Enzyme capacity-based genome scale modelling of CHO cells

Hock Chuan Yeo, Jongkwang Hong, Meiyappan Lakshmanan, Dong‐Yup Lee

TL;DR

This paper introduces a new genome-scale metabolic model for CHO cells (iCHO2291) that incorporates enzyme kinetic information, leading to more accurate predictions of cellular metabolism. This enhanced model, termed ecFBA, can better simulate complex cellular behaviors like overflow metabolism and differentiate between cell line-specific metabolic traits, paving the way for improved biomanufacturing processes.

What Problem Does This Paper Address?

Chinese Hamster Ovary (CHO) cells are workhorses in biomanufacturing, producing a vast majority of therapeutic proteins. However, optimizing their growth and productivity is challenging due to the complexity of their metabolic networks and the variability introduced by different cell lines and culture conditions. Traditional genome-scale metabolic models (GEMs) often struggle to accurately predict intracellular fluxes and identify key engineering targets because they lack detailed kinetic information and enzyme capacity constraints. This leads to a large degree of freedom in model predictions, making it difficult to pinpoint critical bottlenecks. The authors aim to address this by developing a more sophisticated GEM that accounts for enzyme kinetics and capacity, thereby improving the predictive power of metabolic models for CHO cells.

Methods

The core of this work is the development of an enzyme capacity-constrained flux balance analysis (ecFBA) framework. This approach builds upon standard Flux Balance Analysis (FBA) by incorporating enzyme kinetic parameters, specifically turnover numbers ($k_{cat}$) and molecular weights (MW), as constraints.

Here's a breakdown of the key methodological aspects:

  • Model Update (iCHO2291): The authors significantly expanded and refined a previous CHO GEM (iCHO1766). This involved correcting biochemical inconsistencies, adding new genes and reactions based on updated genome annotations, and improving gene-protein-reaction (GPR) associations.
  • Kinetome Incorporation: A crucial novelty is the integration of enzyme kinetic parameters. This includes:
  • Turnover Numbers ($k_{cat}$): Obtained from databases like BRENDA, with a hierarchical approach to select data for Chinese hamster, rodents, and other species if hamster-specific data was unavailable.
  • Molecular Weights (MW): Estimated based on gene-protein-relationships (GPRs) and translated genomic coding sequences (CDS).
  • Enzyme Capacity Constraint: The ecFBA formulation mathematically limits the total flux through reactions based on the capacity of the enzymes involved, using their $k_{cat}$ and MW. This is represented by the equation: $\sum_{j=1}^{N} \frac{MW_j}{k_{cat,j}} v_j \le C$, where $C$ is a constant representing total enzymatic mass.
  • Comparison with Existing Methods: The performance of ecFBA was benchmarked against established methods like parsimonious FBA (pFBA) and carbon-constraint FBA (ccFBA) using experimental datasets.

Datasets

The study utilized several datasets for model validation and comparison:

  • Experimental Flux Data: Intracellular flux measurements from six different cell culture experiments were used. These datasets were chosen to assess the impact of:
  • Cell Line Engineering: Data from Templeton et al. (2014) involving two CHO clones with varying levels of anti-apoptotic gene Bcl-2 expression.
  • Media Changes: Data from McAtee Pereira et al. (2018) comparing the same CHO clone in three different chemically defined media.
  • Kinetome Data: Turnover numbers ($k_{cat}$) and molecular weights (MW) for enzymes were compiled from various sources, including the BRENDA database and human/rodent orthologs.

Key Insights and Ideas

  • Enzyme Capacity is Key to Overflow Metabolism: The study strongly suggests that overflow metabolism (e.g., excessive lactate production) in CHO cells under high glucose conditions is not solely due to limitations in oxidative phosphorylation but is significantly influenced by the capacity of the enzymes involved. When enzyme capacity is limited, cells resort to less efficient but less resource-intensive pathways.
  • Kinetic Information Improves Flux Prediction Accuracy: Incorporating enzyme kinetic parameters (turnover numbers and molecular weights) into FBA significantly reduces flux variability and leads to predictions that are much more consistent with experimental measurements compared to methods that only consider stoichiometry.
  • Differentiating Cell Line and Media Effects: The ecFBA model was able to capture subtle metabolic differences between CHO clones and the impact of different media compositions on lactate metabolism, highlighting its utility for understanding clone-specific behaviors.
  • Lactate as an Energy Source: The model suggests that lactate can be re-oxidized in mitochondria and potentially enter the TCA cycle as an energy source, a finding consistent with recent experimental evidence.
  • Lactate-Pyruvate Cycling: The simulations also pointed to a potential lactate-pyruvate cycling pathway between the cytosol and mitochondria, which could play a role in balancing cellular redox states and supporting glycolysis.

Key Results

  • Improved Correlation with Experimental Data: ecFBA achieved significantly higher Pearson correlation coefficients (r) with experimentally measured intracellular fluxes across all six tested datasets compared to pFBA and ccFBA. For example, in the control condition dataset, ecFBA achieved an r of 0.941, while ccFBA and pFBA achieved 0.794 and 0.901, respectively.
  • Accurate Prediction of Overflow Metabolism: ecFBA successfully predicted the onset of lactate secretion under glucose excess conditions, a phenomenon not captured by standard FBA or pFBA without explicit oxygen limitations.
  • Reduced Blocked Reactions and Dead-End Metabolites: The updated iCHO2291 model showed a substantial reduction in blocked reactions (from 24% to 13%) and dead-end metabolites (from 23% to 9%) compared to its predecessor (iCHO1766).
  • Clone-Specific Metabolic Differences: The model accurately simulated the distinct lactate metabolism of two CHO clones, with clone A showing higher mitochondrial lactate transport and re-oxidation compared to clone B.

Limitations and Future Work

  • Limited Kinetic Data Availability: A significant challenge acknowledged by the authors is the limited availability of reliable in vitro kinetic data ($k_{cat}$ values) for many enzymes in the CHO GEM. While they addressed this by using data from orthologs and employing random permutations, the accuracy of predictions is still dependent on the quality of this data.
  • Experimental Verification: While the model provides strong predictions, the authors suggest that experimental verification, particularly using C13-isotope labeling, is needed to confirm the existence and significance of certain metabolic pathways like lactate cycling.
  • Digital Twin Development: The authors envision that iCHO2291 with ecFBA can serve as a foundational platform for developing "Digital Twin" models for bioreactors. This would involve integrating real-time culture data with the model to predict metabolic changes, cell growth, and product synthesis, ultimately accelerating bioprocess development and optimization.
  • Machine Learning for Kinetic Parameter Prediction: To overcome the data scarcity for kinetic parameters, the authors suggest that machine learning approaches could be used to predict turnover numbers based on enzyme biochemistry and protein structure.

Summary updated

Abstract

ABSTRACT Chinese hamster ovary (CHO) cells are most prevalently used for producing recombinant therapeutics in biomanufacturing. Recently, more rational and systems approaches have been increasingly exploited to identify key metabolic bottlenecks and engineering targets for cell line engineering and process development based on the CHO genome-scale metabolic model which mechanistically characterizes cell culture behaviours. However, it is still challenging to quantify plausible intracellular fluxes and discern metabolic pathway usages considering various clonal traits and bioprocessing conditions. Thus, we newly incorporated enzyme kinetic information into the updated CHO genome-scale model ( i CHO2291) and added enzyme capacity constraints within the flux balance analysis framework (ecFBA) to significantly reduce the flux variability in biologically meaningful manner, as such improving the accuracy of intracellular flux prediction. Interestingly, ecFBA could capture the overflow metabolism under the glucose excess condition where the usage of oxidative phosphorylation is limited by the enzyme capacity. In addition, its applicability was successfully demonstrated via a case study where the clone- and media-specific lactate metabolism was deciphered, suggesting that the lactate-pyruvate cycling could be beneficial for CHO cells to efficiently utilize the mitochondrial redox capacity. In summary, i CHO2296 with ecFBA can be used to confidently elucidate cell cultures and effectively identify key engineering targets, thus guiding bioprocess optimization and cell engineering efforts as a part of digital twin model for advanced biomanufacturing in future.
BibTeX
@inproceedings{yeo2020enzyme,
  author = {Hock Chuan Yeo and Jongkwang Hong and Meiyappan Lakshmanan and Dong‐Yup Lee},
  title = {{Enzyme capacity-based genome scale modelling of CHO cells}},
  booktitle = {Metabolic},
  year = {2020},
  doi = {10.1016/j.ymben.2020.04.005},
}
Figure