Development of a core Clostridium thermocellum kinetic metabolic model consistent with multiple genetic perturbations

Abstract

Background: Clostridium thermocellum is a Gram-positive anaerobe with the ability to hydrolyze and metabolize cellulose into biofuels such as ethanol, making it an attractive candidate for consolidated bioprocessing (CBP). At present, metabolic engineering in C. thermocellum is hindered due to the incomplete description of its metabolic repertoire and regulation within a predictive metabolic model. Genome-scale metabolic (GSM) models augmented with kinetic models of metabolism have been shown to be effective at recapitulating perturbed metabolic phenotypes.

Results: In this effort, we first update a second-generation genome-scale metabolic model (iCth446) for C. thermocellum by correcting cofactor dependencies, restoring elemental and charge balances, and updating GAM and NGAM values to improve phenotype predictions. The iCth446 model is next used as a scaffold to develop a core kinetic model (k-ctherm118) of the C. thermocellum central metabolism using the Ensemble Modeling (EM) paradigm. Model parameterization is carried out by simultaneously imposing fermentation yield data in lactate, malate, acetate, and hydrogen production pathways for 19 measured metabolites spanning a library of 19 distinct single and multiple gene knockout mutants along with 18 intracellular metabolite concentration data for a Δgldh mutant and ten experimentally measured Michaelis-Menten kinetic parameters.

Conclusions: The k-ctherm118 model captures significant metabolic changes caused by (1) nitrogen limitation leading to increased yields for lactate, pyruvate, and amino acids, and (2) ethanol stress causing an increase in intracellular sugar phosphate concentrations (~1.5-fold) due to upregulation of cofactor pools. Robustness analysis of k-ctherm118 alludes to the presence of a secondary activity of ketol-acid reductoisomerase and possible regulation by valine and/or leucine pool levels. In addition, cross-validation and robustness analysis allude to missing elements in k-ctherm118 and suggest additional experiments to improve kinetic model prediction fidelity. Overall, the study quantitatively assesses the advantages of EM-based kinetic modeling towards improved prediction of C. thermocellum metabolism and develops a predictive kinetic model which can be used to design biofuel-overproducing strains.

Keywords: Clostridium thermocellum; Ensemble modeling; Ethanol stress; Genome-scale metabolic model; Kinetic model; Nitrogen limitation.

Fig. 1

Summary of modifications in the iCth446 GSM model after updating it from the previous iSR432 reconstruction. Updated reactions in iCth446: The cofactors highlighted in green (in reactions HEX1, PFK, PGK, and ME) were added in the model and those in red (for reactions HEX1, PFK, and ME) were removed. In addition, reactions ME and PEPCK were made reversible. Dashed lines in gray indicate an example of thermodynamically infeasible cycle of three reactions (ODC, MDH, and ME). The cycle was resolved by removing ODC. NFN was added to the model to allow electron transfer between reducing equivalents. The values alongside the reactions are their FBA-predicted fluxes (in mmol/gDW/h) consistent with the wild-type experimental cellobiose uptake and growth rates [65]

Fig. 2

List of 19 fermentation mutants [36]. The figure shows the corresponding gene knockout (X), downregulation (↓), or upregulation (↑) followed by a comparison of ethanol yield ranges predicted by the GSM and the kinetic model with the experimentally reported values. The iCth446 predictions were performed by restricting all the metabolite yields to their measured ranges except for ethanol. Table 3 enumerates the strains associated with specific mutants

Fig. 3

Impact of limiting nitrogen source in the media on C. thermocellum metabolism. The iCth446 simulations were performed by restricting the ammonia uptake flux to 20% of its wild-type value and maximizing the yield of specific metabolites one at a time

Fig. 4

Core metabolic map of C. thermocellum. The arrows in orange represent the extracellular metabolites, the concentrations and molar yields of which were experimentally measured

Fig. 5

Cross-validation analysis of k-ctherm118 model. The gray bars represent the average scaled deviation of the predicted steady-state fluxes upon cross-validation of the training dataset. The first 19 datasets represent mutants with experientially measured fermentation yields, the penultimate dataset represents Δgldh mutant with measured intracellular concentrations, and the final dataset represents the experimentally measured Michaelis–Menten constants. The white bars correspond to the average scaled deviation of the predicted steady-state flux distribution from the experimental measurements while including all datasets. The difference between two bars represents the reduction in the accuracy of the model predictions upon excluding the flux dataset of the corresponding mutant

Fig. 6

Impact of limiting nitrogen source in the media on C. thermocellum metabolism. a Change in pathway fluxes and metabolite concentration on reducing ammonium uptake activity to 20% of the wild-type activity. The numbers represent the flux values (mol/mol cellobiose) normalized to cellobiose uptake (except for those in parentheses represent concentration change) for the wild-type (gray color) to ammonia-limited (green or red color representing upregulation or downregulation, respectively) conditions. All fluxes were normalized to cellobiose uptake. b Comparison of model-predicted yield change with the experimentally measured values

Fig. 7

Overall impact of ethanol stress on C. thermocellum metabolism. a Affected pathway fluxes and metabolite concentrations under ethanol stress condition. The numbers represent the flux values (mol/mol cellobiose) normalized to cellobiose uptake (except for those in brackets represent concentration change) for the wild-type (gray color) to ammonia-limited (green or red color representing up- or downregulation, respectively) conditions. All fluxes were normalized to cellobiose uptake. b Comparison of model-predicted cytosolic concentration changes with the experimentally measured values

Fig. 8

Carbon dioxide export upregulation revealing the ketol-acid reductoisomerase reaction with non-robust kinetic parameters. Possible substrate-level regulation (I) and secondary function (II) observed in C. glutamicum can improve the parameter robustness