Pathway Analysis of Pichia pastoris to Elucidate Methanol Metabolism and Its Regulation for Production of Recombinant Proteins Pornkamol Unrean Biochemical Engineering and Pilot Plant Research and Development Unit, King Mongkut’s University of Technology Thonburi, Thakhum, Bangkhuntien, Bangkok 10150, Thailand National Center for Genetic Engineering and Biotechnology (BIOTEC), Klong 1, Klong Luang, Pathum Thani 12120, Thailand DOI 10.1002/btpr.1855 Published online December 21, 2013 in Wiley Online Library (wileyonlinelibrary.com)

This research rationally analyzes metabolic pathways of Pichia pastoris to study the metabolic flux responses of this yeast under methanol metabolism. A metabolic model of P. pastoris was constructed and analyzed by elementary mode analysis (EMA). EMA was used to comprehensively identify the cell’s metabolic flux profiles and its underlying regulation mechanisms for the production of recombinant proteins from methanol. Change in phenotypes and flux profiles during methanol adaptation with varying feed mixture of glycerol and methanol was examined. EMA identified increasing and decreasing fluxes during the glycerol–methanol metabolic shift, which well agreed with experimental observations supporting the validity of the metabolic network model. Analysis of all the identified pathways also led to the determination of the metabolic capacities as well as the optimum metabolic pathways for recombinant protein synthesis during methanol induction. The network sensitivity analysis revealed that the production of proteins can be improved by manipulating the flux ratios at the pyruvate branch point. In addition, EMA suggested that protein synthesis is optimum under hypoxic culture conditions. The metabolic modeling and analysis presented in this study could potentially form a valuable knowledge base for future research on rational design and optimization of P. pastoris by determining target genes, pathways, and culture conditions for enhanced recombinant protein synthesis. The metabolic pathway analysis is also of considerable value for production of therapeutic proteins by P. pastoris in biopharC 2013 American Institute of Chemical Engineers Biotechnol. maceutical applications. V Prog., 30:28–37, 2014 Keywords: Pichia pastoris, methanol metabolism, metabolic pathway analysis, recombinant protein synthesis

Introduction Pichia pastoris is a methylotrophic yeast that is considered to be a valuable and cost-effective host for production of a wide variety of therapeutic proteins used in the biopharmaceutical industry.1–3 This yeast expression system offers several advantages, including its ability to perform protein folding and glycosylation, similar to higher eukaryotes.4 Over the last 2 decades, the acquisition of profound knowledge of genetics, molecular biology, biochemistry, and entire genome sequences of the organism5,6 has made it possible to pursue a systems biology approach for the understanding of P. pastoris cellular processes. P. pastoris has been developed as an expression platform for foreign protein production using a methanol-inducible alcohol oxidase promoter (AOX). This promoter permits the expression of protein through induction with methanol. The

Additional Supporting Information may be found in the online version of this article. Correspondence concerning this article should be addressed to P. Unrean at [email protected]. 28

typical process of protein production in P. pastoris involves three phases: glycerol growth phase, methanol adaption phase, and methanol induction phase. First, the cells are grown on glycerol, which allows the culture to reach high cell density. Following the glycerol growth mode, a mixture of glycerol and methanol is added to the culture, which permits the cell to adapt to the methanol substrate. Subsequently, the production of recombinant proteins is induced by feeding methanol to the culture, which activates the AOX promoter controlling heterologous genes for recombinant protein synthesis. Although comprehensive investigations of the methanol metabolism for other methylotrophic yeasts such as Hansenula polymorpha, Kloeckera sp., Candida boidinii are available, little is known about the underlying regulation mechanisms of methanol metabolism that govern the production of recombinant proteins in the P. pastoris expression system. It is well known that environmental perturbations have major effects on the metabolic behavior of cells. Previous studies have reported that the shift of the substrate to methanol exerts stress and causes major physiological changes in the yeast cells.7–9 The extent of such metabolic shift during C 2013 American Institute of Chemical Engineers V

Biotechnol. Prog., 2014, Vol. 30, No. 1

the adaptation is certainly an important issue for the protein production process. It is of significance to elucidate how this substrate shift during the methanol adaptation phase influences the metabolic flux profiles of P. pastoris. The questions of which key enzymes are induced during the adaptation to growth on methanol are also of interest. Additionally, the impact of oxygen availability on heterologous protein production during the methanol induction is essentially unexplored. As a consequence, we applied metabolic pathway analysis to study the effects of such perturbations during adaptation and induction on the production of recombinant proteins, with the aim to systematically identify the metabolic mechanisms ruling efficient protein production from methanol. Studies of the flux responses under these stresses might provide an important platform to identify targets for rational engineering of P. pastoris to enhance its performance in protein synthesis. A metabolic pathway analysis is a valuable tool for metabolic flux distribution studies. A pathway analysis technique commonly used is elementary mode analysis (EMA). EMA identifies all feasible flux distributions available in the reaction network on the basis of stoichiometric mass balance constraints.10 Cells use these defined fluxes based on their genetic backgrounds and/or the culture conditions given. The functioning metabolic fluxes of a cell can be described as a non-negative linear combination of these identified flux patterns. Thus, with EMA, the pathway structure as well as the theoretical metabolic capabilities of the cell can be determined. The EMA approach has been successfully used, for instance, to study metabolic network properties of Escherichia coli,11 and to guide the metabolic engineering of several desired products such as amino acids.12 The same approach is also used to study flux patterns affected by culture conditions such as type of substrates13 or oxygen levels.14 In this study, we used EMA to elucidate the flux distributions of P. pastoris metabolism under culture conditions that simulate the experimental conditions during recombinant protein production. The study focuses on assessing and characterizing the metabolism of P. pastoris during the methanol adaption phase (growing on glycerol–methanol mixture), and during the induction phase (growing on methanol). Specifically, EMA was exploited to study the metabolic flux responses to the substrate shift to methanol during adaption, and to examine the change in flux distributions involved in the expression of recombinant proteins during induction. This analysis could provide comprehensive insight into the methanol metabolism of P. pastoris during recombinant protein synthesis. Metabolic pathway analysis of P. pastoris presented in this work might facilitate strain optimization for efficient production of recombinant proteins. For a comprehensive investigation of heterologous protein biosynthesis, we examined the production of two example recombinant proteins, green fluorescent protein (GFP) and human growth hormone (hGH) in P. pastoris. Green fluorescent protein was selected because this protein is commonly used for studying protein expression mechanisms due to its autofluorescence. Human growth hormone is a protein of commercial interest that is used for treating patients with growth hormone deficiency and other growth disorders. Analysis of the pathways inherent to the cell metabolic network identified the most efficient paths for the synthesis of these proteins. The metabolic flux distributions and flux ratios for different proteins were also compared.

29

Materials and Methods Construction of P. pastoris metabolic reaction network A metabolic network model of P. pastoris was constructed based on the previously published literature on its sequenced genomic information, its metabolism and its available pathway databases. The culturing period of methanol adaptation and induction during protein expression serves as the basis for the model simulation. The model analyzed two culturing phases: (1) glycerol–methanol mixture feeding during adaption and (2) methanol feeding during induction. The constructed network model, therefore, represented P. pastoris growing on glycerol/methanol mixture during adaptation and on methanol during induction. Details on the constructed metabolic model are described in Supporting Information Table S1 and Figure S1. It should be noted that the direction of glycerol transport reaction (R109) was designated as being consumed for growth on methanol–glycerol mixture. However, the direction was changed to being produced for growth on methanol. The production of recombinant proteins, GFP and hGH, was incorporated in the metabolic model as case studies. There were 85 reactions included in the model with 28 classified as reversible, while 57 reactions were classified as irreversible. For simplicity, a linear series of reactions were lumped together as one. Thus, the reactions might not necessarily represent one gene or enzyme. Twelve external metabolites involved in the reaction network were methanol, glycerol, NH3, acetate, ethanol, succinate, biomass, recombinant proteins (GFP and hGH), O2, CO2, and ATP. The internal metabolites are defined as metabolites with no accumulation and are constrained by steady-state mass balance. The external metabolites serve as source or sink of the network. Thus, they can be accumulated and are not constrained by the steady-state assumption.11 The model constituted reactions from intermediary metabolism: glycolysis, pentose-phosphate pathway, citric acid cycle, glycerol utilization pathway, methanol utilization pathway, biomass formation, and biosynthesis pathways of common byproducts of P. pastoris cultures. Reactions and metabolites were localized into compartments: cytosol, peroxisome, and mitochondria. Transporters between the compartments were also incorporated in the model. Metabolites and cofactors were treated as two distinct species, distributed between mitochondrial and cytosolic reactions. Reactions in peroxisome were treated the same as in cytosol. Information on reversibility and localization of reactions was taken from a MetaCyc pathway database (www.metacyc.org).

Biomass synthesis pathway The biomass synthesis pathway was reconstructed by accounting for metabolite drain from the central metabolic pathways as previously described in Varma and Palsson.15 Estimation of the biomass term was based on the cell dry weight composition reported in the literature.5,16 This information was used for calculating the molar requirements of biosynthetic precursors and redox cofactors required for production of biomass, which is composed of proteins, nucleotides, lipids, and carbohydrates. Energy requirements for biomass production, including polymerization energy cost and energy costs for the synthesis of biosynthetic precursors, were also included. The maintenance energy represented by excess ATP to be consumed for maintenance processes was treated separately. The biomass term included nine central metabolism precursor molecules, which

30

Biotechnol. Prog., 2014, Vol. 30, No. 1

Table 1. Metabolite requirements and overall stoichiometry of making one unit of biomass, GFP, and hGH. G6P G3P R5P E4P PEP PYR ACCOA AKG OA NH3 Biomass GFP hGH

5,629 0 0

603 34 30

426 23 17

116 12 9

384 47 43

1,511 116 90

1,171 20 26

1,046 40 46

961 85 51

5,556 319 263

ATP

NAD

NADPH

35,315 333 267

1,506 0 0

7,554 530 420

A different metabolic precursors/cofactors requirement from the intermediate metabolism of each protein is a result of the difference in its amino acid composition. Abbreviations: G6P: glucose-6-phosphate, G3P: glyeraldehyde-3-phosphate, R5P: ribose-5-phosphate. E4P: erythose-4-phosphate, PEP: phosphoenolpyruvate, PYR: pyruvate, ACCOA: acetyl coenzyme A, AKG: a-ketoglutarate and OAA: oxaloacetate. All units are in mmol.

were glucose-6-phosphate (G6P), glyceraldehyde-3-phosphate (G3P), ribose-5-phosphate (R5P), erythrose-4-phosphate (E4P), phosphoenolpyruvate (PEP), pyruvate (PYR), acetyl coenzyme A (ACCOA), a-ketoglutarate (AKG), and oxaloacetate (OAA) along with ATP, NAD/NADH, NADP/NADPH, and NH3 (see Supporting Information Table S1 for details). Recombinant protein synthesis pathway The synthesis pathways for GFP and hGH were reconstructed by computing the metabolic requirement for the synthesis of each considered protein. The pathways account for all the corresponding metabolic drain from the central metabolic pathway to synthesize all required amino acids necessary to make up the target protein. Information about amino acid composition of the expressed proteins was obtained from National Center for Biotechnology Information database. The energy requirements for the production of each peptide bond were also considered. Approximately four ATP equivalents are required to form each polypeptide bond.17 The precursor and cofactor requirements for the synthesis of biomass and recombinant proteins (GFP and hGH) are illustrated in Table 1. It is to be noted that the metabolic requirement for synthesis of the proteins varied, due to the unique amino acid composition of each protein. In addition, the production of biomass in the model does not include the production of the expressed proteins under investigation. That is, the biomass and the recombinant proteins were treated as two separate metabolites.

Results Analysis of P. pastoris metabolic network The network model of P. pastoris was analyzed by EMA to attain more comprehensive understanding of the metabolic capabilities, the metabolic flux distributions, and the regulation of metabolic flux profiles of the organism for the production of heterologous proteins from methanol. The model analysis was used first to examine existing network characteristics of the P. pastoris model, and then to identify catalytic activities and culturing conditions that impact the expressed proteins. We analyzed the model on two culturing phases of the protein production process: methanol adaptation and methanol induction phases. The metabolic richness of a network is determined based on the total number of possible available pathways.11 The constructed network of the P. pastoris system yielded 18,437 pathways during methanol adaptation based on growth supported on a glycerol–methanol mixture, and 6,627 pathways during methanol induction based on growth supported on methanol. Of these pathways, there existed 17,730 and 4,752 available pathways supporting cell growth on glycerol–methanol mixture and on methanol, respectively. Of the pathways available during the induction, there were 664 pathways for the recombinant GFP protein synthesis and 675 pathways for the recombinant hGH protein synthesis. The pathways identified by EMA can be sorted according to the protein yield, which is defined as the ratio of produced protein (mmol) to used substrates (gram), to determine their efficiency. The most efficient pathways for GFP and hGH on methanol produced a maximum yield of 29.32 mmol-GFP/gmethanol and 35.73 mmol-hGH/g-methanol, respectively.

Metabolic network analysis The publicly available program METATOOL version 5.1 (http://pinguin.biologie.uni-jena.de/bioinformatik/networks/ metatool/metatool5.0/metatool5.0.html) was used for EMA. The program decomposed a network into all possible, unique, and balanced metabolic pathways, allowing the calculation of a phenotypic space from all the pathways. Network characteristics were examined by sorting through all the identified pathways. The output files from the METATOOL, which presented all relative fluxes through each reaction for every pathway in matrix format, were analyzed using a Microsoft Excel spreadsheet (Microsoft) for sorting, filtering, and plotting of pathway results based on product yield, such as biomass and expressed proteins. The metabolic capabilities of P. pastoris for growth and for production of recombinant proteins were assessed by comparing relative fluxes and/or product yields among the identified pathways. The yield is defined as the flux ratio of the amount of the product produced to that of the substrates (methanol and/or glycerol) consumed. The yields were sorted to identify maximum capacity for the product synthesis of the studied network. Efficiency of the pathways was determined on the yield basis.

Phenotypic changes during the methanol adaption phase It is of importance to understand how the substrate change from glycerol to methanol during the adaption phase impacts P. pastoris metabolism. We used EMA to predict the phenotypic response of the organism on the glycerol–methanol feed ratio. Feed ratio of glycerol–methanol consumption between 0.01 and 2 was investigated because, within this range, co-assimilation of glycerol and methanol has been observed.18 The impact of glycerol–methanol feed ratio on biomass synthesis (Figure 1A) revealed that the biomass yield decreased significantly as the fraction of methanol in the feed was increased. The result is expected, as higher biomass yield can be achieved when growing on glycerol compared to on methanol. Substrate shift from glycerol to methanol also affected the respiratory quotient (RQ), a ratio of carbon evolution rate, and oxygen uptake rate (Figure 1B). The RQ was decreased with increasing methanol feed fraction. EMA also suggested higher secretion of ethanol and succinate during the metabolic shift of the cells from glycerol to methanol substrate (result not shown). The experimental results also fell within the predictive range, thus validating the model prediction.

Biotechnol. Prog., 2014, Vol. 30, No. 1

31

mixture, and methanol were quantitatively different (Figure 2). The molar fluxes through key enzymes, for example, formaldehyde dehydrogenase (R106r; FALDH), formate dehydrogenase (R107; FDH), and formylglutathione hydralase (R108; SFGTH) in the methanol utilization pathway were elevated during the adaptation to growth on methanol. It can also be seen that the fluxes corresponding to the interconversion of succinate–malate in cytosol and mitochondria, R27 (FRD), R28r (FUM), R79r (SUC; SDH), R80r (FUM; OSM), R89r, and R100, were increased during the transition from glycerol growth to methanol growth conditions. Interestingly, the model also predicted inverted directionality of the flux in glycerol-3-phosphate dehydrogenase (R13r) during the metabolic shift. The activity of anaplerotic reactions, phosphoenolpyruvate carboxykinase (R42) and malic enzyme (R83), appeared to be increased when growing on methanol as predicted by EMA. On the contrary, glycolytic fluxes, R3r (PGI), R7r (TPI), and R8r (TDH; PGK; PGM; ENO), were reduced during the metabolic shift towards methanol. Several cytosol–mitochondria transport reactions (R88r, R90r, R92r, and R94r) were also affected during the metabolic shift. Comparative flux distribution profiles of P. pastoris growing on different glycerol–methanol feed ratios were simulated (Figure 3). Fluxes through R106r, R107, and R108 for methanol assimilation pathways were only activated at high fractions of methanol in the feed. In addition, the activation of the succinate–malate conversion pathway at high methanol feed ratio suggested the significance of this pathway in methanol metabolism. The model also predicted zero flux through glucose 6-phosphate dehydrogenase (R15) at low fractions of methanol feed and non-zero flux at high methanol feed fractions. Activation of this enzyme at increasing methanol feed fractions may be involved in sustaining the redox balance for methanol metabolism. Several of key fluxes in the citric acid cycle, R24 (KGD; LPD), R25r (LSC), R75 (IDH; IDP), and R76 (ICDH) also appeared to be inactive as methanol fraction in the feed increased.

Figure 1. Phenotypic changes of P. pastoris during methanol adaptation phase. Effect of glycerol–methanol feed ratio on (A) biomass and (B) RQ. The yield is based on gram of total used substrates, methanol, and glycerol. RQ is defined as flux ratio of carbon dioxide secretion rate and oxygen uptake rate. The feed ratio of glycerol–methanol is ranging from 0.01 to 2 mmol-methanol/mmol-glycerol. The triangle symbols represent metabolic states of the P. pastoris system observed experimentally as reported in Sola et al.18 These experimental points lie within the shade area which illustrates the predictive possible solution space derived from EMA. The consistency between the experimental observation and the model prediction confirms the accuracy of the constructed network. Circles (P1–P7) represent the most efficient pathways at different glycerol–methanol consumption ratio. All culturing conditions under these feeding of glycerol–methanol mixture can be represented as an envelope of line segments that links the seven identified pathways (P1–P7). All these metabolic states are then defined by linear combinations of the seven pathways.

Comparative flux patterns during growth of glycerol–methanol mixture Molar fluxes through each reaction in the P. pastoris metabolic network growing on glycerol, glycerol–methanol

Metabolic flux profiles during glycerol–methanol metabolic shift During the adaption, a mixture of glycerol and methanol was fed to the culture at increasing ratios of methanol. The analysis of P. pastoris metabolism during the metabolic shift from glycerol to methanol substrates was performed with the aim of understanding the regulatory network. All identified pathways under growth on the glycerol–methanol mixture can be represented through the most efficient metabolic states P1 through P7 for growth on the glycerol–methanol mixture, which formed the envelope as shown in Figure 1A. This envelope captured all available paths under a culture experiencing different glycerol/methanol ratios ranging from 0.01 to 2.0. All pathways can be represented by a linear combination of these seven pathways. For a better understanding of regulatory networks during methanol adaptation, the flux profiles affected during the metabolic shift towards the methanol growth state were examined on the basis of these pathways. The summary of metabolic flux profiles at increasing glycerol/methanol ratios in Figure 4 reveals that the transition from glycerol growth to methanol growth required the up-regulation of several key reactions in the methanol metabolism, R104 (DAK) and R105r (DAS); for glycolysis, R5 (FBP) and R10 (PYK); and for the pentose phosphate pathway, R17 (RPE), R18 (TKT), and R19 (TKT). The fluxes through reaction R23

32

Biotechnol. Prog., 2014, Vol. 30, No. 1

Figure 2. Predictive range of molar flux through different reactions in the P. pastoris network under growth on glycerol, glycerol– methanol, and methanol. The plot shows the different flux values through each of the reactions as predicted by EMA upon change in substrate: glycerol in red, glycerol–methanol in green, and methanol in blue. EMA identifies reactions with increasing (A) or decreasing (B) fluxes during the metabolic shift from growth on glycerol to growth on methanol. These reactions could be targeted for a genetic switch to control between growth on glycerol and on methanol. Fluxes are represented as molar flux of the reaction per fluxes of substrate consumed. The corresponding enzyme and gene for each reaction are given in Supporting Information Table S1.

(IDH), R30 (GUT), R41 (ALD), R3r (PGI), R7r (TPI), and R13r (GPD) are also predicted to be downregulated as the fraction of methanol in the feed is increased. Optimum pathways for recombinant protein synthesis EMA permits an identification of all alternative operational pathways of a metabolic network. These pathways can be compared according to their efficiency in terms of recombinant protein yield under methanol induction conditions. The reactions involved for the most efficient production for the two proteins analyzed, GFP and hGH, in comparison with the biomass synthesis, are shown in Table 2. These optimum pathways shared certain common features of used reactions. For example, reactions in the pentose phosphate pathway were reversed, and the citric acid cycle was incomplete. Gluconeogenesis reactions were also used. No byproduct other than carbon dioxide was produced. An intriguing difference between the optimum pathways for recombinant proteins and biomass was fluxes around the pyruvate branch point. The protein-producing pathways did not use malic enzyme (R82), while the biomass-producing pathways did. In addition, the directions of the flux through malate dehydrogenase (R81r) and the pyruvate mitochondrial transporter (R94r) for the protein synthesis were reversed compared with the corresponding reactions for biomass synthesis. EMA also revealed that the flux through some of the reactions for protein synthesis was quantitatively different, compared with that for the biomass synthesis (Figure 5). Fluxes entering glycolytic and pentose phosphate reactions were higher during protein synthesis than biomass synthesis on methanol. Knowing all the essential reactions that must be active for the most efficient protein production during the induction phase could aid in identifying target genes for further strain development. Ideally, the other reactions that are not required might be removed through target gene deletion to force the cellular metabolism into operating via the most efficient pathways.

Network sensitivity analysis of recombinant protein synthesis The network sensitivity to protein production was examined by investigating the influence of the change of relative flux ratios at the branch node of PYR on the recombinant protein yield (Figure 6). The PYR node is defined as a branch point between glycolysis and the citric acid cycle. The fluxes around the pyruvate node include pyruvate carboxylase (R37), pyruvate decarboxylase (R38), and the pyruvate mitochondrial transporter (R94r). It was found that the protein yield of both GFP and hGH trended to decrease as the flux partitioning at the pyruvate node increased. The yield of GFP decreased dramatically by 98% as the flux ratio at the pyruvate node increased by 57%. Likewise, the yield of hGH declined by 47% as the flux partitioning at the pyruvate node increased by 49%. It is likely that increasing flux partitioning at this node results in more flux being drained towards biomass instead of protein synthesis. Ideally, the existing network could be manipulated through genetic engineering of reactions around the PYR branch point, such that the flux partition is redirected towards the desired protein synthesis pathway.

Effect of oxygen availability on protein synthesis Production of heterologous proteins is impacted by a number of culturing parameters, including oxygen levels. Oxygen availability strongly influences P. pastoris metabolism by causing energy deprivation and affecting cellular redox reactions, which could have an impact on the protein synthesis pathway. Thus, EMA was performed to investigate the impact of oxygen supply on the metabolic capabilities of protein production. The plot in Figure 7 presents the yield of the expressed proteins vs. the oxygen consumption. Although both GFP and hGH required different precursors/cofactors for expression in P. pastoris, the optimum oxygen consumption for the production of the two proteins was quite similar.

Biotechnol. Prog., 2014, Vol. 30, No. 1

Figure 3.

33

Optimum flux distributions of P. pastoris network growing under different mixture of glycerol–methanol feeding as predicted by EMA. (A) Glycerol–methanol feed ratio at 0.7 mmol/mmol, (B) glycerol–methanol feed ratio at 1.9 mmol/mmol. These two feed ratios were selected for comparison purpose to study the effect of methanol feed fraction on flux distributions. Active fluxes (non-zero fluxes) are shown in red, while inactive fluxes (zero fluxes) are shown in gray.

It appeared that the yield of recombinant proteins increased along with an increase in oxygen consumption to 1.78 mmol-oxygen/mmol-methanol, suggesting a correlation between oxygen availability and recombinant protein synthe-

sis. The yields of proteins were decreased dramatically as oxygen levels increased further. The results revealed that high oxygenation was undesirable for protein synthesis. We hypothesize that at high oxygen levels, more flux entered the

34

Biotechnol. Prog., 2014, Vol. 30, No. 1

citric acid cycle and more NADH2 cofactor (produced via the citric acid cycle) was supplied for biomass production instead of protein synthesis. Thus, the increased protein production under oxygen limitation was the result of decreased fluxes towards the citric acid cycle and oxidative phosphorylation. The finding was consistent with previous studies that reported the increased production of heterologous proteins under hypoxic conditions compared to fully aerobic conditions.19–21 According to the model, the production of recombinant proteins could be improved by maintaining the oxygen levels below a threshold of 1.78 mmol of oxygen per mmol of methanol. Therefore, fermentation strategies can be designed to optimize recombinant protein synthesis by controlling the oxygen supply.

Discussion Understanding the pathway structure of the cellular metabolic network is important for characterizing an organism. Herein we used a pathway analysis tool called EMA for deconstructing the complex metabolic network of P. pastoris into simplified pathway units. P. pastoris is considered a valuable expression platform for many recombinant proteins. Analysis of these pathway possibilities provides unique insight into the structure of the P. pastoris metabolic network that can guide possible modification and optimization strategies for improving recombinant protein synthesis. The number of available pathways identified by EMA revealed different degrees of complexity and cellular flexibility under the two culturing phases examined, that is, adaptation and

Figure 4. Effect of glycerol–methanol feed fraction on metabolic fluxes in Pichia pastoris. Reactions (genes) predicted by EMA to be up-regulated as methanol feed ratio increases are shown in green. The down-regulated reactions as methanol feed fraction increases are shown in red. Metabolic flux is defined as reaction rate normalized by the consumption rate of both glycerol and methanol. Glycerol–methanol ratio is defined as the molar uptake rate of glycerol per that of methanol. Note that only reactions (gene) that has the statistically significant relationship of reactions and glycerol–methanol uptake ratio are shown.

Table 2. The Optimum Flux Distribution of Methanol Metabolism in P. pastoris Identified by EMA for the Most Efficient Production of GFP Protein, hGH Protein, and Biomass Extracellular Fluxes (mmol/cell-h)

Methanol Utilization

Pathway

JProduct

JO2

JCO 2

JNH 3

RQ

R104

R105r

R106r

R107

R108

GFP hGH Biomass

0.939 1.145 0.023

1.780 1.770 1.720

0.763 0.755 0.717

0.299 0.301 0.129

0.428 0.426 0.416

1.450 1.478 1.364

1.450 1.478 1.364

0.549 0.521 0.635

0.549 0.521 0.635

0.549 0.521 0.635

Glycolysis/Gluconeogenesis

Pentose Phosphate Pathway

Pathway

R3r

R5

R6r

R7r

R8r

R10

R15

R16r

R17r

R18r

R19r

R20r

GFP hGH Biomass

20.220 20.199 20.177

1.058 1.068 1.064

21.058 21.068 21.064

20.391 20.409 20.300

0.326 0.345 0.142

0.282 0.296 0.133

0.220 0.199 0.046

20.391 20.409 20.431

0.612 0.609 0.478

20.413 20.429 20.441

20.424 20.439 20.444

20.413 20.429 20.441

Citric Acid Cycle Pathway

R22r

R23

R27

R28r

R29r

R72

R73

R74r

R79r

R80r

R81r

GFP hGH Biomass

0.038 0.052 0.024

0.037 0.052 0.024

1.485 1.435 1.414

21.485 21.435 21.414

21.522 21.488 21.438

0.037 0.052 0.027

0.037 0.052 0.024

0.038 0.052 0.024

22.970 22.870 22.828

1.485 1.435 1.414

1.523 1.488 21.438

Pathway

R58

R88r

R89r

R90r

R94r

R100

R102r

R36

R38

R41

R37

R82

GFP hGH Biomass

1.523 1.488 1.438

20.037 20.052 20.024

21.522 21.487 21.438

0.038 0.052 0.024

0.037 0.052 20.007

1.485 1.435 1.414

21.523 21.487 21.437

0.018 0.029 0.024

0.018 0.029 0.024

0.018 0.029 0.024

0.117 0.111 0.081

0 0 0.034

Cytosol-Mitochondrial Transport

Acetate Synthesis

Anaplerotic Pathway

Only non-zero fluxes are shown. All fluxes are normalized with methanol uptake. Fluxes in reversed direction are shown in negative values. The fluxes around pyruvate node that are significantly distinct between biomass and protein synthesis fluxes are italicized. The most efficient pathway for protein synthesis does not require malic enzyme (R82) while the biomass does. This reaction may be used as a genetic switch for controlling between biomass and protein synthesis. The reactions malate dehydrogenase (R81r) and pyruvate mitochondrial transporter (R94r) are reversed for protein synthesis in comparison with the corresponding biomass flux. The maximum carbon yield is 29.32 mmol/g-methanol for GFP, 35.73 mmol/g-methanol for hGH, and 0.73 g/g-methanol for biomass.

Biotechnol. Prog., 2014, Vol. 30, No. 1

Figure 5.

35

Comparison of molar fluxes through different reactions in the most efficient pathways for production GFP, hGH, and biomass. The plot shows the different flux values through each reaction. The activities of these fluxes are reduced during growth in reference to protein synthesis on methanol. Fluxes are represented as molar fluxes of the reactions per fluxes of methanol consumed, which are shown in blue for GFP, red for hGH, and green for biomass. The corresponding enzyme and gene for each reaction are given in Supporting Information Table S1.

induction of methanol. The cellular system is less restricted to adjust itself for optimal fitness under mixed glycerol– methanol feed conditions, due to the larger number of available pathways, than under methanol feed conditions. This is in agreement with previous reports that feeding mixtures of glycerol and methanol (instead of only methanol) to the culture is an effective means for improving the methanol adaption process by shortening the adaption time.9,22 Phenotypic analysis predicted by EMA suggested reductions in both cell growth and RQ with increasing methanol feed ratio during the adaption (Figures 1A, B), which is consistent with experimental observations by Sola et al.18 The analysis results implied that the cell required higher oxygen consumption growing on methanol than on glycerol. This is expected, as the oxidation of methanol requires oxygen as precursor for both energy production and carbon assimilation. Therefore, growth on methanol is an aerobic process and oxygen limitation can lead to growth limitation.23 The underlying regulatory mechanisms of cellular networks permit a cell’s ability to adapt to available nutrient conditions by selecting the appropriate pathways. Thus, it is of importance to investigate changes in flux profiles during the change in substrate from glycerol to methanol. The change in flux distributions of P. pastoris during methanol adaptation (Figures 2 and 3) agrees well with other experimental studies. The molar fluxes through key enzymes in the methanol utilization pathway were found to be induced upon a shift to methanol culture.24 One previous study reported that the methanol oxidation pathway (R106r; R107; R108) is only activated at high levels of methanol,25 as the methylotrophic yeast uses this pathway for methanol detoxification under conditions of high methanol concentration; this is consistent with the model prediction in Figure 3. The increase in anaplerotic enzyme activities during methanol metabolism has also been reported.18 The predictive results by EMA were also supported by earlier transcriptome analysis and 13 C labeling experimental studies18,26,27 that showed a decrease in glycolytic fluxes (i.e., R8r) when the fraction of methanol in the feed was increased. Several glycolytic genes were also reported to be down-regulated upon adaptation to methanol metabolism. Also consistent with the experimental observations is the exchange flux of pyruvate between

Figure 6. Effect of flux partition on protein synthesis at pyruvate node under methanol induction phase, (A) for GFP and (B) for hGH. The pyruvate node represents flux ratio of glycolysis and citric acid cycle, defined as percentage of R94r per sum of R37, R38, and R94r. It was found that the flux ratio at this branch point has significant impact on the production of protein synthesis. Each circle symbol is a predicted yield of the identified pathway based on EMA. Yield is defined as the ratio of mmol in product to gram in the methanol substrate.

cytosol and mitochondria that was reported to be decreased with increasing methanol feed fraction. A gene expression study by Sauer et al.7 also reported the deregulation of several genes in the citric acid cycle upon the shift from glycerol to methanol, which is in agreement with the EMA prediction in Figure 3. Figure 4 shows up- and down-regulation of several key fluxes in the P. pastoris metabolic network during methanol adaptation. The EMA prediction of flux profiles is consistent with reported experimental data that showed the genetic and enzymatic level of these reactions—R5 (FBP); R10 (PYK); R17 (RPE); R18 (TKT); R19 (TKT); R104 (DAK); R105r (DAS)—were elevated upon a shift from glycerol to methanol substrate.7,18,27,28 The adaptation time of P. pastoris to methanol can take as long as 14 h, depending on the strain’s genetic background.29 Thus, overexpression of these genes based on the flux profiles of the EMA prediction could permit an improvement in methanol adaptation in P. pastoris by reducing the adaptation time. In addition, these genes could

36

Biotechnol. Prog., 2014, Vol. 30, No. 1

point to a different regulatory mechanism for controlling the two products. Interestingly, the reaction differences between protein synthesis and biomass synthesis occur around the pyruvate node. This implied that P. pastoris has an underlying mechanism to control the flux partitioning through this node during biomass and protein production. Thus, these reactions could be targeted for engineering to specifically select between production of biomass and production of recombinant proteins. Culturing conditions could also be used to enhance protein synthesis. The impact of the oxygen stress response on recombinant protein production in Figure 7 suggests that recombinant protein is best produced under hypoxic conditions. Oxygen limitation leads to growth limitation on methanol. Under this condition, a higher methanol concentration is maintained, which might further induce the AOX1 promoter, resulting in more protein being expressed.23,30 In the methanol pathway, the oxidation of methanol with the oxygen molecule is the first step of both energy production and carbon assimilation. Therefore, surplus of oxygen might lead to accumulation of formaldehyde, especially during the induction phase on methanol. The EMA results suggest that manipulation of the oxygen levels available to the culture could regulate which pathways are available to the cell, and permit the efficient production of recombinant proteins by carefully controlling oxygen at its optimal level.

Conclusion

Figure 7. Influence of oxygen levels on recombinant protein synthesis as predicted by EMA under methanol induction phase. Yield is defined as the ratio of mmol in product to that in the methanol substrate. Shade area presents performance correlation of the recombinant protein yield and the oxygen levels. The results suggest the production of recombinant proteins are desired under oxygen limitation culture condition. The optimum protein synthesis could be achieved by controlling oxygen levels at 1.78 mmol of oxygen levels per mmol of methanol used.

also be used as genetic switches to control the utilization of methanol substrate during recombinant protein synthesis in P. pastoris. The consistency between model prediction and experimental result confirms EMA as a powerful tool capable of predicting a regulatory system of the cellular metabolic network that permits the selection of optimal pathways under a specified growth condition. Additionally, EMA was used to study a recombinant P. pastoris system for the production of two recombinant proteins, GFP and hGH. The analysis revealed pathway possibilities and capacities of the system, based on reaction stoichiometry, compartmentalization, and thermodynamics constraints. For the considered network, EMA revealed different numbers of possible pathways for the different proteins (664 pathways for GFP and 675 pathways for hGH). This result suggested that the number of existing pathways available to the cell is strongly dependent on the amino acid composition of the expressed proteins. Differences in flux patterns for synthesis of biomass and recombinant proteins

EMA of the recombinant P. pastoris system has provided insight into the basic pathway structure of the organism for the production of heterologous proteins. EMA permits the evaluation of the metabolic capabilities that can be used to analyze and predict the cell’s performance under a wide variety of culturing conditions during recombinant protein synthesis. As case studies, we analyzed the identified pathways for the production of two recombinant proteins, GFP and hGH, in P. pastoris. Here, we demonstrated that the approach is well-suited for exploring the possible impacts of environmental stresses such as changes in feeding substrate or oxygen levels on the P. pastoris metabolism. The analysis allows understanding of the global mechanisms connecting these environmental conditions to growth and levels of expressed protein. EMA revealed that P. pastoris is capable of coping with culturing perturbations during the switch from glycerol to methanol by readjusting its metabolic flux profiles. The model was able to capture key characteristics of P. pastoris metabolism, including flux profiles during methanol adaptation and induction reported in experimental studies, thus validating the model. Knowing which fluxes/ enzymes are required for growth and recombinant protein synthesis during methanol metabolism, it is possible to develop genetic strategies for an efficient production of recombinant proteins from methanol. The metabolic flux profiles of P. pastoris to methanol identified by EMA could contribute to possible strategies for strain optimization to reduce the time required for the methanol adaptation process. Ideally, genetic switch targeting of the identified reactions could be developed to encourage metabolite fluxes towards certain reactions, favoring faster adaptation to a change in substrate conditions to methanol. Study of the environmental effects on protein synthesis could also serve as the basis for future research on rational optimization of strains to further enhance production of recombinant proteins during the

Biotechnol. Prog., 2014, Vol. 30, No. 1

induction process of methanol metabolism. Thus, EMA can serve as a valuable tool for metabolic studies that should be generally useful for biopharmaceutical industry for the production of therapeutic proteins by P. pastoris. The approach could be applied to study the metabolic activity for the production of other recombinant proteins, as well as whole-cell biotransformation by P. pastoris.

37

15. 16.

Literature Cited

17.

1. Gerngross T. Advances in the production of human therapeutic proteins in yeasts and filamentous fungi. Nat Biotechnol. 2004; 22:1409–1414. 2. Walsh G. Biopharmaceutical benchmarks. Nat Biotechnol. 2006; 24:769–776. 3. Idiris A, Tohda H, Kumagai H, Takegawa K. Engineering of protein secretion in yeast: strategies and impact on protein production. Appl Microbiol Biotechnol. 2010;86:403–417. 4. Daly R, Hearn M. Expression of heterologous proteins in Pichia pastoris: a useful experimental tool in protein engineering and production. J Mol Recognit. 2005;18:119–138. 5. De Schutter K, Lin Y, Tiels P, Van Hecke A, Glinka S, WeberLehmann J, Rouze P, Van de Peer Y, Callewaert N. Genome sequence of the recombinant protein production host Pichia pastoris. Nat Biotechnol. 2009;27:561–566. 6. Mattanovich D, Graf A, Stadlmann J, Dragosits M, Redl A, Maurer M, Kleinheinz M, Sauer M, Altmann F, Gasser B. Genome, secretome and glucose transport highlight unique features of the protein production host Pichia pastoris. Microb Cell Fact. 2009;8:29. 7. Sauer M, Branduardi P, Gasser B, Valli M, Maurer M, Porro D, Mattanovich D. Differential gene expression in recombinant Pichia pastoris analysed by heterologous DNA microarray hybridisation. Microb Cell Fact. 2004;3:1–17. 8. Mattanovich D, Gasser B, Hohenblum H, Sauer M. Stress in recombinant protein producing yeasts. J Biotechnol. 2004;113: 121–135. 9. Cos O, Ramon R, Montesinos JL, Valero F. Operational strategies, monitoring and control of heterologous protein production in the methylotrophic yeast Pichia pastoris under different promoters: a review. Microb Cell Fact. 2006;5:17. 10. Schuster S, Hilgetag C, Woods JH, Fell DA. Reaction routes in biochemical reaction systems: algebraic properties, validated calculation procedure and example from nucleotide metabolism. J Math Biol. 2002;45:153–181. 11. Schuster S, Fell DA, Dandekar T. A general definition of metabolic pathways useful for systematic organization and analysis of complex metabolic networks. Nat Biotechnol. 2000;18:326– 332. 12. Kr€omer JO, Wittmann C, Schr€oder H, Heinzle E. Metabolic pathway analysis for rational design of L-methionine production by Escherichia coli and Corynebacterium glutamicum. Metab Eng. 2006;8:353–369. 13. Stelling J, Klamt S, Bettenbrock K, Schuster S, Gilles E. Metabolic network structure determines key aspects of functionality and regulation. Nature. 2002;420:190–193. 14. Unrean P, Nguyen NH. Metabolic pathway analysis of Scheffersomyces (Pichia) stipitis: effect of oxygen availability on

18.

19. 20.

21.

22.

23.

24. 25. 26. 27.

28. 29. 30.

ethanol synthesis and flux distributions. Appl Microbiol Biotechnol. 2012;94:1387–1398. Varma A, Palsson BO. Metabolic capabilities of Escherichia coli: I. synthesis of biosynthetic precursors and cofactors. J Theor Biol. 1993;165:477–502. Carnicer M, Baumann K, T€oplitz I, Sanchez-Ferrando F, Mattanovich D, Ferrer P, Albiol J. Macromolecular and elemental composition analysis and extracellular metabolite balances of Pichia pastoris growing at different oxygen levels. Microb Cell Fact. 2009;8:65. Mathews CK, Holde KE, Ahern KG. Biochemistry. San Francisco: Addison Wesley Longman; 2000. Sola A, Jouhten P, Maaheimo H, Sanchez-Ferrando F, Szyperski T, Ferrer P. Metabolic flux profiling of Pichia pastoris grown on glycerol/methanol mixtures in chemostat cultures at low and high dilution rates. Microbiology. 2007;153:281–290. Trentmann O, Khatri NK, Hoffmann F. Reduced oxygen supply increases process stability and product yield with recombinant Pichia pastoris. Biotechnol Prog. 2004;20:1766–1775. Baumann K, Maurer M, Dragosits M, Cos O, Ferrer P, Mattanovich D. Hypoxic fed-batch cultivation of Pichia pastoris increases specific and volumetric productivity of recombinant proteins. Biotechnol Bioeng. 2008;100:177–183. Baumann K, Carnicer M, Dragosits M, Graf AB, Stadlmann J, Jouhten P, Maaheimo H, Gasser B, Albiol J, Mattanovich D, Ferrer P. A multi-level study of recombinant Pichia pastoris in different oxygen conditions. BMC Syst Biol. 2010;4:141. Zhang W, Hywood Potter K, Plantz B, Schlegel V, Smith L, Meagher M. Pichia pastoris fermentation with mixed-feeds of glycerol and methanol: growth kinetics and production improvement. J Ind Microbiol Biotechnol. 2003;30:201–215. Charoenrat T, Ketudat-Cairns M, Stendahl-Andersen H, Jahic M, Enfors SO. Oxygen-limited fed-batch process: an alternative control for Pichia pastoris recombinant protein processes. Bioprocess Biosyst Eng. 2005;27:399–406. Tuttle DL, Dunn WA Jr. Divergent modes of autophagy in the methylotrophic yeast Pichia pastoris. J Cell Sci. 1995;108:25–35. Jones JG, Bellion E. Methanol oxidation and assimilation in Hansenula polymorpha. An analysis by 13C N.M.R. in vivo. Biochem J. 1991;280:475–481. Sola A, Maaheimo H, Yl€onen K, Ferrer P, Szyperski T. Amino acid biosynthesis and metabolic flux profiling of Pichia pastoris. Eur J Biochem. 2004;271:2462–2470. Zutphen T, Baerends RJ, Susanna KA, de Jong A, Kuipers OP, Veenhuis M, van der Klei IJ. Adaptation of Hansenula polymorpha to methanol: a transcriptome analysis. BMC Genomics. 2010;11:1. Klei Van der IJ, Yurimoto H, Sakai Y, Veenhuis M. The significance of peroxisomes in methanol metabolism in methylotrophic yeast. Biochim Biophys Acta. 2006;1763:1453–1462. Dietzsch C, Spadiut O, Herwig C. A fast approach to determine a fed batch feeding profile for recombinant Pichia pastoris strains. Microb Cell Fact. 2011;10:85. Lee CY, Lee SJ, Jung KH, Katoh S, Lee EK. High dissolved oxygen tension enhances heterologous protein expression by recombinant Pichia pastoris. Process Biochem. 2003;38:1147–1154.

Manuscript received Mar. 28, 2013, and revision received Oct. 25, 2013.

Pathway analysis of Pichia pastoris to elucidate methanol metabolism and its regulation for production of recombinant proteins.

This research rationally analyzes metabolic pathways of Pichia pastoris to study the metabolic flux responses of this yeast under methanol metabolism...
683KB Sizes 0 Downloads 0 Views