Elsevier

Journal of Biotechnology

Volume 184, 20 August 2014, Pages 172-178
Journal of Biotechnology

Reducing Recon 2 for steady-state flux analysis of HEK cell culture

https://doi.org/10.1016/j.jbiotec.2014.05.021Get rights and content

Highlights

  • Reduced Recon 2 into a concise metabolic model for flux analysis in four overall steps.

  • Original model has 7440 reactions; reduced model has 357 reactions.

  • Curated Recon 2's metabolites and reactions for inconsistencies and duplication.

  • Glutaminolysis rate increased when l-alanyl-l-glutamine dipeptide was depleted.

Abstract

A representative stoichiometric model is essential to perform metabolic flux analysis (MFA) using experimentally measured consumption (or production) rates as constraints. For Human Embryonic Kidney (HEK) cell culture, there is the opportunity to use an extremely well-curated and annotated human genome-scale model Recon 2 for MFA. Performing MFA using Recon 2 without any modification would have implied that cells have access to all functionality encoded by the genome, which is not realistic. The majority of intracellular fluxes are poorly determined as only extracellular exchange rates are measured. This is compounded by the fact that there is no suitable metabolic objective function to suppress non-specific fluxes. We devised a heuristic to systematically reduce Recon 2 to emphasize flux through core metabolic reactions. This implies that cells would engage these dominant metabolic pathways to grow, and any significant changes in gross metabolic phenotypes would have invoked changes in these pathways. The reduced metabolic model becomes a functionalized version of Recon 2 used for identifying significant metabolic changes in cells by flux analysis.

Introduction

Metabolism is a set of cellular processes that can be quantified using a constraint-based modeling approach (Lewis et al., 2012). The main constraints used in Metabolic Flux Analysis (MFA) are the metabolic model, which describes the cellular processes available to the cell, and the time profiles of accumulation (or depletion) of substrates and products. The activity of cellular processes can be estimated using constraints derived from observed gross changes in the medium (Mahadevan and Schilling, 2003, Schellenberger et al., 2011b). The activity of the TCA cycle, for example, depends on the catabolism of glucose and amino acids, the production of lactate, the biosynthetic demand for precursors and the consumption of oxygen. More importantly, we can identify plausible shifts in metabolism based on changes in the measured rates. Flux analysis is performed on mammalian cells to understand how have cells metabolized the variety of resources available in the medium, and ultimately to improve cell culture performance and productivity (Ahn and Antoniewicz, 2012, Altamirano et al., 2013, Bonarius et al., 1996, Dietmair et al., 2012, Follstad et al., 1999, Xie and Wang, 1994).

Compared to microbial fermentations, a greater customization of the metabolic model is required when dealing with mammalian cell cultures. Mammalian cells in culture display a fastidious metabolism, typically characterized by high rates of glycolysis and glutaminolysis (Warburg, 1956, Zeng and Deckwer, 1995). Mammalian cells have specific auxotrophies, at the same time are able to consume a wide variety of substrates if available. Moreover, the consumptions of these substrates may be in excess of the amount required for biosynthesis. The major by-products of mammalian cell cultures are lactate and ammonium, but may include glutamate, alanine, aspartate, proline and glycine. Depending on the medium composition and the growth phase, the same by-products can be reversibly consumed (Martínez et al., 2013). Other aspects which must be considered are the decomposition of free glutamine (Ozturk and Palsson, 1990), the presence of viable and dead cells, and cells showing non-steady or arrested growth (Ahn and Antoniewicz, 2012, Martínez et al., 2013, Sengupta et al., 2011).

Recon 2 is a comprehensive and high quality knowledge database of human metabolism (Duarte et al., 2007, Thiele et al., 2013). It is a refined and consolidated compendium of gene, protein, reaction and metabolite interactions from existing reconstructions, published literature and personal expertise. While Recon 2 is simulation-ready, flux results obtained tend to be loose and non-specific. The high degree of redundancy in the model masks metabolic contribution of individual reactions. For example, it is not possible to determine the contribution of the oxidative pentose-phosphate pathway in the production of NADPH, as other, less significant pathways can accomplish the same.

Here, we will use an improved and reduced version of Recon 2 to investigate the metabolism of human embryonic kidney (HEK) cells, a common host for the production of recombinant proteins, vaccines and viruses (Martínez et al., 2010). The aim of the model reduction process is to remove redundant pathways in the peripheral metabolism involved in biosynthesis and catabolism, but at the same time place emphasis on metabolic flexibilities in the central metabolism in the attempt to resolve the latter based on experimental data. Peripheral pathways are often poorly determined, and this simplification implies that peripheral pathways are linear if a flux constraint can be assigned, or trivial (non-existent) if no flux constraint can be provided. For example, it is often assumed that nucleotide salvage and fatty acid beta-oxidation pathways are inactive in growing cells.

This work describes the “functionalization” of Recon 2 by model reduction for steady-state MFA of a HEK cell culture. The model reduction process is carefully conceived, and it becomes a ledger to formalize the assumptions incorporated into MFA that can be revised if necessary. While MFA is fundamentally limited in terms of resolution, it can be effectively used to identify plausible shifts in metabolism suggested by experimental data that warrant subsequent investigations. This work bridges the gap between a highly curated but generic Recon 2 and its specialized application in investigating the metabolism of human cell lines in culture.

Section snippets

Model curation

Recon 2 (MODEL1109130000) obtained from BioModels Database (Li et al., 2010) was further curated for inconsistencies and duplication, ensuring that the overall network is balanced in terms of mass and energy. Reactions containing unbalanced protons were replaced by reactions obtained from Recon X website (www.humanmetabolism.org) (see Supplementary 2.1). Inconsistent names, charges, formula and annotation information (KEGG, PubChem, ChEBI) were identified and resolved (see Supplementary 2.2).

Recon 2 reduction

Table 1 shows the number of active reactions contained in Recon 2, and the decline in model size as Recon 2 was being modified and reduced. The degrees-of-freedom of the models are shown as an indicator of the number of underdetermined metabolic pathways. The constrained reactions are listed in Supplementary 3.2.

Discussion

Instead of taking MFA results as absolute, MFA was used to identify metabolic changes in HEK cells when l-alanyl-l-glutamine dipeptide was depleted. We pursued an aggressive approach of reducing Recon 2 for the purpose of performing comparative flux analysis emphasizing on the central metabolism. The reduction process was operated under the theme that peripheral pathways are linear if a flux constraint can be assigned, or trivial (non-existent) if no flux constraint can be provided. Reactions

Conclusion

Recon 2 is a large-scale but highly accurate reaction database of human metabolism, which was functionalized by reduction into a concise model for MFA. The proposed reduction process is a necessary step in the attempt to perform MFA using genome-scale models that are highly generic and underdetermined, nonetheless subsequent analyses are still bound by known limitations of conventional MFA.

References (41)

  • J. Schellenberger et al.

    Elimination of thermodynamically infeasible loops in steady-state metabolic models

    Biophys. J.

    (2011)
  • W.S. Ahn et al.

    Towards dynamic metabolic flux analysis in CHO cell cultures

    Biotechnol. J.

    (2012)
  • C. Altamirano et al.

    Advances in improving mammalian cells metabolism for recombinant protein production

    Electron. J. Biotechnol.

    (2013)
  • S.A. Becker et al.

    Context-specific metabolic networks are consistent with experiments

    PLoS Comput. Biol.

    (2008)
  • H.P.J. Bonarius et al.

    Determination of the respiration quotient in mammalian cell culture in bicarbonate buffered media

    Biotechnol. Bioeng.

    (1995)
  • H.P.J. Bonarius et al.

    Metabolic flux analysis of hybridoma cells in different culture media using mass balances

    Biotechnol. Bioeng.

    (1996)
  • H.P.J. Bonarius et al.

    Activity of glutamate dehydrogenase is increased in ammonia-stressed hybridoma cells

    Biotechnol. Bioeng.

    (1998)
  • A.P. Burgard et al.

    Minimal reaction sets for Escherichia coli metabolism under different growth requirements and uptake environments

    Biotechnol. Prog.

    (2001)
  • S. Dietmair et al.

    A multi-omics analysis of recombinant protein production in Hek293 cells

    PLoS ONE

    (2012)
  • N.C. Duarte et al.

    Global reconstruction of the human metabolic network based on genomic and bibliomic data

    Proc. Natl. Acad. Sci. U. S. A.

    (2007)
  • Cited by (48)

    • Process modeling of recombinant adeno-associated virus production in HEK293 cells

      2022, Current Opinion in Chemical Engineering
      Citation Excerpt :

      Several constraint-based models of HEK293 metabolism have been developed based on a reconstruction of a human genome-scale model. Quek et al. (2014) first described the protocol for reducing Recon 2.0 (the second version of the reconstruction of human metabolism with 7440 reactions and 2626 metabolites [27]) into a steady-state MFA for HEK293 cells (357 reactions) [28]. The model was used to analyze the depletion of L-alanyl-L-glutamine dipeptide, which caused HEK cells to produce alanine and consume glutamine.

    • Microbial metabolomics: recent advancements and applications in infectious diseases and drug discovery

      2022, Recent Advances and Future Perspectives of Microbial Metabolites: Applications in Biomedicine
    • Modeling and optimization of bioreactor processes

      2022, Current Developments in Biotechnology and Bioengineering: Advances in Bioprocess Engineering
    • Cell culture metabolomics and lipidomics

      2022, Metabolomics Perspectives: From Theory to Practical Application
    View all citing articles on Scopus
    View full text