Metabolic flux profiling of recombinant protein secreting Pichia pastoris growing on glucose:methanol mixtures

Growth and product formation of recombinant P. pastoris strains

Impact of methanol co-assimilation on the central carbon metabolism of P. Pastoris growing on glucose methanol mixtures

The 2D ¹H-¹³C-HSQC spectra were analysed as described by [28] and [29], yielding the relative abundances of intact C2 and C3 fragments in proteinogenic amino acids f-values (Additional file 2). Analysis of the f-values were coherent with the biosynthetic pathways of proteinogenic amino acids in yeast, as already shown in previous studies of P. pastoris[5]. The use of the C6 source glucose and C1 source methanol for BDF ¹³C-labelling of proteinogenic amino acids enabled the determination of the flux ratios for reactions associated with the assimilation of C1 source by the cell. When yeast are grown on glucose as a sole carbon source, the f-values of His-Cα and Phe-Cα must be equal or practically equal due to the effect of transaldolase and transketolase reactions [15, 30]. However, these two patterns were different in our experiments (Additional file 2), providing a direct evidence of assimilation of methanol for cell growth and maintenance, as well as proving that glucose limiting conditions allow for induction of the methanol assimilation pathways by the latter substrate. Similar evidence has also previously been observed in P. pastoris cells, growing in glycerol:methanol mixtures, under carbon-limiting conditions [6].

When comparing glucose vs glucose:methanol carbon flux distributions in P. pastoris cells growing aerobically in carbon-limited chemostat cultures, a clear impact of methanol assimilation is observed on the metabolic network operation (Figure 1). The split ratio between the glycolytic and the oxidative branch of the pentose phosphate pathway (PPP) fluxes was clearly shifted to the latter pathway in glucose:methanol grown cells, probably reflecting the demand of pentose phosphates for methanol assimilation. Also, cells growing in the mixed substrate presented a significantly lower flux through the tricarboxylic acids (TCA) cycle (normalized to the glucose uptake rate). This difference was also reflected in the calculated NADH regeneration rates (Figure 1). Interestingly, Solà and the co-workers [6] did not detect significant differences in the relative TCA cycle activity when comparing cells grown on glycerol vs glycerol:methanol mixtures. As expected from early mixed substrate studies of methylotrophic yeasts [31], methanol co-assimilation resulted in slightly lower biomass yield (14.3 ± 0.3 g CDW/ C-mol) compared with glucose-only grown cultures (16.8 ± 1.2 g CDW/ C-mol). Nevertheless, considering the standard deviations of these calculated values, further experimental data would be required to confirm this tendency. Most notably, most methanol (about 80%) was directly dissimilated to CO₂, as opposite to cells growing on methanol as sole carbon source under limiting conditions [16]. Methanol co-assimilation has an important effect on the bioenergetics of methylotrophic yeasts. For instance, direct oxidation of methanol to CO₂ via formaldehyde and formate (see Figure 1) yields two 2 NADH mols per methanol mol. Early studies on methanol metabolism showed that the enzymes for primary oxidation and assimilation of methanol or formaldehyde are under inhibitory control of energy equivalents, e.g. NADH and ATP (formaldehyde and formate dehydrogenases), ADP (dihydroxyacetone kinase) and AMP (fructose-l,6-bisphosphatase) [32]. In methylotrophic yeasts these metabolic pathways operate in accordance with the balance of consumed and produced energy equivalents, thus providing energetic regulation of formaldehyde oxidation and assimilation. Overall, the results seem to reflect the fact that the glucose-only cultivation is NADH-limited and the co-assimilation of methanol as auxiliary substrate may provide an extra direct source of NADH [33].

Impact of Rol secretion on the central carbon metabolism of P. pastoris

As previously observed [12], Rol overproduction had an impact on the substrate specific consumption rate. In addition, both Rol-expressing strains showed slightly lower but significant biomass yields, as well as higher CER and OUR values, compared to the reference strain. This phenomenon might be related to higher energy demand caused by Rol secretion, resulting in higher maintenance-energy requirements. Since Rol amounts were very small relative to the total cell protein, one is tempted to speculate that such metabolic burden was mainly associated to the secretion stress triggered by [Rol 23, 34], rather than to an increased demand for building blocks. The latter situation seems to occur when recombinant proteins are produced (intracellularly) at high levels in P. pastoris[17, 18] and other hosts.

Metabolic flux analysis allowed identification of limited but statistically significant changes in the fluxes through central carbon metabolism (Figure 2 and Additional file 3). Since glucose uptake rates were higher in the Rol-producing strains, glycolytic fluxes were concomitantly increased compared to the reference strain. In addition, the flux of pentose phosphates to the methanol assimilation pathway was significantly decreased in Rol-producing strains, consistent with lower fluxes of methanol being assimilated to glyceraldehyde 3-phosphate (GAP), particularly in the ROL 2-copy strain (Figure 3). Notably, although the biomass yield of the Rol-producing strains was somewhat lower compared to the reference strain (also reflected in lower flux to biomass synthesis, Figure 2), the flux through the oxidative branch of the PPP appeared to be similar in all strains. Since one would expect a correlation between the biomass yield and the activity of PPP [35], such effect could be the result of an increased demand of reduction equivalents (NADPH) to regenerate reduced glutathione (GSH) in the endoplasmatic reticulum (ER), the electron donor in the protein folding oxidative process [36, 37]. Interestingly, direct methanol dissimilation to CO₂ also involves an oxidative step using GSH as electron donor.

Further evidence for a metabolic burden derived from Rol expression was indicated by a significantly increased relative flux through the TCA cycle (normalized to the glucose uptake rate) in the Rol-producing strains. In addition, a tendency to increase the flux through the methanol dissimilatory pathway to CO₂ was observed in the Rol-producing strains compared to the reference strain, also reflected in a slight increase in the split ratio between this pathway and the assimilatory pathway to Pep; however, such tendency was not statistically significant. Coherent with these two observations, the respiration rate increased (see CER, OUR and respiratory quotient (RQ) values in Table 1), reflecting a significant increase in NADH regeneration reactions and, probably, in ATP generation (Figure 2). This would suggest that the metabolic burden caused by Rol secretion is at least partially compensated by increased energy production. That is, the Rol-producing strains appear to have higher maintenance requirements compared to the control strain. Since Rol production levels are moderate, the correlation observed between the specific Rol secretion rates and NADH regeneration rates provide indirect evidence on the metabolic burden associated with protein folding and conformational stress. Interestingly, Heyland and co-workers [17] observed a similar effect (increased TCA cycle flux of 1.1 mmol g CDW h⁻¹ in the producing strain compared to 0.7 mmol g CDW h⁻¹ in the reference strain) in P. pastoris cells producing an intracellular recombinant protein growing on glucose in a fed-batch culture at a controlled growth rate of 0.12 h⁻¹. However, in a recent study by the same group [18], where a series of P. pastoris strains expressing different levels of a model recombinant protein exponentially growing in shake cultures (that is, at maximum specific growth rate) were compared, revealed that although there was a relative increase in the relative TCA cycle activity in the producing strains compared to the reference strain, the absolute TCA cycle activity remained constant around 2.1 ± 0.1 mmol g CDW h⁻¹ in all strains, suggesting an upper limit of TCA cycle activity and, thereby postulating that cells do not have the capacity to catabolize a sufficient amount of carbon through the TCA cycle to fully compensate the higher energy demand derived from recombinant protein overproduction. In the present case, the value of the TCA cycle activity was 0.39 ± 0.03 mmol gCDW h⁻¹, 0.55 ± 0.03 mmol gCDW h⁻¹, and 0.57 ± 0.04 mmol gCDW h⁻¹ for the control strain, single and 2-copy Rol-producing strains, respectively, clearly below the reported hypothetical activity upper TCA cycle limit for P. pastoris growing aerobically on glucose [18]. Furthermore, since there is a significant increase in the absolute flux through the TCA cycle in Rol-producing strains, it seems plausible that there is no energetic limitation at this level in cells growing on glucose:methanol mixtures. In fact, co-assimilation of methanol as auxiliary substrate might be a mechanism by which the increased energy demand of the producing strains is compensated.

Strains and media

A series of recombinant P. pastoris X-33 (Invitrogen) derived strains were used in this study. Namely, a control strain harbouring pGAPAα (Invitrogen) as mock plasmid [26], and two different Rol producing strains: i) the X-33/pPICZAα-ROL strain, previously obtained by [24], regarded as a strain containing a single copy of the ROL expression vector integrated at the host’s AOX1 genomic locus and, ii) a strain newly generated in this study following a second transformation of strain X-33 with pPICZAα-ROL using a electroporation procedure described by [38]. Prior to transformation, plasmid DNA was linearized to promote integration at the AOX1 locus. Transformants were selected on YPD agar plates containing 100 mg L⁻¹ zeocin (Invivogen) and were subsequently replica-plated onto selection YPD agar plates containing 1000 mg L⁻¹ zeocin, as a strategy to select transformants containing multiple copies of the Rol expression vector integrated in their genome [12, 38, 39]. To select a multi-copy strain with higher expression levels, 10 independent transformants from agar plates containing 1000 mg L⁻¹ zeocin were tested for extracellular lipolytic activity in 500 mL Erlenmeyers following standard procedures described in the Pichia Expression Kit Manual (Invitrogen). The best clone, was further selected for chemostat studies. Both single and multicopy Rol-producing strains were further characterised in terms of ROL gene dosage by quantitative real-time PCR.

qRT-PCR assay

Quantitative real-time PCR was carried out in 20 μL volume reactions using semi-skirted iQ 96-well PCR plates and SsoFast^TM EvaGreen® Supermix (both from Bio-Rad). Samples were measured in triplicate and standards were measured in duplicate on the iCycler Thermal Cycler (Bio-Rad). A non template control was run in every experiment for each of the primer pairs to avoid detection of unspecific priming. The reactions were incubated at 95°C for 5 min to activate the Taq polymerase, and then subjected to a three-step cycling protocol including melting (94°C, 15 s), annealing (58°C, 15 s) and extension (72°C, 30 s) for a total of 40 cycles. Each extension was followed by data collection at 72°C and a short incubation step at 78°C (1 s) for a second plate reading closer to the melting point. Following a final extension of 5 min at 72°C, we generated a melting-curve profile collecting data along 70 cycles with variable temperature starting at 60°C, with 0.5°C increments/cycle (1-s intervals). The primers used for the amplification reaction were 5′ CCCTGTCGTCCAAGAACAAC 3′ and 5′ GAGGACCACCAACAGTGAAG 3′ (forward and reverse primers, respectively) for the ROL gene; for the reference amplification reaction of the β-actin gene (ACT1), primers were the same as described previously in [15]. The relative gene expression level was calculated for each sample in triplicate measurements giving a maximum standard deviation around 10%. Since the amplification efficiencies of the target and reference genes were not the same in our experiments, we used the Pfaffl method [40] for the relative quantification of our qRT-PCR results.

Chemostat cultivations

Two duplicate aerobic, carbon-limited continuous cultures for each of the three strains were carried out in a 3 L vessel bioreactor (Applikon Biotechnology) that was controlled at 25°C. The working volume was kept at 1 L by means of an overflow system. The pH was controlled at 5.0 with 1 M NH₃. An aeration rate of 1 vvm, controlled by the bioreactor’s mass flow meters, and a stirring rate of 800 rpm allowed maintaining dissolved oxygen levels at a minimum of 15% of air saturation. An overpressure of 0.2 bar was applied to the system to facilitate sampling of broth. The chemostat cultures were set at a D of 0.09 h⁻¹ by feeding a defined growth medium [41] containing 50 g L⁻¹ of glucose/methanol mixture (80% glucose / 20% methanol, w/w) as a carbon source. The bioreactor off-gas was cooled in a condenser (4°C), dried by means of two silica gel columns and subsequently analyzed with BCP-CO₂ and BCP-O₂ sensors (Blue-Sens). Sensors were calibrated using a series of 3 calibration gases containing CO₂/O₂/N₂ mixtures in the following percentages, respectively: 1/20.9/78.042; 3/5/91.97; 7/ 0/93. Steady state samples were taken after the cultures had been in constant conditions for a minimum of five residence times. Steady states were assessed over 4 to 6.5 residence times for constant biomass production CER, OUR, and detectable extracellular metabolites.

Analytical procedures

Biomass analyses. The cell concentration was monitored by measuring the optical density of cultures at 600 nm (OD₆₀₀). For cell dry weight (CDW) measurement, 5 mL of culture broth was filtered using pre-weighed dried glass fiber filters (Millipore). Cells were washed twice using the same volume of distilled water and dried overnight at 100°C. Triplicate samples (5 mL) were taken for all optical density and cell dry weight measurements. Biomass samples for the determination of the elemental composition, as well as amino acid, total protein and carbohydrate contents were prepared and analyzed as described by [26]. The measured amino acid content of the biomass allowed on the one hand, estimating the total protein content of the biomass and, on the other hand, calculating a specific protein composition to be used in a synthesis equation for the metabolic flux calculations. The experimentally measured elemental components (C, H, N, S and ash content). Oxygen content was calculated by difference as the remaining component. Major macromolecular biomass components (proteins and carbohydrates) were reconciled as previously described [27]. DNA, RNA and lipid content considered in this data consistency analysis were taken from previous measurements [26]. The resulting balanced biomass macromolecular composition was subsequently used for ¹³ C-constrained metabolic flux analysis. In all chemostat cultivations, the C recovery data was above 92% before applying a data consistency and reconciliation step. The experimental data was verified using standard data consistency and reconciliation procedures [42–44], under the constraint that the elemental conservation relations were satisfied. For all chemostat cultivations performed, the statistical consistency test carried out with a confidence level of 95% was acceptable, and consequently accepting that there was no proof for gross measurement errors.

Quantification of extracellular metabolites. Triplicate samples (5 mL) for extracellular metabolite analyses were centrifuged at 6,000 rpm for 3 min in a micro centrifuge (Minispin, Eppendorf) to remove the cells, and subsequently filtered through 0.45 mm-filters (Millipore type HAWP). Glucose, methanol, and other potential extracellular compounds were analyzed by HPLC (Dionex Ultimate 3000) analysis using an ionic exchange column, (ICSep ICE-COREGEL 87 H3, Transgenomic). The mobile phase was 6 mM sulphuric acid. The injection volume was 20 μL and the chromatogram was quantified with the CROMELEON software (Dionex).

Phosphoenolpyruvate carboxykinase (PepCK) assay. The activity of PepCK was assayed following a method previously described by [45], using the soluble fraction of cell crude extracts obtained after mechanical disruption using glass beads, as described by [46]. Briefly, cell extracts were prepared by mechanical disruption of 15–20 mg of lyophilized cells resuspended in 0.5 mL of extraction buffer (20 mM Hepes, pH 7.1, 1 mM DTT, 100 mM KCl), Complete protease inhibitor cocktail (Roche) and 1 g of glass beads. Cell suspensions were subsequently vortexed for 6 periods of 30 s, with a 30 s interval in ice between each vortexing cycle. Samples were centrifuged at 3000 ×g for 5 min at 4°C and the supernatant was subsequently transferred into an Eppendorf tube. A final centrifugation step was carried out in Eppendorf tubes at 5000 ×g for 15 min and 4°C to ensure that the final supernatant was totally clear. The resulting supernatant was used as cell-free extract. PepCK activity was determined following the 340 nm absorbance of the reduced pyridine nucleotide cofactor (E_340nm = 6.22 mM⁻¹). The reaction mixture (1 mL) contained 100 μmol Imidazole-HCl buffer, pH 6.6, 50 μmol NaHCO₃, 2 μmol MnCl₂, 2 μmol reduced glutathione, 2.5 μmol ADP, 0.15 μmol NADH, 3 U of malate dehydrogenase (Roche), and cell-free extract, 10~50 μL. The reaction was started by adding 2.5 μmol Pep. Enzyme activity was measured in a spectrophotometer (Varian Cary 300) at 30°C and 340 nm [45]. One unit activity is defined as the amount of enzyme that catalyzes the formation of 1 μmol of reduced pyridine nucleotide per min.

Lipase activity assay. The lipolytic activity was performed as previously described in [47].

Biosynthetically directed fractional (BDF) ¹³ C-labelling

P. pastoris cells were fed with a minimal medium containing 50 g L⁻¹ of a glucose:methanol mixture (80% glucose / 20% methanol, w/w) for five bioreactor volume changes until reaching a metabolic steady state, as indicated by a constant cell density in the bioreactor and constant O₂ and CO₂ concentrations in the exhaust gas. The ¹³C-labelling experiments were performed in two replicate cultures for each strain.

BDF ¹³C labelling of cells growing at steady state on a mix of two carbon source has been described elsewhere [6]. Briefly, as two carbon sources (namely, glucose and methanol) were used, the BDF ¹³C labelling step involved feeding the reactor with the medium containing about 12% (w/w) of uniformly ¹³C-labelled and 88% unlabelled amounts of each substrate simultaneously fed for 1.5 volume changes. [U-¹³C] glucose (isotopic enrichment 99%) and ¹³C-methanol (isotopic enrichment 99%) were purchased from Cortecnet (Voisins le Bretonneux, France). The labelled substrates were fed for a period of 1.5 residence times, after which, a volume of about 500 mL of culture broth was harvested, centrifuged at 4000 ×g for 10 min, resuspended in 20 mM Tris·HCl, pH 7.6, and centrifuged again. The recovered and washed cell pellets were freeze dried (Benchtop 5 L Vitris Sentry, Virtis Co., Gardiner, NY, USA). Finally, 100 mg of the freeze dried cell pellets were suspended into 10 mL of 6 M HCl and the biomass was hydrolysed in sealed glass tubes at 110°C for 22 h. The suspensions were dried overnight in an oven at 90°C, dissolved in H₂O and filtered through 0.2 μm filters (Millipore). The filtrates were vacuum-dried and dissolved in D₂O for NMR experiments. The final pH of the samples was below 1 due to residual HCl.

NMR spectroscopy

¹H-¹³C-HSQC nuclear magnetic resonance (NMR) spectra of the samples were acquired at 40°C on a Varian Inova spectrometer operating at a ¹H-resonance frequency of 600 MHz essentially as described in [28]. For each sample two spectra focusing on the aliphatic and aromatic regions were acquired as previously reported in [28]. The spectra were processed using the standard Varian spectrometer software VNMR (version 6.1, C).

Metabolic flux ratio (METAFoR) analysis

As described previously [5, 28–30], the calculation of metabolic flux ratios when using fractional ¹³C-labeling of amino acids is based on assuming both a metabolic (see above) and an isotopomeric steady state. As stated above, to establish a cost-effective protocol for a larger number of ¹³C labelling experiments, we fed a chemostat operating in metabolic steady state for the duration of 1.5 volume changes with the medium containing the ¹³C-labelled substrates before harvesting the biomass. Then, the fraction of unlabeled biomass produced prior to the start of the supply with ¹³C-labelled medium can be calculated following simple wash-out kinetics [5].

¹³C-metabolic flux analysis

¹³C-constrained metabolic flux analysis (¹³C-MFA) was performed using a stoichiometric model comprising the major pathways of P. pastoris central carbon metabolism. To calculate the intracellular net fluxes, the model was constrained with extracellular flux parameters (evolution rates of biomass, methanol and glucose uptake rate, CO₂ uptake rate) and 3 intracellular ratios derived from the METAFoR analysis (see Table 5), as described by [50], thereby constituting a determined system. Therefore, redox cofactors were not used as mass balance constraints to solve the ¹³C-MFA system. Cofactor mass balances are potential sources of errors since the correct balancing requires detailed knowledge of the relative activities of different isoenzymes and the enzyme cofactor specificities on a cell wide scale. Error minimization for the flux calculations in the determined network was carried out as described by [15]. The stoichiometric model of central carbon metabolism of P. pastoris was formulated following the model utilized by [15], complemented with the methanol assimilation pathways (Additional file 4). Glyoxylate cycle and malic enzyme reaction were omitted from the model on the grounds of the inspection of the METAFoR analysis, as previously described [48]. In this model, the consumption of central metabolic pathway’s intermediate metabolites for formation of the major biomass macromolecular components (proteins, carbohydrates, lipids and nucleic acids), was calculated as previously described [26] and considering P. pastoris biosynthetic pathways [5, 6, 51, 52]. The metabolic fluxes were considered as net fluxes so that a net flux in the forward direction was assigned a positive value and a net flux in the reverse direction was assigned a negative value.

Calculation of NADH regeneration rates

The rate of NADH regeneration was derived from the determined fluxes. Once a solution of the metabolic system was found, the metabolic fluxes were used to perform a theoretical calculation of the oxygen consumed. For this purpose, all major steps involved in oxygen consumption were taken into account (essentially, methanol and lipid biosynthesis pathways, as well as all relevant electron balances). Furthermore, it was assumed that all NADPH generated was consumed in biosynthetic reactions. Therefore, all the remaining reduction equivalents were assumed to be recycled through the respiratory chain as any other relevant possibility for recycling has already been taken into account. This allowed calculating the theoretical oxygen consumption rates. The theoretical oxygen consumption rates calculated represented 92% of the experimentally measured ones. Those results indicate that under the tested experimental conditions, the calculated variables are highly consistent with the experimental ones.

Statistical analyses

Data are given as mean ± SEM. Where appropriate, values were compared by a t-test, and significant differences were considered if above a 95% confidence level (p < 0.05).