic flux redistributions (such as the Entner-Doudoroff pathway) were detected. Under metabolic burdens, the reduced TCA fluxes forced the engineered strain to rely more on substrate-level phosphorylation for ATP production, which increased acetate overflow. Acetate (as low as 10 mM) in the media was highly toxic to silk-producing strains, which reduced 4mer production by 43% and 16mer by 84%, respectively. Due to the high toxicity of large-size silk proteins, 16mer's productivity was limited, particularly in the minimal medium. Therefore, metabolic burden, overflow acetate, and toxicity of silk proteins may form a vicious positive feedback loop that fractures the metabolic network. Three solutions could be applied: 1) addition of building block supplements (i.e., eight key amino acids: His, Ile, Phe, Pro, Tyr, Lys, Met, Glu) to reduce metabolic burden; 2) disengagement of growth and production; and 3) use of non-glucose based substrate to reduce acetate overflow. Other reported strategies were also discussed in light of decoupling this positive feedback loop.

Central carbon metabolite profiling reveals vector-associated differences in the recombinant protein production host Escherichia coli BL21

The Gram-negative bacterium Escherichia coli is the most widely used host for recombinant protein production, both as an industrial expression platform and as a model system at laboratory scale. The recombinant protein production industry generates proteins with direct applications as biopharmaceuticals and in technological processes central to a plethora of fields. Despite the increasing economic significance of recombinant protein production, and the importance of E. coli as an expression platform and model organism, only few studies have focused on the central carbon metabolic landscape of E. coli during high-level recombinant protein production. In the present work, we applied four targeted CapIC- and LC-MS/MS methods, covering over 60 metabolites, to perform an in-depth metabolite profiling of the effects of high-level recombinant protein production in strains derived from E. coli BL21, carrying XylS/Pm vectors with different characteristics. The mass-spectrometric central carbon metabolite profiling was complemented with the study of growth kinetics and protein production in batch bioreactors. Our work shows the robustness in E. coli central carbon metabolism when introducing increased plasmid copy number, as well as the greater importance of induction of recombinant protein production as a metabolic challenge, especially when strong promoters are used.

Dynamic metabolic modelling of overproduced protein secretion in Streptomyces lividans using adaptive DFBA

BACKGROUND

Streptomyces lividans is an appealing host for the production of proteins of biotechnological interest due to its relaxed exogenous DNA restriction system and its ability to secrete proteins directly to the medium through the major Sec or the minor Tat routes. Often, protein secretion displays non-uniform time-dependent patterns. Understanding the associated metabolic changes is a crucial step to engineer protein production. Dynamic Flux Balance Analysis (DFBA) allows the study of the interactions between a modelled organism and its environment over time. Existing methods allow the specification of initial model and environment conditions, but do not allow introducing arbitrary modifications in the course of the simulation. Living organisms, however, display unexpected adaptive metabolic behaviours in response to unpredictable changes in their environment. Engineering the secretion of products of biotechnological interest has systematically proven especially difficult to model using DFBA. Accurate time-dependent modelling of complex and/or arbitrary, adaptive metabolic processes demands an extended approach to DFBA.

RESULTS

In this work, we introduce Adaptive DFBA, a novel, versatile simulation approach that permits inclusion of changes in the organism or the environment at any time in the simulation, either arbitrary or interactively responsive to environmental changes. This approach extends traditional DFBA to allow steering arbitrarily complex simulations of metabolic dynamics. When applied to Sec- or Tat-dependent secretion of overproduced proteins in S. lividans, Adaptive DFBA can overcome the limitations of traditional DFBA to reproduce experimental data on plasmid-free, plasmid bearing and secretory protein overproducing S. lividans TK24, and can yield useful insights on the behaviour of systems with limited experimental knowledge such as agarase or amylase overproduction in S. lividans TK21.

CONCLUSIONS

Adaptive DFBA has allowed us to overcome DFBA limitations and to generate more accurate models of the metabolism during the overproduction of secretory proteins in S. lividans, improving our understanding of the underlying processes. Adaptive DFBA is versatile enough to permit dynamical metabolic simulations of arbitrarily complex biotechnological processes.

Quantification of proteomic and metabolic burdens predicts growth retardation and overflow metabolism in recombinant Escherichia coli

Escherichia coli has been the host organism most frequently investigated for efficient recombinant protein production. However, the production of a foreign protein in recombinant E. coli often leads to growth deterioration and elevated secretion of acetic acid. Such observed phenomena have been widely linked with cell stress responses and metabolic burdens originated particularly from the increased energy demand. In this study, flux balance analysis and dynamic flux balance analysis were applied to investigate the observed growth physiology of recombinant E. coli, incorporating the proteome allocation theory and an adjustable maintenance energy level (ATPM) to capture the proteomic and energetic burdens introduced by recombinant protein synthesis. Model predictions of biomass growth, substrate consumption, acetate excretion, and protein production with two different strains were in good agreement with the experimental data, indicating that the constraint on the available proteomic resource and the change in ATPM might be important contributors governing the growth physiology of recombinant strains. The modeling framework developed in this work, currently with several limitations to overcome, offers a starting point for the development of a practical, model-based tool to guide metabolic engineering decisions for boosting recombinant protein production.

Modelling the metabolism of protein secretion through the Tat route in Streptomyces lividans

Background

Streptomyces lividans has demonstrated its value as an efficient host for protein production due to its ability to secrete functional proteins directly to the media. Secretory proteins that use the major Sec route need to be properly folded outside the cell, whereas secretory proteins using the Tat route appear outside the cell correctly folded. This feature makes the Tat system very attractive for the production of natural or engineered Tat secretory proteins. S. lividans cells are known to respond differently to overproduction and secretion of Tat versus Sec proteins. Increased understanding of the impact of protein secretion through the Tat route can be obtained by a deeper analysis of the metabolic impact associated with protein production, and its dependence on protein origin, composition, secretion mechanisms, growth phases and nutrients. Flux Balance Analysis of Genome-Scale Metabolic Network models provides a theoretical framework to investigate cell metabolism under different constraints.

Results

We have built new models for various S. lividans strains to better understand the mechanisms associated with overproduction of proteins secreted through the Tat route. We compare models of an S. lividans Tat-dependent agarase overproducing strain with those of the S. lividans wild-type, an S. lividans strain carrying the multi-copy plasmid vector and an α-amylase Sec-dependent overproducing strain. Using updated genomic, transcriptomic and experimental data we could extend existing S. lividans models and produce a new model which produces improved results largely extending the coverage of S. lividans strains, the number of genes and reactions being considered, the predictive behaviour and the dependence on specification of exchange constraints. Comparison of the optimized solutions obtained highlights numerous changes between Tat- and Sec-dependent protein secreting strains affecting the metabolism of carbon, amino acids, nucleotides, lipids and cofactors, and variability analysis predicts a large potential for protein overproduction.

Conclusions

This work provides a detailed look to metabolic changes associated to Tat-dependent protein secretion reproducing experimental observations and identifying changes that are specific to each secretory route, presenting a novel, improved, more accurate and strain-independent model of S. lividans, thus opening the way for enhanced metabolic engineering of protein overproduction in S. lividans.

Electronic supplementary material

The online version of this article (10.1186/s12866-018-1199-3) contains supplementary material, which is available to authorized users.

A novel cytosolic NADH:quinone oxidoreductase from Methanothermobacter marburgensis

Methanothermobacter marburgensis is a strictly anaerobic, thermophilic methanogenic archaeon that uses methanogenesis to convert H₂ and CO₂ to energy. M. marburgensis is one of the best-studied methanogens, and all genes required for methanogenic metabolism have been identified. Nonetheless, the present study describes a gene (Gene ID 9704440) coding for a putative NAD(P)H:quinone oxidoreductase that has not yet been identified as part of the metabolic machinery. The gene product, MmNQO, was successfully expressed, purified and characterized biochemically, as well as structurally. MmNQO was identified as a flavin-dependent NADH:quinone oxidoreductase with the capacity to oxidize NADH in the presence of a wide range of electron acceptors, whereas NADPH was oxidized with only three acceptors. The 1.50 Å crystal structure of MmNQO features a homodimeric enzyme where each monomer comprises 196 residues folding into flavodoxin-like α/β domains with non-covalently bound FMN (flavin mononucleotide). The closest structural homologue is the modulator of drug activity B from Streptococcus mutans with 1.6 Å root-mean-square deviation on 161 Cα atoms and 28% amino-acid sequence identity. The low similarity at sequence and structural level suggests that MmNQO is unique among NADH:quinone oxidoreductases characterized to date. Based on preliminary bioreactor experiments, MmNQO could provide a useful tool to prevent overflow metabolism in applications that require cells with high energy demand.

A novel NADH:quinone oxidoreductase, MmNQO, from Methanothermobacter marburgensis was identified. MmNQO oxidizes NADH with several electron acceptors and is structurally similar to bacterial MdaB. It is localized in the cytosol and may provide a useful tool to prevent overflow metabolism.

Metabolic impact assessment for heterologous protein production in Streptomyces lividans based on genome-scale metabolic network modeling

The metabolic impact exerted on a microorganism due to heterologous protein production is still poorly understood in Streptomyces lividans. In this present paper, based on exometabolomic data, a proposed genome-scale metabolic network model is used to assess this metabolic impact in S. lividans. Constraint-based modeling results obtained in this work revealed that the metabolic impact due to heterologous protein production is widely distributed in the genome of S. lividans, causing both slow substrate assimilation and a shift in active pathways. Exchange fluxes that are critical for model performance have been identified for metabolites of mouse tumor necrosis factor, histidine, valine and lysine, as well as biomass. Our results unravel the interaction of heterologous protein production with intracellular metabolism of S. lividans, thus, a possible basis for further studies in relieving the metabolic burden via metabolic or bioprocess engineering.

Quantification of metabolic limitations during recombinant protein production in Escherichia coli

Escherichia coli is one of the major microorganisms for recombinant protein production because it has been best characterized in terms of molecular genetics and physiology, and because of the availability of various expression vectors and strains. The synthesis of proteins is one of the most energy consuming processes in the cell, with the result that cellular energy supply may become critical. Indeed, the so called metabolic burden of recombinant protein synthesis was reported to cause alterations in the operation of the host's central carbon metabolism. To quantify these alterations in E. coli metabolism in dependence of the rate of recombinant protein production, (13)C-tracer-based metabolic flux analysis in differently induced cultures was used. To avoid dilution of the (13)C-tracer signal by the culture history, the recombinant protein produced was used as a flux probe, i.e., as a read out of intracellular flux distributions. In detail, an increase in the generation rate rising from 36 mmol(ATP)g(CDW)(-1)h(-1) for the reference strain to 45 mmol(ATP)g(CDW)(-1)h(-1) for the highest yielding strain was observed during batch cultivation. Notably, the flux through the TCA cycle was rather constant at 2.5±0.1 mmol g(CDW)(-1)h(-1), hence was independent of the induced strength for gene expression. E. coli compensated for the additional energy demand of recombinant protein synthesis by reducing the biomass formation to almost 60%, resulting in excess NADPH. Speculative, this excess NADPH was converted to NADH via the soluble transhydrogenase and subsequently used for ATP generation in the electron transport chain. In this study, the metabolic burden was quantified by the biomass yield on ATP, which constantly decreased from 11.7g(CDW)mmol(ATP)(-1) for the reference strain to 4.9g(CDW)mmol(ATP)(-1) for the highest yielding strain. The insights into the operation of the metabolism of E. coli during recombinant protein production might guide the optimization of microbial hosts and fermentation conditions.

Improving N-glycosylation efficiency in Escherichia coli using shotgun proteomics, metabolic network analysis, and selective reaction monitoring

Recently, the prospect of using Escherichia coli as a host for human glycoprotein production has increased due to detailed characterization of the prokaryotic N-glycosylation process and the ability to transfer the system into this bacterium. Although functionality of the native Campylobacter jejuni N-glycosylation system in E. coli has been demonstrated, the efficiency of the process using the well-characterized C. jejuni glycoprotein AcrA, was found to be low at 13.4±0.9% of total extracted protein. A combined approach using isobaric labeling of peptides and probability-based network analysis of metabolic changes was applied to forward engineer E. coli to improve glycosylation efficiency of AcrA. Enhancing flux through the glyoxylate cycle was identified as a potential metabolic manipulation to improve modification efficiency and was achieved by increasing the expression of isocitrate lyase. While the overall recombinant protein titre did not change significantly, the amount of glycosylated protein increased by approximately 300%.

Amino acid uptake profiling of wild type and recombinant Streptomyces lividans TK24 batch fermentations

Streptomyces lividans is considered an interesting host for the secretory production of heterologous proteins. To obtain a good secretion yield of heterologous proteins, the availability of suitable nitrogen sources in the medium is required. Often, undefined mixtures of amino acids are used to improve protein yields. However, the understanding of amino acid utilization as well as their contribution to the heterologous protein synthesis is poor. In this paper, amino acid utilization by wild type and recombinant S. lividans TK24 growing on a minimal medium supplemented with casamino acids is profiled by intensive analysis of the exometabolome (metabolic footprint) as a function of time. Dynamics of biomass, substrates, by-products and heterologous protein are characterized, analyzed and compared. As an exemplary protein mouse Tumor Necrosis Factor Alpha (mTNF-α) is considered. Results unveil preferential glutamate and aspartate assimilation, together with glucose and ammonium, but the associated high biomass growth rate is unfavorable for protein production. Excretion of organic acids as well as alanine is observed. Pyruvate and alanine overflow point at an imbalance between carbon and nitrogen catabolism and biosynthetic fluxes. Lactate secretion is probably related to clump formation. Heterologous protein production induces a slowdown in growth, denser clump formation and a shift in metabolism, as reflected in the altered substrate requirements and overflow pattern. Besides glutamate and aspartate, most amino acids are catabolized, however, their exact contribution in heterologous protein production could not be seized from macroscopic quantities. The metabolic footprints presented in this paper provide a first insight into the impact and relevance of amino acids on biomass growth and protein production. Type and availability of substrates together with biomass growth rate and morphology affect the protein secretion efficiency and should be optimally controlled, e.g., by appropriate medium formulation and substrate dosing. Overflow metabolism as well as high biomass growth rates must be avoided because they reduce protein yields. Further investigation of the intracellular metabolic fluxes should be conducted to fully unravel and identify ways to relieve the metabolic burden of plasmid maintenance and heterologous protein production and to prevent overflow.

Identification of cellular objective for elucidating the physiological state of plasmid-bearing Escherichia coli using genome-scale in silico analysis

The presence of multiple copies of plasmids in Escherichia coli could induce a complex cascade of physiological changes known as the metabolic burden response. In this work, the physiological effect of such plasmid metabolic burden on E. coli metabolism was investigated by constraint-based genome-scale flux modeling. We systematically applied three cellular objectives: (a) maximizing growth rate, (b) maximizing plasmid production, and (c) maximizing maintenance energy expenditure to quantify in silico flux distributions. These simulated results were compared with experimental flux information to identify which of these cellular objectives best describes the physiological and metabolic states of plasmid-bearing (P+) E. coli. Unlike the wild-type E. coli cells that have directed the metabolism toward an optimum growth rate under the nutrient-limited condition, the maximum growth rate objective could not correctly predict the metabolic state of recombinant P+ cells. Instead, flux simulations by maximizing maintenance energy expenditure showed good consistency with experimental observation, indicating that the P+ cells are energetically less efficient and could require higher maintenance energy. This study demonstrates that the cellular objective of maximizing maintenance energy expenditure provides a better description of the underlying physiological state in recombinant microorganisms relevant to biotechnological applications.

Metabolic engineering of microorganisms for biofuels production: from bugs to synthetic biology to fuels

The ability to generate microorganisms that can produce biofuels similar to petroleum-based transportation fuels would allow the use of existing engines and infrastructure and would save an enormous amount of capital required for replacing the current infrastructure to accommodate biofuels that have properties significantly different from petroleum-based fuels. Several groups have demonstrated the feasibility of manipulating microbes to produce molecules similar to petroleum-derived products, albeit at relatively low productivity (e.g. maximum butanol production is around 20 g/L). For cost-effective production of biofuels, the fuel-producing hosts and pathways must be engineered and optimized. Advances in metabolic engineering and synthetic biology will provide new tools for metabolic engineers to better understand how to rewire the cell in order to create the desired phenotypes for the production of economically viable biofuels.

Metabolic flux and robustness analysis of glycerol metabolism in Klebsiella pneumoniae

The knowledge of the mechanism of flux distribution will benefit understanding cell physiology and regulation of metabolism. In this study, the measured fluxes obtained under steady-state conditions were used to estimate intracellular fluxes and identify the robustness of branch points of the anaerobic glycerol metabolism in Klebsiella pneumoniae for the production of 1,3-propanediol by metabolic flux analysis. The biomass concentration increased as NADH(2)/NAD(+) decreased at low initial concentration and inversed at high initial glycerol concentration. The flux distribution revealed that the branch points of glycerol and dihydroxyacetonephosphate were rigid to the environmental conditions. However, the pyruvate and acetyl coenzyme A metabolisms gave cells the flexibility to regulate the energy and intermediate fluxes under various environmental conditions. Additionly, it was found that the formation rate of ethanol and the ratio of pyruvate dehydrogenase to pyruvate formate lyase appeared visible fluctuations at high glycerol uptake rate.

Enabling technologies: fermentation and downstream processing

Efficient parallel tools for bioprocess design, consequent application of the concepts for metabolic process analysis as well as innovative downstream processing techniques are enabling technologies for new industrial bioprocesses from an engineering point of view. Basic principles, state-of-the-art techniques and cutting-edge technologies are briefly reviewed. Emphasis is on parallel bioreactors for bioprocess design, biochemical systems characterization and metabolic control analysis, as well as on preparative chromatography, affinity filtration and protein crystallization on a process scale.