Contextualized genome-scale model unveils high-order metabolic effects of the specific growth rate and oxygenation level in recombinant Pichia pastoris

Abstract

Pichia pastoris is recognized as a biotechnological workhorse for recombinant protein expression. The metabolic performance of this microorganism depends on genetic makeup and culture conditions, amongst which the specific growth rate and oxygenation level are critical. Despite their importance, only their individual effects have been assessed so far, and thus their combined effects and metabolic consequences still remain to be elucidated. In this work, we present a comprehensive framework for revealing high-order (i.e., individual and combined) metabolic effects of the above parameters in glucose-limited continuous cultures of P. pastoris, using thaumatin production as a case study. Specifically, we employed a rational experimental design to calculate statistically significant metabolic effects from multiple chemostat data, which were later contextualized using a refined and highly predictive genome-scale metabolic model of this yeast under the simulated conditions. Our results revealed a negative effect of the oxygenation on the specific product formation rate (thaumatin), and a positive effect on the biomass yield. Notably, we identified a novel positive combined effect of both the specific growth rate and oxygenation level on the specific product formation rate. Finally, model predictions indicated an opposite relationship between the oxygenation level and the growth-associated maintenance energy (GAME) requirement, suggesting a linear GAME decrease of 0.56 mmol ATP/g_DCW per each 1% increase in oxygenation level, which translated into a 44% higher metabolic cost under low oxygenation compared to high oxygenation. Overall, this work provides a systematic framework for mapping high-order metabolic effects of different culture parameters on the performance of a microbial cell factory. Particularly in this case, it provided valuable insights about optimal operational conditions for protein production in P. pastoris.

Keywords: DO, dissolved oxygen; Dissolved oxygen; Experimental design; FBA, flux balance analysis; FVA, flux variability analysis; GAME, growth-associated maintenance energy; GSMM, genome-scale metabolic model; Metabolic modelling; NGAME, non-growth-associated maintenance energy; Pichia pastoris; Recombinant protein; Thaumatin; ll-FBA, loopless flux balance analysis.

Fig. 1

Overview of the experimental design and GSMM contextualization workflow employed in this study. A) Doehlert fractional experimental design for evaluating the metabolic impact of the specific growth rate (μ) and dissolved oxygen (DO) level on the metabolic performance of P. pastoris. This design considers 1 central point and 6 extreme points over 3 DO levels (4%, 30% and 56% oxygen saturation), and 5 μ levels (0.05, 0.075, 0.1, 0.125 and 0.15 h⁻¹). B) Manual GSMM contextualization workflow for flux simulations. The latest P. pastoris GSMM (iMT1026 v3.0) was incrementally constrained for accurately describing metabolic fluxes under glucose-limited conditions using reported experimental data and flux feasibility criteria.

Fig. 2

Response surfaces of P. pastoris macroscopic culture parameters under different DO and μ conditions. The fitted response surfaces were for the (encoded) biomass yield (A), specific thaumatin production rate (B) and thaumatin yield (C) are, respectively: $Y_{\text{S,X}}=0.565+0.036\cdot \underset{‾}{D}O-0.02\cdot \underset{‾}{\mu }\cdot \underset{‾}{D}O-0.017\cdot \underset{‾}{D}{O}^2$ (R² = 0.948), $q_{\text{P}}=0.033-0.004\cdot \underset{‾}{D}O+0.008\cdot \underset{‾}{\mu }\cdot \underset{‾}{D}O$ (R² = 0.68), and $Y_{\text{S,P}}=0.186-0.127\cdot \underset{‾}{\mu }+0.08\cdot {\underset{‾}{\mu }}^2$ (R² = 0.959), where $\mu =0.05+0.05\cdot (\underset{‾}{\mu }+1)\text{ , }\underset{‾}{\mu }\in [-\mathrm{1,1}]$ and $DO=4+30\cdot (\underset{‾}{D}O+0.8666)\text{ , }\underset{‾}{D}O\in [-0.8666,0.8666]$. Further details about the regression and statistical effects can be found in Supplementary Table S6.

Fig. 3

Evaluation of contextualized P. pastoris GSMM for metabolic flux prediction under glucose-limited conditions. A) Step-by-step contextualization process of the P. pastoris iMT1026 v3.0 GSMM for describing growth under glucose-limited conditions. B) Evaluation of maximum specific growth rate predictive performance of the initial versus contextualized GSMM. C) Comparison of the predicted intracellular flux through cytoplasmic reactions of the initial (slope = 0.911, R² = 0.783) and contextualized GSMM (slope = 1.047, R² = 0.972) against experimental data under appropriate growth conditions. D) Comparison of the predicted intracellular flux through mitochondrial reactions of the initial (slope = 0.112, R² = 0.076) and contextualized GSMM (slope = 0.721, R² = 0.867) against experimental data under the same conditions used in C). For more information about the employed experimental conditions refer to Materials and Methods.

Fig. 4

Evaluation of the DO influence on the GAME and NGAME requirements of P. pastoris under glucose-limited conditions. A) Comparison of the specific growth rate predictions of the model with a fixed GAME for all DO conditions (slope = 0.752, R² = 0.871), the model with a variable GAME for each DO condition (slope = 0.844, R² = 0.949), and the model with a variable NGAME for each DO condition (slope = 0.890, R² = 0.886) against the experimental growth data of this study. The model with a variable GAME for each DO condition has stronger statistical support as shown by the lower AIC_c value (-158.9 versus -146.8 and -154.7). B) Predicted GAME values for different DO conditions (95 mmol ATP g_DCW⁻¹ for the L condition, 75 mmol ATP g_DCW⁻¹ for the M condition, and 66 mmol ATP g_DCW⁻¹ for the E conditions).

Fig. 5

Flux distribution analysis of P. pastoris metabolism under extreme DO conditions. A) Predicted metabolic flux distributions under extreme DO conditions (4% and 56% oxygen saturation) at μ = 0.075 h⁻¹. Parsimonious FBA (Lewis et al., 2010) was employed to obtain a representative flux distribution under each condition. B) Feasible flux ranges for main central carbon reactions of P. pastoris under extreme oxygenation conditions. Flux ranges were calculated using loopless FVA (Saa and Nielsen, 2016a) as described in the Materials and Methods section. Abbreviations: HEX, hexokinase; PGI, glucose-6-phosphate isomerase; TPI, triose-phosphate isomerase; GAPD, glyceraldehyde-3-phosphate dehydrogenase; ENO, enolase; PYK, pyruvate kinase; G6PDH2, glucose-6-phosphate dehydrogenase; PGL, 6-phosphogluconolactonase; GND, phosphogluconate dehydrogenase; RPI, ribose-5-phosphate isomerase; TKT1(2), transketolase 1(2); PYRt2m, pyruvate transport (mitochondrial); PDH1a, pyruvate dehydrogenase; CSm, citrate synthase (mitochondrial); ACONTm, aconitate hydratase (mitochondrial); ICDHxm, isocitrate dehydrogenase NAD-dependent (mitochondrial); ICDHym, isocitrate dehydrogenase NADP-dependent (mitochondrial); AKGDbm, oxoglutarate dehydrogenase dihydrolipoamide S-succinyltransferase (mitochondrial); SUCOASm, oxoglutarate dehydrogenase dihydrolipoamide S-succinyltransferase (mitochondrial); MDHm, malate dehydrogenase (mitochondrial); PC, pyruvate carboxylase; PYRDC, pyruvate decarboxylase; PGMT, phosphoglucomutase.

Fig. 6

Phase plot of optimal P. pastoris growth and thaumatin production as a function the specific growth rate and dissolved oxygen. The optimal conditions for efficient growth (dashed line) and increased thaumatin production (continuous line) are shown in thick black lines, whereas suboptimal conditions are represented with thin lines using the same symbols as before. As shown in the figure, optimal growth lie opposite to optimal production conditions.