Fina et al. Microbial Cell Factories (2023) 22:117 Microbial Cell Factories https://doi.org/10.1186/s12934-023-02123-0 RESEARCH Open Access High throughput 13C-metabolic flux analysis of 3-hydroxypropionic acid producing Pichia pastoris reveals limited availability of acetyl- CoA and ATP due to tight control of the glycolytic flux Albert Fina1, Pierre Millard2, Joan Albiol1, Pau Ferrer1* and Stephanie Heux2 Abstract Background Production of 3-hydroxypropionic acid (3-HP) through the malonyl-CoA pathway has yielded promising results in Pichia pastoris (Komagataella phaffii), demonstrating the potential of this cell factory to produce this platform chemical and other acetyl-CoA-derived products using glycerol as a carbon source. However, further metabolic engineering of the original P. pastoris 3-HP-producing strains resulted in unexpected outcomes, e.g., significantly lower product yield and/or growth rate. To gain an understanding on the metabolic constraints underlying these observations, the fluxome (metabolic flux phenotype) of ten 3-HP-producing P. pastoris strains has been characterized using a high throughput 13C-metabolic flux analysis platform. Such platform enabled the operation of an optimised workflow to obtain comprehensive maps of the carbon flux distribution in the central carbon metabolism in a parallel-automated manner, thereby accelerating the time-consuming strain characterization step in the design- build-test-learn cycle for metabolic engineering of P. pastoris. Results We generated detailed maps of the carbon fluxes in the central carbon metabolism of the 3-HP producing strain series, revealing the metabolic consequences of different metabolic engineering strategies aimed at improving NADPH regeneration, enhancing conversion of pyruvate into cytosolic acetyl-CoA, or eliminating by-product (arabitol) formation. Results indicate that the expression of the POS5 NADH kinase leads to a reduction in the fluxes of the pentose phosphate pathway reactions, whereas an increase in the pentose phosphate pathway fluxes was observed when the cytosolic acetyl-CoA synthesis pathway was overexpressed. Results also show that the tight control of the glycolytic flux hampers cell growth due to limited acetyl-CoA biosynthesis. When the cytosolic acetyl-CoA synthesis pathway was overexpressed, the cell growth increased, but the product yield decreased due to higher growth- associated ATP costs. Finally, the six most relevant strains were also cultured at pH 3.5 to assess the effect of a lower pH on their fluxome. Notably, similar metabolic fluxes were observed at pH 3.5 compared to the reference condition at pH 5. *Correspondence: Pau Ferrer Pau.Ferrer@uab.cat Full list of author information is available at the end of the article © The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Fina et al. Microbial Cell Factories (2023) 22:117 Page 2 of 16 Conclusions This study shows that existing fluoxomics workflows for high-throughput analyses of metabolic phenotypes can be adapted to investigate P. pastoris, providing valuable information on the impact of genetic manipulations on the metabolic phenotype of this yeast. Specifically, our results highlight the metabolic robustness of P. pastoris’s central carbon metabolism when genetic modifications are made to increase the availability of NADPH and cytosolic acetyl-CoA. Such knowledge can guide further metabolic engineering of these strains. Moreover, insights into the metabolic adaptation of P. pastoris to an acidic pH have also been obtained, showing the capability of the fluoxomics workflow to assess the metabolic impact of environmental changes. Keywords Fluxomics, 13C-Metabolic flux analysis, Pichia pastoris, Komagataella phaffii, High throughput, 3-hydroxypropionic acid, acetyl-CoA Background are further characterized at a bioreactor scale. However, The methylotrophic yeast Pichia pastoris (syn. Komaga- due to the development of high throughput (HT) biore- taella phaffii) has gained a lot of attention in the recent actor platforms, it is now possible to characterize a whole years due to its increased use in metabolic engineering set of genotypic variants at the fluxomic level providing applications [1–3]. Recent studies have shown the great the most complete and systematic description of the met- potential of P. pastoris to produce 3-HP [4, 5]. 3-HP is abolic state of a cell [15–17]. A strain’s fluxome is a result a bulk chemical with a large interest due to its multiple derived from the combination of its genome, transcrip- applications. It was listed among the top-value added tome, proteome, and metabolome, and the regulatory products to be obtained from biomass by the Department interactions between these components. Thus, fluxomics of Energy of the United States [6]. 3-HP can be converted can be used to identify regulatory mechanisms that may to acrylic acid, which is used to produce superabsorbent be hampering a strain’s performance [18, 19]. 13 plastics, as well as to other chemicals of interest, such as C-Metabolic Flux Analysis (13C-MFA) is the most malonic acid or 1,3-propanediol [6, 7]. widespread technique for the quantification of the fluxes 3-HP production in P. pastoris has recently been [20]. It consists in using a 13C-labelled substrate as tracer achieved by heterologously expressing the bi-functional to latter infer the metabolic reaction rates. HT analysis enzyme malonyl-CoA reductase (MCR) from Chloro- of multiple strains has focused on stationary 13C-MFA flexus aurantiacus [4]. Further strain optimization based (i.e., cells are collected at a metabolic and isotopic steady on rational strain engineering resulted in somewhat lim- state) because the sampling protocol is easier to autom- ited improvement (ca. 50% increase) of product yield [5] atize compared to instationary 13C-MFA, as there is a compared to similar strategies to increase the production remarkably lower number of samples to process and of 3-HP in other yeasts [8, 9]. So far, the highest 3-HP analyse [16]. The use of 13C-MFA for the characterization production reported in P. pastoris is 37.1 g L− 1 of 3-HP at of P. pastoris fluxes has been largely reported using glu- a volumetric productivity of 0.71 g L− 1 h− 1 in a fed-batch cose, glycerol, or combinations of thereof with methanol culture using glycerol as a carbon source [5]. [21–23]. Nonetheless, the exploitation of HT 13C-MFA One of the advantages of using P. pastoris to produce tools and methodologies for the characterisation of P. 3-HP is its ability to grow at a low pH. Performing the pastoris metabolism remains largely unexplored. In this cultures at an acidic pH allows the organic acids extrac- study, we have applied a workflow that allows the char- tion from the fermentation media using non-toxic sol- acterisation of the fluxome of multiple strains in parallel. vents. This process is simpler, and it generates less waste Such workflow integrates all relevant steps from experi- products than classical downstream processes [10, 11]. mental design to data acquisition and processing, and Moreover, in situ product recovery systems can be imple- flux calculation (Fig. 1), providing detailed supporting mented, avoiding reaching toxic levels of 3-HP [12]. information to facilitate its transferability to other inves- However, while the cultivation of P. pastoris at pH as low tigations. Notably, the workflow includes the generation as 3 has been widely studied as a strategy to minimise of a genome-scale reduced metabolic model applying a the activity of some endogenous proteases [13, 14], there compression protocol to a previously described Genome are no previous studies on the impact of low pH on the Scale Metabolic (GSM) model of P. pastoris [23, 24]. The fluxome of P. pastoris. aim of this study was to understand the impact of differ- Recent advances in the field of synthetic biology allow ent genetic modifications into the fluxome of a set of P. the generation of a high number of recombinant micro- pastoris strains that produce 3-HP using glycerol as a sole bial strains in a short amount of time within a single carbon source. The fluxomics analyses provided mean- metabolic engineering project. Typically, the strains are ingful insights of the bottlenecks of this yeast’s metabo- tested in small scale cultivation systems such as microti- lism towards 3-HP production. Moreover, the fluxome of ter plates or shake flasks, and the best performing ones the 3-HP producing strains was also characterized at pH Fina et al. Microbial Cell Factories (2023) 22:117 Page 3 of 16 Table 1 Strains used in this study. The genotype indicates the overexpressed genes with the corresponding promoter, and the deleted genes Strain name Genotype Reference X-33 Invitrogen-Thermo Fisher Scientific PpHP1 pGAP_mcrCa [4] PpHP2 pGAP_mcr-NCa pGAP_mcr-CCa [4] PpHP5 pGAP_mcr-NCa pGAP_mcr-CCa [4] pGAP_cPOS5Sc (Genbank’s [25] accession number ON52858*) PpHP6 pGAP_mcr-NCa pGAP_mcr-CCa [4] pGAP_ACC1Yl pGAP_cPOS5Sc PpHP8 pGAP_mcr-NCa pGAP_mcr-CCa [5] pGAP_mcr-CCa pGAP_ACC1Yl pGAP_cPOS5Sc PpHP13 pGAP_mcr-NCa pGAP_mcr-CCa [5] pGAP_mcr-CCa pGAP_ACC1Yl pGAP_cPOS5Sc pTEF1_acsSeL641P PpHP15 pGAP_mcr-NCa pGAP_mcr-CCa [5] pGAP_mcr-CCa pGAP_ACC1Yl pGAP_cPOS5Sc ∆ArDH PpHP17 pGAP_mcr-NCa pGAP_mcr-CCa [5] pGAP_mcr-CCa pGAP_ACC1Yl pGAP_cPOS5Sc pTEF1_acsSeL641P ∆ArDH PpHP18 pGAP_mcr-NCa pGAP_mcr-CCa [5] pGAP_mcr-CCa pGAP_ACC1Yl pGAP_cPOS5Sc pTEF1_acsSeL641P ∆ArDH pPDC1_PDC1 *Available from: https://www.ncbi.nlm.nih.gov/nuccore/ON528582. Model generation The genome scale model iMT1026v3 [24, 26] was com- pressed using CellNetAnalyzer 2021.1 [27] in Matlab R2020b (Matlab inc., Mathworks, MA, USA). A con- textualized core model was generated conserving the reactions, metabolites, and phenotypes that had been previously described for P. pastoris grown on glycerol as sole carbon source [23]. Additionally, D-arabitol secre- tion was conserved, as it has been previously reported Fig. 1 Diagram illustrating the complete workflow for conducting high that theses strains produce D-arabitol as by-product in throughput metabolic flux profiling on a set of Pichia pastoris strains. The both batch and fed-batch cultures [4]. Cytosolic acetate diagram encompasses all essential steps, starting from model generation production was also kept during model compression, as and C-source labelling optimization to metabolic flux calculation cytosolic acetate is used as substrate for 3-HP produc- tion. The models and scripts to compress iMT1026v3 3.5, a relevant condition for industrial production of car- model into a core model can be found in the Supplemen- boxylic acids. tary File S1. The newly generated core model was manually curated Materials and methods considering relevant literature-based knowledge. First, Strains production of isoleucine consumes mitochondrial pyru- The P. pastoris strains used in this study are listed in vate instead of 2-(α-hydroxyethyl)thiamine diphosphate. Table 1. These were all derived from the parental strain The production of lysine uses mitochondrial acetyl-CoA X-33 (Invitrogen-Thermo Fisher Scientific, MA, USA), and α-ketoglutarate from the cytosol, instead of vice which was used as a reference strain for this study. The versa. Finally, alanine production is derived from mito- construction of all strains has been previously described chondrial pyruvate, instead of cytosolic pyruvate [28– elsewhere [4, 5]. 30]. In addition, as biosynthesis of glutamate, glutamine, Fina et al. Microbial Cell Factories (2023) 22:117 Page 4 of 16 aspartate, and asparagine may take place in both the Media, cultivation conditions, and automated sampling cytoplasm and the mitochondria, both pathways were Seed cultures for the mini-bioreactor cultivations were included. Finally, the non-oxidative branch of the pen- prepared as follows: 50 mL falcon tubes containing 5 mL tose phosphate pathway was described using the half- YPG (1% w/v yeast extract, 2% w/v peptone, and 1% v/v reactions model, which considers the kinetic mechanism glycerol) medium were inoculated from cryostock and of the transketolase and transaldolase reactions [31]. The incubated overnight at 30oC and 200 rpm. The over- 3-HP production pathway from cytosolic acetate was also night cultures were transferred to a 250 mL shake flask included. containing 25 mL YPG at a starting OD600 of 1 and sub- All the necessary files to generate the new core model sequently grown for 8 h at 30oC and 200 rpm. The cul- (PpaCore_3HP.mat) from the GSM model can be found tures were harvested at the exponential phase and used in the Supplementary File S2. The final core stoichiomet- to inoculate the mini-bioreactor cultures. ric model contains 151 species and 145 reactions. A HT fluxomics platform consisting of 48 mini bio- For 13C-flux calculations, some of the amino acid bio- reactors was used for the cultivation and the sampling synthetic reactions were manually lumped to reduce the [36]. Briefly, the 50-mL bioreactors of this platform are number of overall reactions and metabolites. Some of the equipped with pH and dissolved oxygen sensors for pro- intermediary metabolites of the lumped reactions were cess monitoring and control. The bioreactors are agitated present in the biomass production reactions. The stoi- using a magnetic stirrer bar and they are designed to be chiometry of the biomass formation equation was cor- aerated through the headspace of each reactor to avoid rected to remove these metabolites. These modifications the accumulation of CO2. can be found in the Supplementary File S3. This model The batch medium contained 1.8 g L− 1 citric acid, was converted into FTBL (Flux TaBuLar) format [20, 32]. 0.02 g L− 1 CaCl2 · 2 H2O, 12.6 g L− 1 (NH4)2HPO4, 0.5 g The reactions producing 3-HP from cytosolic acetate L− 1 MgSO4 · 7 H2O, 0.9 g L− 1 KCl, 0.4 mg L− 1 biotin, (the closest intermediary metabolite to malonyl-CoA and 4.6 ml L− 1 of PTM1 trace salts [37]. The pH of the in the core model) were introduced, and carbon atom medium was adjusted to 5 or 3.5 using HCl. The batch transitions were added for all reactions to simulate label medium was supplemented with 0.5 g L− 1 glycerol propagation. The final core isotopic model for the P. pas- labelled at position 1 and 2 g L− 1 glycerol labelled at toris 3-HP producing strains, which contains 113 species position 2 (20% 1-13C-glycerol and 80% 2-13C-glycerol). and 123 reactions, can be found in the Supplementary Labelled substrates were obtained from Innovachem SAS File S4 (PpaCore_3HP.ftbl). The FTBL model can also (France). All the components of the medium except for be obtained from BIOMODELS [33], accession number: the trace salts, the biotin, and the labelled glycerol were MODEL2210090004. mixed and autoclaved. The other components were filter- sterilized using a syringe filter with 0.2 μm pore size and In silico design of 13C-labelling experiments introduced into the batch medium under sterile condi- The context-specific core model of P. pastoris metabolism tions. Each bioreactor was filled with 15 mL of medium. including the 3-HP formation reactions (PpaCore_3HP. All the strains listed in Table 1 were tested in tripli- ftbl) was used to calculate the optimal isotopic composi- cate in bioreactor cultures at pH 5. The cultures were tion of the substrate using IsoDesign v1.2.1 [34]. IsoDesign inoculated at a starting OD600 of 0.025. The strains X-33, uses influx_si [35] to calculate the precision of the fluxes PpHP1, PpHP6, PpHP8, PpHP15, and PpHP18 were for each (mixture of ) label input to be tested. tested in triplicate at pH 3.5. The cultures were inocu- To determine the optimal label input, the 13C-glycerol lated at a starting OD600 of 0.05. The temperature was combinations yielding the highest number of fluxes set to 28oC. The stirring rate was set to 2200 rpm. The air with a SD < 1 was calculated. The commercially avail- flow into the sterile gas cover, which ensures identical gas able 13C-labelled glycerol variants include 1-13C-glycerol, distribution into the head spaces of all the 48 bioreactors, 2-13C-glycerol, 1,3-13C-glycerol, and U-13C-glycerol. The was controlled to 104 mL min− 1 per bioreactor by a mass substrate combinations were ranked based on the sum of flow controller (Aalborg Instruments, NY, USA). SDs of the reactions with a SD < 1. Afterwards, the sub- Automated sampling was programmed to measure the strates providing the highest precision of the fluxes of the OD600 off-line every two hours for the first 16 h of culti- upper glycolysis (UG) and the pentose phosphate path- vation. After 16 h of cultivation, sampling of the OD600 way (PPP) were ranked. The IsoDesign results indicated was performed every hour (during the exponential phase that the best substrate combination consisted of 20% of the cultures). From the 16 to the 24 h of cultivation, 1-13C-glycerol and 80% 2-13C-glycerol (Supplementary 250 µL samples were withdrawn hourly from each cul- File S5). ture for supernatant analysis. The samples were placed on 96-well plates with a 0.45 μm filter bottom and vac- uum filtered immediately. A final supernatant sample Fina et al. Microbial Cell Factories (2023) 22:117 Page 5 of 16 was withdrawn after 27 h cultivation. After 27 h cultiva- of 20 × 2.1 mm with particle size 5 μm (Supelco Belle- tion, OD600 samples were withdrawn every 2.5 h for 10 h. fonte, PA, USA) and a column Discovery HS F5 HPLC Sampling and OD600 data from the mini-bioreactor cul- column of 150 × 2.1 mm with particle size 5 μm (Supelco tures can be found on Supplementary File S6. When each Bellefonte). Solvent A consisted of 0.1% v/v formic acid culture reached an OD600 above 1, a 500 µL sample was in ultrapure water and solvent B of 0.1% v/v formic acid withdrawn from the culture. The samples were quenched in acetonitrile. The flow rate of the eluent was set to 0.25 on 3.5 mL quenching solution (40% v/v acetonitrile, 40% mL min− 1, and the temperature of the sampler and the v/v methanol, 20% water and 0.1 M formic acid) at -20oC. column were set to 4 and 30 °C, respectively. Solvent B Samples were stored at -20oC between 1 and 3 h until set points were varied as follows: 0 min: 2%; 2 min: 2%; further processing. 10 min: 5%; 16 min: 35%; 20 min: 100%; 24 min: 100%. Finally, the set points of the initial conditions (2% solvent Analyses B) were set for 6 min before the injection of the next sam- Exometabolites analysis with NMR ple. The injection volume was 5 µL. Glycerol, 3-HP, and D-arabitol were quantified from The MS detector was set to detect the positive ions the filtered supernatant samples using Nuclear Mag- on FTMS mode. The resolution was set to 70,000 (at netic Resonance (NMR). 180 µL of sample and 20 µL of m/z = 400), the capillary temperature to 320 °C, and the 10 mM TSP (3-(trimethylsilyl)-[2,2,3,3-2H4]-propionic source heater temperature to 300 °C. Sheath gas and aux- acid sodium salt) were mixed. TSP was used as internal iliary gas flow rates were set to 40 and 10 arbitrary units, standard and quantified using a commercial succinate respectively. The S-lens RF level was set to 40% and the solution at 1 g L− 1 (43,057, Sigma Aldrich, MO, USA). source voltage to 5 kV. The MS was set to measure the 1D-1 H analyses were carried out on a Bruker Advance exact mass of all carbon isotopologues of all the amino III 800 MHz spectrometer (Bruker BioSpin, Germany) acids (Supplementary File S8). equipped with a 5 mm CQPI cryoprobe. Samples were The isotopologue distribution of the amino acids was analysed at 280 K using a 30o pulse sequence with water corrected considering the isotopologue labelling of the suppression (zgpr30), with a relaxation delay of 7 s. Top- inoculum and the natural abundance of all isotopes using Spin v3.6.4 (Bruker BioSpin, Germany) was used for the IsoCor v2.2.0 (https://github.com/MetaSys-LISBP/Iso- analysis of the NMR spectra. Commercial succinate at 1 g Cor) [39]. Raw and processed data can be found in Sup- L− 1 (43,057, Sigma Aldrich, MO, USA) was used for the plementary file S9. re-quantification of TSP. All data was corrected based on the quantification of succinate. The peak integration data Bioprocess parameters from the NMR analyses can be found in Supplementary The OD600 and the NMR supernatant analyses results File S7. were used to calculate the growth rate (µ) and the spe- cific production or consumption rates (q-rates) for glyc- Sample processing and analysis of the 13C-proteinogenic erol, 3-HP, and D-arabitol. To this end, PhysioFit v1.0.2 amino acids using LC/MS (https://github.com/MetaSys-LISBP/PhysioFit) was used Quenched samples were vortexed for 30 s and centri- [40]. The absorbance at OD600 was correlated to the bio- fuged at 5,000 g for 5 min in a swing-rotor centrifuge. mass concentration using a constant conversion factor Pellets were dried using a Rotavapor (Büchi, Switzerland). for all the strains, the rationale being that all the strains The biomass was treated with 150 µL 6 N HCl at 110 °C used in this study shared the same conversion factor in for 16 h. The acid was evaporated in the rotavapor. Each the culture conditions used. This assumption is sup- sample was washed using 100 µL milliQ water and sub- ported by experimental data from our previous studies sequently dried in the rotavapor. This step was repeated [4, 5], where the conversion factor of the biomass absor- twice to eliminate all the acid traces. Finally, the pellets bance (OD600) to the cell dry weight (CDW) was statis- were resuspended on 200 µL milliQ water. The samples tically identical for strains PpHP6, PpHP8, and PpHP18 were centrifuged at 10,000 g for 5 min to eliminate the in batch cultures using the same growth medium. The biomass debris. 40 µL of supernatant (containing the conversion factor value was derived from calculating the amino acids) were diluted onto 460 µL milliQ water. µ (which is independent from the conversion factor) of The diluted amino acids were analyzed using a previ- the triplicate cultures of the reference strain (X-33) at ously described HPLC-MS method [38]. The UHPLC pH 5. Afterwards, the genome scale model iMT1026v3 Vanquish (Thermo Fisher Scientific, MA, USA) was was used to calculate the uptake rate of glycerol at such coupled to a mass spectrometry (MS) detector Orbitrap growth rate (0.21 h− 1) considering a non-growth associ- Q-Exactive plus with a heated electrospray ionization ated ATP consumption of 2.51 mmol gCDW−1 h− 1 [24], source (Thermo Fisher Scientific). HPLC-MS was used which resulted 3.4 mmolGlyc gCDW−1 h− 1. This calculation with a precolumn Discovery HS F5 Supelguard Cartridge was done in Matlab using the CobraToolbox v2.26.0 [41]. Fina et al. Microbial Cell Factories (2023) 22:117 Page 6 of 16 Finally, the conversion factor that correlates the experi- fluxes through all producing and consuming reactions mental data and the computational results was calculated for each strain. The Matlab script, the input file with to be 0.563 gCDW/OD600. The biomass yield for strain the Monte Carlo results, and the FBA results file can be X-33 using this calculated conversion factor resulted in found in the Supplementary File S11. 0.73 gCDW gGlyc−1, which falls within the reported experi- mental values [24]. Results and discussion The biomass concentration data can be found together HT cultivation bioprocess parameters with the OD600 data in Supplementary File S6. The series of previously available 3-HP-producing P. pas- The bioprocess parameters results calculated using toris strains[4, 5] were grown in glycerol batch cultures at PhysioFit can be found in the Supplementary File S10. pH 5 using a HT 13C-fluxomics platform [36]. The stoi- chiometry of the engineered reactions and correspond- 13 C-Flux calculation ing enzymes in the 3-HP-producing P. pastoris strains Flux calculations were performed from metabolic and are depicted in Fig. 2. Strain engineering strategies were isotopic steady-state data using influx_si v5.3.0 (https:// aimed at increasing the delivery of the substrates of the influx-si.readthedocs.io, [35]) and the previously malonyl-CoA to 3-HP pathway (i.e., cytosolic acetyl- obtained FTBL format model. The bioprocess parameters CoA, malonyl-CoA, and NADPH), and the reduction of and the corrected data of the isotopologue distribution the production of the main by-product (D-arabitol). of the amino acids were used as input data. All fluxes Measurement of the extracellular metabolites and the could be determined accurately from the extracellular biomass concentration allowed calculating the biopro- fluxes and labelling data except for fluxes through the cess parameters (specific substrate consumption and MDHc, MDHm, Tr_AKGMALtm and Tr_AKGOAAtm (by)products production rates (q-rates) and µmax) for reactions. each strain (Fig. 3), which is a prerequisite for metabolic All fluxes were normalized to the rate of the substrate flux calculations. Results were consistent with previ- uptake rate. A sensitivity analysis was performed using ously reported cultivation data [4, 5]. In a nutshell, com- the Monte-Carlo method with 100 independent simu- parison between strains PpHP1 and PpHP2 reveals that lations. Statistical analysis (chi-square test) was used to dissection of MCR from C. aurantiacus into each of its determine sufficient goodness of the fit (based on a 95% two subunits resulted in increased specific malonyl- confidence level) for each experiment. The average fluxes CoA reductase activity [4, 46], hence leading to a 10-fold and associated standard deviations of the mean were increase in the 3-HP yield. The overexpression of the calculated from the biological replicates with sufficient genes encoding for an acetyl-CoA carboxylase from Yar- goodness of the fit. rowia lipolytica (ACC1Yl) and a cytosolic version of the Throughout this study, the reference guidelines for the mitochondrial NADH kinase from Saccharomyces cere- calculation of fluxes derived from 13C-data have been fol- visiae (cPOS5) in strain PpHP6 led to a further increase lowed [42]. Therefore, all the relevant data is adequately in 3-HP yield by enhancing the conversion of acetyl-CoA shared to the community for reproducibility and further into malonyl-CoA (3-HP’s metabolic precursor) and analyses. NADPH supply, respectively. In addition, the expression Flux distributions were visualized using Omix v2.0.7 of a second copy of the gene encoding the C-terminal (Omix Gmbh, Germany) [43]. The figures show the aver- domain of MCR yielded the highest 3-HP-producing age absolute or relative fluxes and the standard deviation strain (PpHP8), which was also the slowest growing strain of the triplicates. and the strain producing the largest amount of D-arabitol in batch cultures. As previously reported, the overexpres- Flux balance analysis sion of the genes encoding for the cytosolic acetyl-CoA Flux Balance Analysis (FBA) [44] was performed using synthesis pathway (i.e., acsSeL641P and PDC1, Fig. 2) in the the CobraToolbox v2.26.0 in Matlab R2020b and the pre- strain PpHP8 (yielding strain PpHP18), aimed to increase viously described PpaCore_3HP.mat model. The experi- the conversion of pyruvate to cytosolic acetyl-CoA, mental confidence intervals obtained from the Monte restored the growth rate, but the 3-HP yield dropped Carlo analyses were used as upper and lower bounds drastically [5]. These results pointed at a limitation of for each flux. Afterwards, FBA was performed using the resources in PpHP8 when growing at maximal growth maximization of the flux of the ATP of maintenance reac- rate [5], thereby resulting in an opposite trend between tion (ATPM) as objective function. This is the objective biomass and product yields. function resulting in the best description of the intracel- Results in Fig. 3 also show that not only the growth lular fluxes of a cell culture growing on a batch culture rate of strain PpHP8 was remarkably lower than the with excess of substrate [45]. FBA results were also used growth rate of the reference strain (0.13 h− 1 and 0.22 h− 1, to calculate the ATP and NADPH balance by summing respectively), but also the substrate uptake rate (2.10 and Fina et al. Microbial Cell Factories (2023) 22:117 Page 7 of 16 Fig. 2 Metabolic pathway from pyruvate to 3-HP through cytosolic acetyl-CoA and malonyl-CoA in P. pastoris. The reactions catalysed by all the enzymes linked to overexpressed or deleted genes are displayed. Enzyme abbreviations: Pdc1: Pyruvate decarboxylase 1; Ald: Endogenous cytosolic aldehyde de- hydrogenase; ACSSeL641P: Acetyl-CoA synthase from Salmonella enterica harbouring the point mutation L641P to avoid post-translational inhibition of the enzyme by acetylation; Acc1Yl: Acetyl-CoA carboxylase from Y. lipolytica; MCR-CCa: C-terminal domain of the malonyl-CoA reductase from C. aurantiacus; MCR-NCa: N-terminal domain of the malonyl-CoA reductase from C. aurantiacus; cPos5Sc: NADH kinase from S. cerevisiae located on the cytosol; ArDH: Arabitol dehydrogenase. Green and red arrows and enzyme abbreviations indicate whether the corresponding genes were overexpressed or deleted, respectively 3.37 mmol gCDW− 1 h− 1). In fact, the glycerol specific yields. Strikingly, deletion of the main D-arabitol dehy- uptake rate (qS) of PpHP8 was significantly lower than drogenase encoding gene (ArDH) in strains PpHP15 and the mean qS value observed for the rest of strains (based PpHP17 resulted in increased growth rates but reduced on a one-way ANOVA test, p = 0.02). A high activity of 3-HP production, compared to their corresponding the malonyl-CoA to 3-HP pathway probably led to a parental strains (PpHP8 and PpHP13, respectively). reduced availability of acetyl-CoA for this strain. Acetyl- CoA plays a central role on the biosynthesis of precur- Fluxome ofP. pastoris3-HP-producing strains at pH 5 sors, and it plays a key role in physiological regulation The bioprocess parameters of the series of strains processes, such as the acylation of histones [47]. Growth obtained from batch cultivations and the amino acids iso- defects have also been reported in S. cerevisiae strains topologue distributions derived from the corresponding harbouring a high acetyl-CoA carboxylase activity, where labelling experiments were used to calculate the intracel- depletion of acetyl-CoA was described as the most likely lular fluxes. The fluxome of each strain can be found in cause of hampered growth [48]. Moreover, the observed the Supplementary File S12 (see also Supplementary File increase in the growth rate and the uptake rate when S13 for a summary of the results). acsSeL641P was overexpressed support this hypothesis. The metabolic flux profile of the reference strain (X-33) Strikingly, the biomass yield for PpHP8-derived strains obtained using the HT robotic platform is comparable to overexpressing acsSeL641P (PpHP13, ppHP17), and PDC1 the previously reported results for the same strain grow- plus acsSeL641P (PpHP18) were lower than the biomass ing in glycerol chemostats on a similar medium [23]. yield of strain PpHP8. Noticeably, the fluxes of the upper glycolysis (UG) and Notably, all strains showed production of D-arabitol pentose phosphate pathway (PPP) in our batch experi- under the tested conditions. This by-product is typically ments (i.e., at µmax) were higher than the ones observed produced due to an imbalance on the NADPH regen- in chemostat cultures at lower growth rates (0.10 and eration [4, 5], resulting in reduced biomass and product 0.16 h− 1), coherent with the positive correlation between Fina et al. Microbial Cell Factories (2023) 22:117 Page 8 of 16 Fig. 3 Bioprocess parameters of the parental P. pastoris strain and nine 3-HP-producing strains cultivated in glycerol batch mini bioreactor cultures at pH 5. Orange bars show the biomass yield, grey bars show the 3-HP yield, green bars show the D-arabitol yield, blue diamond depict the µmax, and yellow circles show the substrate uptake rate (qS). Standard deviation of the replicates is depicted growth rate and the UG and PPP fluxes previously ratio controls the fluxes towards the oxidative branch observed in glycerol chemostats (Additional Figure S1). of the PPP [49]. In addition, overexpression of cPOS5 Similarly, fluxes of the lower glycolysis (LG) and the tri- in P. pastoris provides the cell with an additional source carboxylic acid (TCA) cycle reactions were lower than of NADPH, leading to a higher NADPH/NADP+ ratio the ones observed in chemostat cultures at lower growth [50]. Therefore, decrease of UG and PPP fluxes in PpHP5 rates, also coherent with the reported inverse correla- compared to PpHP2 was consistent with an increased tion between growth rate and LG and TCA cycle fluxes. NADPH/NADP+ ratio. Lower LG and TCA cycle fluxes are consistent with No major changes were observed when ACC1Yl was increased UG and PPP fluxes, as already reported [23]. overexpressed (i.e., in strain PpHP6, compared to strain To facilitate comparison of metabolic flux distributions PpHP5). For the strain PpHP8, which harboured an addi- amongst strains, a heat map illustrating the fold-change tional copy of the gene encoding for MCR-CCa, the high- between the relative fluxes (i.e., normalised to the spe- est fluxes towards 3-HP production were observed, while cific glycerol uptake rate) of each recombinant strain the UG and PPP fluxes were the lowest among all strains. compared to the relative fluxes of the reference strain Such observation agrees with the results for the strain cultivated at pH 5 is shown in Fig. 4. The most drastic PpHP5, as the increase in the NADPH requirements due changes were observed in the relative fluxes through the to production of 3-HP followed independent trends with UG and PPP reactions. First, when the MCR activity was the fluxes of the oxidative branch of the PPP. increased by expressing separately the two MCR domains Heterologous expression of acsSeL641P in strain PpHP8 (i.e., PpHP2 compared to PpHP1), the fluxes of the UG (i.e., generating strain PpHP13) led to a drastic switch in and PPP increased noticeably (10∼25%). Such trend can the strain’s fluxome. The relative fluxes through the UG be explained by increased NADPH requirements, as and PPP increased remarkably in PpHP13, while showing NADPH is used as the electron donor for the two consec- a lower relative flux towards 3-HP production compared utive reactions catalysed by MCR. When the gene encod- to PpHP8 or PpHP6 strains. Moreover, the biomass ing the heterologous cytosolic NADH kinase (cPOS5Sc) yield of PpHP13 was lower. Therefore, considering the was overexpressed (PpHP5), the observed fluxes of the NADPH requirements for biomass and 3-HP produc- UG and PPP reactions were 15∼30% lower compared tion, increased production of NADPH through the PPP to PpHP2. It is well known that the NADPH/NADP+ seems unfounded. Deletion of the gene encoding for Fina et al. Microbial Cell Factories (2023) 22:117 Page 9 of 16 Fig. 4 Reaction flux Log2 fold-change between each one of the ten 3-HP-producing recombinant strains compared to the reference strain (X-33). All co- lours are referred to the − 1 to + 1 colour scale, except the flux of the ‘MCR’ reaction, where the upper boundary of the colour scale is set to + 4. Moreover, in the case of ‘MCR’, the Log2 fold-change of the fluxes is referred to the flux of strain PpHP1. YX/S is the biomass to substrate yield. The reaction stoichiometry corresponding to each reaction abbreviation can be found in the Supplementary File S3 the NADPH-dependent arabitol dehydrogenase enzyme (resulting in PpHP18) led to the highest relative flux (ArDH) in strains PpHP8 and PpHP13 (i.e., obtaining through the UG and PPP. However, neither the biomass strains PpHP15 and PpHP17, respectively) led to minor yield nor the 3-HP yield were affected (Fig. 3). changes in the strains’ fluxomes under the tested growth Changes in the fluxes through LG and the TCA cycle conditions. Finally, overexpression of PDC1 in PpHP17 reactions followed the opposite trend to the UG and the Fina et al. Microbial Cell Factories (2023) 22:117 Page 10 of 16 PPP. Small differences in the fluxes through the glyoxylate Further comparison of the absolute flux distributions cycle were also observed. However, as the absolute values (i.e., non-normalised to the glycerol specific uptake rate) of these fluxes were low (below 0.025 mmol mmol− 1 h− 1), in the reference and the 3-HP-producing strains PpHP8 absolute changes of these fluxes did not have an impact and PpHP18 (Fig. 5) provided additional insights. on the strain’s biomass and product yields. Fig. 5 Flux map for the reference (X-33), PpHP8, and PpHP18 strains growing in glycerol batch cultures at pH 5. The average and the standard deviation of the absolute fluxes for each triplicate experiment are displayed. Fluxes are given in mmol gCDW− 1 h− 1. Abbreviations: G6P: Glucose-6-phosphate; F6P: Fructose-6-phosphate; FBP: Fructose bisphosphate; Glyc: Glycerol; DHAP: Dihydroxyacetone phosphate; G3P: Glyceraldehyde-3-phosphate; R5P: Ribose-5-phosphate; S7P: Sedoheptulose-7-phosphate; E4P: Erythrose-4-phosphate; ABT: D-arabitol; E2: glycolaldehyde moiety of the non-oxidative PPP reactions; E3: dihydroxyacetone moiety of the non-oxidative PPP reactions; 13DPG: 1,3-Bisphosphoglycerate; 3PG: 3-phosphoglycerate; 2PG: 2-phospho- glycerate; PEP: Phosphoenolpyruvate; Pyr: Pyruvate; Acald: Acetaldehyde: Ac: Acetate; AcCoA: Acetyl-CoA; MalCoA: Malonyl-CoA; 3-HP: 3-Hydroxypropi- onic acid: CIT: Citrate; ICIT: Isocitrate; αKG: α-ketoglutarate; SUCC: Succinate; FUM: Fumarate; MAL: Malate; OAA: Oxaloacetate; GLX: Glyoxylate Fina et al. Microbial Cell Factories (2023) 22:117 Page 11 of 16 First, overexpression of the 3-HP production pathway Strain PpHP18 has remarkably higher UG and PPP led to a higher pyruvate decarboxylase flux at the pyru- fluxes than strain PpHP8 (Fig. 5). The UG and PPP have vate node in PpHP8 compared to the reference strain a low carbon and energy yield. Therefore, while PDC1 is (0.30 mmol gCDW−1 h− 1 and 0.12 mmol gCDW−1 h− 1, being overexpressed, the energy requirements in strain respectively). The overexpression of ACSSeL641P and Pdc1 PpHP18 sinked the pyruvate into the TCA cycle for in strain PpHP8 did not result in a higher flux towards ATP generation, hampering the flux toward cytosolic cytosolic acetyl-CoA in strain PpHP18, as the absolute acetyl-CoA and, ultimately, reducing the 3-HP yield. On flux values for the two strains were identical (0.30 mmol the contrary, as PpHP8 grew at a lower rate, the energy gCDW−1 h− 1). Compared to the reference strain, overex- requirements of the strain were reduced, leaving more pression of the cytosolic acetyl-CoA production path- substrate available to produce 3-HP. way in strain PpHP18 did not increase the LG fluxes. All Overall, comparison of absolute flux distributions sug- these findings agree with previous results in S. cerevisiae, gests that to further increase the 3-HP yield in strain where the overexpression of PDC1 led to a higher flux PpHP8, the glycolytic fluxes would need to be signifi- towards this pathway without increasing the glycolytic cantly increased. Moreover, results also point to high flux [51]. It is well described that the glycolytic flux is ATP requirements in strain PpHP18 are the cause of the tightly controlled in yeast S. cerevisiae, and the glycolytic differences between these two strains. flux cannot be increased by overexpressing individual enzymes [52]. In the case of P. pastoris, increased glyco- ATP and NADPH producing/consuming fluxes of the lytic fluxes have only been described under low oxygen 3-HP-producing P. pastoris strains availability [53] or when a transcription factor control- The results from the previous section show that the ling the expression of all the glycolytic genes was over- fluxes through the PPP decreased when a heterologous expressed [54]. Therefore, as pyruvate is pulled into the cytosolic NADH kinase was expressed. Consistently, production of 3-HP, but the glycolytic flux and the uptake NADPH production rates calculated from the 13C-flux of glycerol do not increase, the overall ATP yield of the data (Fig. 6A) indicate that the NADPH produced by the 3-HP-producing strains decreases. glucose-6-phosphate dehydrogenase (G6PDH) reaction Fig. 6 A. Production and consumption rates of NADPH (estimated from the 13C-MFA results) and specific glycerol uptake rates for each strain. B. FBA results. The y-axis on the left side shows the sum of the fluxes for the main ATP-consuming reactions. The y-axis on the right side shows the percentage of pyruvate entering the mitochondria and the TCA cycle. To do so, the ratio of the fluxes of PYK (the main cytosolic pyruvate producing reaction) and PDH was calculated. Abbreviations: G6PDH, glucose-6-phosphate dehydrogenase; GAME, growth associated maintenance energy; NGAME, non-growth associated maintenance energy; UG/PPP: upper glycolysis and pentose phosphate pathway reactions; Acetyl-CoAcyt: Cytosolic acetyl-CoA production pathway; VPDH/PYK: Ratio between the fluxes of the pyruvate dehydrogenase (PDH) and the pyruvate kinase (PYK) reactions Fina et al. Microbial Cell Factories (2023) 22:117 Page 12 of 16 was lower than the actual NADPH requirements in some PpHP18, respectively) to compensate for the higher ATP of the strains overexpressing the cPOS5 gene (i.e., strains requirements. PpHP5, PpHP6, andPpHP8), supporting that the NADH Altogether, these results show the correlation between kinase reaction contributed to cover the cell’s NADPH acetyl-CoA availability, growth rate, ATP requirements, requirements. and flux distribution at the pyruvate node, and their Strains PpHP13, PpHP17, and PpHP18, which over- impact on the 3-HP production yield. Acetyl-CoA deple- express the cytosolic acetyl-CoA production pathway tion in strain PpHP8 hampered growth rate. When the (i.e., Pdc1 and ACSSeL641P), also produced more NADPH cytosolic acetyl-CoA biosynthetic pathway was overex- through the oxidative branch of the PPP than the actual pressed (strains PpHP13, PpHP17, and PpHP18), growth cell requirements. The substantial increase in the fluxes rate increased. Increase in the growth rate increased ATP through UG and PPP for such strains coincided with requirements, which increased channelling of pyruvate an increase in the specific glycerol uptake rate and the into the TCA cycle, hampering the 3-HP product yield. growth rate (Fig. 6A, and Fig. 3), but the biomass yield decreased (Fig. 3). Despite the large standard deviation Fluxome of P. pastoris strains at pH 3.5 of the fluxes through the UG and PPP fluxes in strains It has been reported that a pH of 3.5 (i.e., 1 unit below PpHP13, PpHP17, and PpHP18, it can be concluded the pKa of 3-HP) was optimal for the further downstream that these strains did not benefit from the heterolo- processing of 3-HP by solvent extraction [12]. Therefore, gous expression of the cPos5, while PpHP5, PpHP6, and the reference strain (X-33) as well as some 3-HP-produc- PpHP8 do. NADPH production by NADH kinase is more ing strains (i.e., PpHP1, PpHP6, PpHP8, PpHP15, and efficient in terms of both carbon and ATP conservation PpHP18) were further tested at pH 3.5. than the use of the UG and PPP, which would explain the No remarkable differences in the fluxome between reduction in the biomass and 3-HP yield in the strains the two conditions were observed for the reference PpHP13, PpHP17, and PpHP18. strain (Fig. 7A). In contrast, higher UG and PPP fluxes To corroborate the consistency of the observed meta- were observed at pH 3.5 than at pH 5 in strains PpHP1, bolic fluxes, FBA was used to verify the redox and energy PpHP6, PpHP8, PpHP15 (See Fig. 7B). In addition, a conservation of the 13C-MFA results. The maximiza- higher production of D-arabitol was also observed in tion of the flux through the ATP sink reaction (ATPM) most strains when growing at low pH (Fig. 7B and Addi- was used as an objective function. This is the function tional Figure S2). best describing the intracellular fluxes of a cell culture The fold-change of the relative fluxes of each recom- growing on a batch culture with an excess of substrate binant strain compared to the reference strain (Fig. 7C) [45]. Consequently, the resulting flux values are based showed that the impact of each genetic modification in on in vivo data, and at the same time, they fulfil the bio- the flux of each strain followed a similar trend at pH 5 logical function of optimizing the biomass yield. FBA and pH 3.5 (Figs. 4 and 7B, respectively). 3-HP was pro- results confirmed that both redox and energy balances duced at pH 3.5, but the yield was slightly lower than could be conserved at the given experimental fluxes. The the one achieved at pH 5 for all the tested strains. For FBA results were also used to calculate the ATP balance instance, the highest 3-HP producing strain at both pH (Fig. 6). values was PpHP8, which produced 3-HP at a yield of A higher ATP of maintenance for strains PpHP13, 0.084 ± 0.007 Cmol Cmol− 1 at pH 3.5 (0.081 ± 0.006 g PpHP17, and PpHP18 was observed, as depicted in g− 1), which is 23% lower than the product yield of the Fig. 6B, being PpHP18 the strain with the highest ATP same strain at pH 5. See Additional Figure S3 for the requirements among all strains. Conversely, the calcu- comparison of the fluxes at pH 5 and 3.5 for this strain. lated ATP requirement for strain PpHP8 was the lowest, Regarding the NADPH production and consumption particularly due to the lowest growth associated mainte- fluxes for each strain, the same trends were also observed nance energy (GAME) requirements, as it is the slowest at pH 3.5 (Additional Figure S4), that is, NADPH require- growing strain. ments in strain PpHP8 exceeded NADPH production Figure 6B shows there is a direct correlation between from the PPP, meaning the flux through the cytosolic the overall ATP requirements of the strain and the frac- NADH kinase reaction compensated for that difference. tion of pyruvate that was directed into the TCA cycle In contrast, the NADPH production through the PPP in for ATP production. While 47.1% of the pyruvate was strain PpHP18 greatly exceeded the requirements. More- channelled into the mitochondria in the reference over, the substrate uptake rate also followed the same strain, 32.1% was directed through the same pathway trend for all the strains in both pH conditions (Additional in strain PpHP8. The mitochondrial transport of pyru- Figure S4). vate raised to 40.2% and 41.9% when the cytosolic ace- Altogether, these results contribute to the understand- tyl-CoA pathway was expressed (strains PpHP13 and ing of the adaptation of this yeast to a low pH at a fluxome Fina et al. Microbial Cell Factories (2023) 22:117 Page 13 of 16 Fig. 7 A. Flux map with the relative fluxes for the parental strain grown at pH 5 and pH 3.5. Metabolite abbreviations can be found in the caption of Fig. 5. B. Log2 fold-change of the relative fluxes at a culture pH of 3.5 of the strains X-33, PpHP1, PpHP6, PpHP8, PpHP15, and PpHP18 compared to the same strains grown at pH 5. The colour scale describes the Log2 fold-change at a -2 to 2 range. C. Log2 fold-change of the relative fluxes at a culture pH of 3.5 of the strains PpHP1, PpHP6, PpHP8, PpHP15, and PpHP18 compared to the parental strain. The colour scale describes the Log2 fold-change at a -1 to 1 range. ‘MCR’ fluxes fold-change are compared to PpHP1. The upper bound of the colour scale of the ‘MCR’ flux is set to 4 Fina et al. Microbial Cell Factories (2023) 22:117 Page 14 of 16 level. Production of D-arabitol at acidic pH was increased this end, we also provide an end-to-end description of the for all the 3-HP-producing strains (from 2 to 20-fold). P. fluxomics workflow, as well as sharing the raw and pro- pastoris produces D-arabitol under several stress condi- cessed datasets following the guidelines of good practices tions, such as under hypoxia or osmotic stress [53, 55]. in publishing 13C-metabolic flux analyses studies, with Thus, higher D-arabitol production at a lower pH is prob- the aim of contributing spreading the use of fluxomics ably due to a stress response. Moreover, the biomass yield analyses in P. pastoris. at pH 3.5 was lower for most strains (Fig. 7B), indicating a higher ATP requirement for maintenance. Such results Supplementary Information have already been described in other yeasts grown at The online version contains supplementary material available at https://doi. org/10.1186/s12934-023-02123-0. lower pH, such as S. cerevisiae, where the decrease in the biomass yield was also attributed to an increase in Additional Figure 1. Flux maps of the parental P. pastoris strain (X-33) the ATP of maintenance [56]. Moreover, the 3-HP yield growing on methanol. Comparison of the flux map obtained in this was lower than the one of the same strains at pH 5, con- study (batch mini bioreactor) with the previously reported flux maps in chemostat cultures (https://doi.org/10.1016/j.nbt.2019.01.005). Additional sistently with previous studies describing 3-HP produc- Figure S2. Bioprocess parameters of the parental P. pastoris strain and tion in S. cerevisiae grown at pH 3.5 [57]. The observed nine 3-HP-producing strains cultivated in glycerol batch mini bioreactor decrease in the product yield when the ATP usage cultures at pH 3.5. Additional Figure S3. Flux map with the relative fluxes for the PpHP8 strain grown at pH 5 and pH 3.5. Additional Figure S4. increases confirms that ATP is a limiting factor towards Production and consumption rates of NADPH and specific glycerol uptake increasing the 3-HP yield. Similarly, increased D-arabitol rates for each strain at pH 3.5. production, which is a NADPH sink, can also explain the decrease in the 3-HP yield observed at low pH. Acknowledgements Still, as metabolic flux profiles at pH 3.5 remained Not applicable. mostly unchanged compared to those at pH 5, it is likely Author contributions that strain engineering strategies at both pH will have the All authors contributed to the study conception and design of experiments. same outcome. AF and SH performed the mini-bioreactor cultures. AF performed the analyses and calculations. AF wrote the first draft of the manuscript. All authors critically reviewed the manuscript and approved its final version. Conclusions This study describes the parallel characterization of a set Funding This work was funded by project PID2019-104666GB-I00 from the Spanish of 3-HP-producing P. pastoris recombinant strains at two Ministry of Science and Innovation; 2017-SGR-1462 from the Agència de relevant process conditions (pH 5 and pH 3.5) using a Gestió d’Ajuts Universitaris i de Recerca (AGAUR) of the Catalan Government. HT approach that has allowed to save time and resources AF PhD scholarship was funded with FPU grant (FPU17/05434) from the Spanish Ministry of Science. AF was granted with a Short-Term Fellowship compared to conventional strain-by-strain sequential (8853) from the European Molecular Biology Organization (EMBO) for a approach. Overall, we show, step-by-step, the setup of an research stay at the Toulouse Biotechnology Institute. optimized workflow for HT metabolic flux profiling of Data availability Pichia pastoris. It provides meaningful insights regarding The datasets supporting the conclusions of this article are included within the the impact of each genetic perturbation on the metabolic article and its additional files. The Files can be downloaded from https://doi. flux distribution of the 3-HP producing strain, point- org/10.34810/data673. The additional figures can be downloaded from the Electronic Supplementary materials. ing to a competition for energy and carbon resources for either cell growth or 3-HP production as the major cause of the observed phenotypes, regardless of the pH Declarations of the culture. Thus, it is concluded that both acetyl-CoA Competing interests and ATP limitations are the main bottlenecks hamper- The authors declare no competing interests. ing 3-HP production in P. pastoris. To overcome such Ethics approval and consent to participate bottlenecks, a strategy to increase the glycolytic fluxes Not applicable. (e.g., rewiring the regulatory mechanisms of this pathway Consent for publication and/or selecting cultivation conditions that favour higher Not applicable. glycolytic fluxes) should be addressed. Overall, this study will contribute towards the improvement of P. pastoris Author details 1 Department of Chemical, Biological and Environmental Engineering, strains and bioprocess engineering strategies to produce Universitat Autònoma de Barcelona, Bellaterra, Catalonia 08193, Spain 3-HP and other acetyl-CoA-derived products. Impor- 2 TBI, Université de Toulouse, CNRS, INRAE, INSA, Toulouse 31077, France tantly, this study showcases the potential of automated fluoxomics workflows for accelerated strain characteri- Received: 10 March 2023 / Accepted: 27 May 2023 sation in the context of the so-called Design-Build-Test- Learn cycle for metabolic engineering of industrially relevant microorganisms, beyond model organisms. To Fina et al. Microbial Cell Factories (2023) 22:117 Page 15 of 16 References 19. Wasylenko TM, Stephanopoulos G. Metabolomic and 13C-metabolic flux anal- 1. Schwarzhans JP, Luttermann T, Geier M, Kalinowski J, Friehs K. Towards sys- ysis of a xylose-consuming Saccharomyces cerevisiae strain expressing xylose tems metabolic engineering in Pichia pastoris. Biotechnol Adv. 2017. https:// isomerase. Biotechnol Bioeng. 2015;112:470–83. https://doi.org/10.1002/ doi.org/10.1016/j.biotechadv.2017.07.009. bit.25447. 2. Peña DA, Gasser B, Zanghellini J, Steiger MG, Mattanovich D. Metabolic engi- 20. Wiechert W, Möllney M, Petersen S, de Graaf AA. A universal framework neering of Pichia pastoris. Metab Eng. 2018;50:2–15. https://doi.org/10.1016/j. for 13C metabolic flux analysis. Metab Eng. 2001;3:265–83. https://doi. ymben.2018.04.017. org/10.1006/mben.2001.0188. 3. De S, Mattanovich D, Ferrer P, Gasser B. Established tools and emerging trends 21. Ferrer P, Albiol J. 2014. 13C-Based metabolic flux analysis of recom- for the production of recombinant proteins and metabolites in Pichia pasto- binant Pichia pastoris. Methods Mol Biol 1191, 291–313. https://doi. ris. Essays Biochem. 2021;65:293–307. https://doi.org/10.1042/EBC20200138. org/10.1007/978-1-4939-1170-7_17. 4. Fina A, Brêda GC, Pérez-Trujillo M, Freire DMG, Almeida RV, Albiol J, Ferrer 22. Rußmayer H, Buchetics M, Gruber C, Valli M, Grillitsch K, Modarres G, Guerra- P. Benchmarking recombinant Pichia pastoris for 3-hydroxypropionic acid sio R, Klavins K, Neubauer S, Drexler H, Steiger M, Troyer C, al, Chalabi A, production from glycerol. Microb Biotechnol. 2021;14:1671–82. https://doi. Krebiehl G, Sonntag D, Zellnig G, Daum G, Graf AB, Altmann F, Koellensperger org/10.1111/1751-7915.13833. G, Hann S, Sauer M, Mattanovich D, Gasser B. Systems-level organization of 5. Fina A, Heux S, Albiol J, Ferrer P. Combining metabolic engineering and mul- yeast methylotrophic lifestyle. BMC Biol. 2015;13:80. https://doi.org/10.1186/ tiplexed screening methods for 3-hydroxypropionic acid production in Pichia s12915-015-0186-5. pastoris. Front Bioeng Biotechnol. 2022;10:942304. https://doi.org/10.3389/ 23. Tomàs-Gamisans M, Ødum ASR, Workman M, Ferrer P, Albiol J. Glycerol fbioe.2022.942304. metabolism of Pichia pastoris (Komagataella spp.) characterised by 13C-based 6. Werpy T, Petersen G. Top value added chemicals from biomass volume I - metabolic flux analysis. N Biotechnol. 2019;50:52–9. https://doi.org/10.1016/j. results of screening for potential candidates from sugars and synthesis gas. nbt.2019.01.005. U.S. Department of Energy; 2004. https://doi.org/10.2172/15008859. 24. Tomàs-Gamisans M, Ferrer P, Albiol J. Fine-tuning the P. pastoris iMT1026 7. della Pina C, Falletta E, Rossi M. A green approach to chemical building genome-scale metabolic model for improved prediction of growth on meth- blocks. The case of 3-hydroxypropanoic acid. Green Chem. 2011;13:1624. anol or glycerol as sole carbon sources. Microb Biotechnol. 2018;11:224–37. https://doi.org/10.1039/c1gc15052a. https://doi.org/10.1111/1751-7915.12871. 8. Kildegaard KR, Jensen NB, Schneider K, Czarnotta E, Özdemir E, Klein T, Maury 25. Clark K, Karsch-Mizrachi I, Lipman DJ, Ostell J, Sayers EW. GenBank Nucleic J, Ebert BE, Christensen HB, Chen Y, Kim IK, Herrgård MJ, Blank LM, Forster J, Acids Res. 2016;44:D67–D72. https://doi.org/10.1093/nar/gkv1276. Nielsen J, Borodina I. Engineering and systems-level analysis of Saccharo- 26. Tomàs-Gamisans M, Ferrer P, Albiol J. Integration and validation of the myces cerevisiae for production of 3-hydroxypropionic acid via malonyl-CoA genome-scale metabolic models of Pichia pastoris: a comprehensive update reductase-dependent pathway. Microb Cell Fact. 2016;15:53. https://doi. of protein glycosylation pathways, lipid and energy metabolism. PLoS ONE. org/10.1186/s12934-016-0451-5. 2016;11:e0148031. https://doi.org/10.1371/journal.pone.0148031. 9. Takayama S, Ozaki A, Konishi R, Otomo C, Kishida M, Hirata Y, Matsumoto T, 27. Klamt S, Saez-Rodriguez J, Gilles ED. Structural and functional analysis of Tanaka T, Kondo A. Enhancing 3-hydroxypropionic acid production in combi- cellular networks with CellNetAnalyzer. BMC Syst Biol. 2007;1:2. https://doi. nation with sugar supply engineering by cell surface-display and metabolic org/10.1186/1752-0509-1-2. engineering of Schizosaccharomyces pombe. Microb Cell Fact. 2018;17:176. 28. Gombert AK, dos Santos MM, Christensen B, Nielsen J. Network identification https://doi.org/10.1186/s12934-018-1025-5. and flux quantification in the central metabolism of Saccharomyces cerevisiae 10. van Maris AJA, Konings WN, van Dijken JP, Pronk JT. Microbial export of under different conditions of glucose repression. J Bacteriol. 2001;183:1441– lactic and 3-hydroxypropanoic acid: implications for industrial fermenta- 51. https://doi.org/10.1128/JB.183.4.1441-1451.2001. tion processes. Metab Eng. 2004;6:245–55. https://doi.org/10.1016/j. 29. Solà A, Maaheimo H, Ylönen K, Ferrer P, Szyperski T. Amino acid biosynthesis ymben.2004.05.001. and metabolic flux profiling of Pichia pastoris. Eur J Biochem. 2004;271:2462– 11. López-Garzón CS, Straathof AJJ. Recovery of carboxylic acids produced by 70. https://doi.org/10.1111/j.1432-1033.2004.04176.x. fermentation. Biotechnol Adv. 2014;32:873–904. https://doi.org/10.1016/j. 30. Lehnen M, Ebert BE, Blank LM. A comprehensive evaluation of constrain- biotechadv.2014.04.002. ing amino acid biosynthesis in compartmented models for metabolic flux 12. Chemarin F, Athès V, Bedu M, Loty T, Allais F, Trelea IC, Moussa M. Towards an analysis. Metab Eng Commun. 2017;5:34–44. https://doi.org/10.1016/j. in situ product recovery of bio-based 3-hydroxypropionic acid: influence of meteno.2017.07.001. bioconversion broth components on membrane-assisted reactive extrac- 31. Kleijn RJ, van Winden WA, van Gulik WM, Heijnen JJ. Revisiting the 13C-label tion. J Chem Technol Biotechnol. 2019;94:964–72. https://doi.org/10.1002/ distribution of the non-oxidative branch of the pentose phosphate pathway jctb.5845. based upon kinetic and genetic evidence. FEBS J. 2005;272:4970–82. https:// 13. Koganesawa N, Aizawa T, Shimojo H, Miura K, Ohnishi A, Demura M, doi.org/10.1111/j.1742-4658.2005.04907.x. Hayakawa Y, Nitta K, Kawano K. Expression and purification of a small 32. Wiechert W, de Graaf AA. Bidirectional reaction steps in metabolic networks. cytokine growth-blocking peptide from armyworm Pseudaletia separata Biotechnol Bioeng. 1997;55:101–35. by an optimized fermentation method using the methylotrophic yeast 33. Chelliah V, Juty N, Ajmera I, Ali R, Dumousseau M, Glont M, Hucka M, Jalowicki Pichia pastoris. Protein Expr Purif. 2002;25:416–25. https://doi.org/10.1016/ G, Keating S, Knight-Schrijver V, Lloret-Villas A, Natarajan KN, Pettit JB, Rodri- S1046-5928(02)00036-0. guez N, Schubert M, Wimalaratne SM, Zhao Y, Hermjakob H, le Novère N, 14. Damasceno LM, Pla I, Chang HJ, Cohen L, Ritter G, Old LJ, Batt CA. An Laibe C. BioModels: ten-year anniversary. Nucleic Acids Res. 2015;43:D542–8. optimized fermentation process for high-level production of a single-chain https://doi.org/10.1093/nar/gku1181. fv antibody fragment in Pichia pastoris. Protein Expr Purif. 2004;37:18–26. 34. Millard P, Sokol S, Letisse F, Portais J-C. IsoDesign: A software for optimizing https://doi.org/10.1016/j.pep.2004.03.019. the design of 13C-metabolic flux analysis experiments. Biotechnol Bioeng. 15. Heux S, Meynial-Salles I, O’Donohue MJ, Dumon C. White biotechnol- 2014;111:202–8. https://doi.org/10.1002/bit.24997/abstract. ogy: state of the art strategies for the development of biocatalysts for 35. Sokol S, Millard P, Portais JC. Influx_s: increasing numerical stability and preci- biorefining. Biotechnol Adv. 2015;33:1653–70. https://doi.org/10.1016/j. sion for metabolic flux analysis in isotope labelling experiments. Bioinformat- biotechadv.2015.08.004. ics. 2012;28:687–93. https://doi.org/10.1093/bioinformatics/btr716. 16. Heux S, Bergès C, Millard P, Portais JC, Létisse F. Recent advances in high- 36. Heux S, Juliette P, Stéphane M, Serguei S, Jean-Charles P. A novel platform for throughput 13C-fluxomics. Curr Opin Biotechnol. 2017;43:104–9. https://doi. automated high-throughput fluxome profiling of metabolic variants. Metab org/10.1016/j.copbio.2016.10.010. Eng. 2014;25:8–19. https://doi.org/10.1016/j.ymben.2014.06.001. 17. Teworte S, Malcı K, Walls LE, Halim M, Rios-Solis L. Recent advances in fed- 37. Maurer M, Kühleitner M, Gasser B, Mattanovich D. Versatile modeling and batch microscale bioreactor design. Biotechnol Adv. 2022;55:107888. https:// optimization of fed batch processes for the production of secreted heter- doi.org/10.1016/j.biotechadv.2021.107888. ologous proteins with Pichia pastoris. Microb Cell Fact. 2006;5:37. https://doi. 18. Kohlstedt M, Becker J, Wittmann C. Metabolic fluxes and beyond - sys- org/10.1186/1475-2859-5-37. tems biology understanding and engineering of microbial metabolism. 38. Heuillet M, Bellvert F, Cahoreau E, Letisse F, Millard P, Portais JC. Methodology Appl Microbiol Biotechnol. 2010;88:1065–75. https://doi.org/10.1007/ for the validation of isotopic analyses by mass spectrometry in stable-isotope s00253-010-2854-2. labeling experiments. Anal Chem. 2018;90:1852–60. https://doi.org/10.1021/ acs.analchem.7b03886. Fina et al. Microbial Cell Factories (2023) 22:117 Page 16 of 16 39. Millard P, Delépine B, Guionnet M, Heuillet M, Bellvert F, Létisse F. IsoCor: recombinant Pichia pastoris (Komagataella phaffii): impact on cell physiology isotope correction for high-resolution MS labeling experiments. Bioinformat- and recombinant production of secreted proteins. Appl Environ Microbiol. ics. 2019;35:4484–7. https://doi.org/10.1093/bioinformatics/btz209. 2020;86:e02038–19. https://doi.org/10.1128/AEM.02038-19. 40. Peiro C, Millard P, de Simone A, Cahoreau E, Peyriga L, Enjalbert B, Heux 51. van Hoek P, Flikweert MT, van der Aart QJM, Steensma HY, van Dijken JP, Pronk S. Chemical and metabolic controls on dihydroxyacetone metabolism JT. Effects of pyruvate decarboxylase overproduction on flux distribution at lead to suboptimal growth of Escherichia coli. Appl Environ Microbiol. the pyruvate branch point in Saccharomyces cerevisiae. Appl Environ Micro- 2019;85:e00768–19. https://doi.org/10.1128/AEM.00768-19. biol. 1998;64:2133–40. https://doi.org/10.1128/aem.64.6.2133-2140.1998. 41. Schellenberger J, Que R, Fleming R, Thiele I, Orth J, Feist A, Zielinski D, Bordbar 52. Schaaff I, Heinisch J, Zimmerman F. Overexpression of glycolytic enzymes in A, Lewis N, Rahmanian S, Kang J, Hyduke D, Palsson B. Quantitative prediction yeast. Yeast. 1989;5:285–90. of cellular metabolism with constraint-based models: the COBRA Toolbox 53. Baumann K, Carnicer M, Dragosits M, Graf AB, Stadlmann J, Jouhten P, v2.0. Nat Protoc. 2011;6:1290–307. https://doi.org/10.1038/nprot.2011.308. Maaheimo H, Gasser B, Albiol J, Mattanovich D, Ferrer P. A multi-level study 42. Crown SB, Antoniewicz MR. Publishing 13C metabolic flux analysis studies: of recombinant Pichia pastoris in different oxygen conditions. BMC Syst Biol. a review and future perspectives. Metab Eng. 2013;20:42–8. https://doi. 2010;4:141. https://doi.org/10.1186/1752-0509-4-141. org/10.1016/j.ymben.2013.08.005. 54. Ata Ö, Rebnegger C, Tatto NE, Valli M, Mairinger T, Hann S, Steiger MG, Çalık P, 43. Droste P, Miebach S, Niedenführ S, Wiechert W, Nöh K. Visualizing multi-omics Mattanovich D. A single Gal4-like transcription factor activates the Crabtree data in metabolic networks with the software omix - a case study. BioSys- effect in Komagataella phaffii. Nat Commun. 2018;9:4911. https://doi. tems. 2011;105:154–61. https://doi.org/10.1016/j.biosystems.2011.04.003. org/10.1038/s41467-018-07430-4. 44. Orth JD, Thiele I, Palsson BØO. What is flux balance analysis? Nat Biotechnol. 55. Dragosits M, Stadlmann J, Graf A, Gasser B, Maurer M, Sauer M, Kreil DP, 2010;28:245–8. https://doi.org/10.1038/Nbt.1614. Altmann F, Mattanovich D. The response to unfolded protein is involved in 45. Schuetz R, Kuepfer L, Sauer U. Systematic evaluation of objective functions osmotolerance of Pichia pastoris. BMC Genomics. 2010;11:207. https://doi. for predicting intracellular fluxes in Escherichia coli. Mol Syst Biol. 2007;3:119. org/10.1186/1471-2164-11-207. https://doi.org/10.1038/msb4100162. 56. Hakkaart X, Liu Y, Hulst M, el Masoudi A, Peuscher E, Pronk J, van Gulik W, 46. Liu C, Wang Q, Xian M, Ding Y, Zhao G. 2013. Dissection of malonyl-coenzyme Daran-Lapujade P. Physiological responses of Saccharomyces cerevisiae to A reductase of Chloroflexus aurantiacus results in enzyme activity improve- industrially relevant conditions: slow growth, low pH, and high CO2 levels. ment. PLoS ONE 8, e75554. Biotechnol Bioeng. 2020;117:721–35. https://doi.org/10.1002/bit.27210. 47. Shi L, Tu BP. Acetyl-CoA and the regulation of metabolism: mechanisms and 57. Borodina I, Kildegaard KR, Jensen NB, Blicher TH, Maury J, Sherstyk S, Schnei- consequences. Curr Opin Cell Biol. 2015;33:125–31. https://doi.org/10.1016/j. der K, Lamosa P, Herrgård MJ, Rosenstand I, Öberg F, Forster J, Nielsen J. Estab- ceb.2015.02.003. lishing a synthetic pathway for high-level production of 3-hydroxypropionic 48. Pereira H, Azevedo F, Domingues L, Johansson B. Expression of Yarrowia acid in Saccharomyces cerevisiae via β-alanine. Metab Eng. 2015;27:57–64. lipolytica acetyl-CoA carboxylase in Saccharomyces cerevisiae and its effect https://doi.org/10.1016/j.ymben.2014.10.003. on in-vivo accumulation of Malonyl-CoA. Comput Struct Biotechnol J. 2022;20:779–87. https://doi.org/10.1016/j.csbj.2022.01.020. 49. Christodoulou D, Kuehne A, Estermann A, Fuhrer T, Lang P, Sauer U. Reserve flux capacity in the pentose phosphate pathway by NADPH binding is con- Publisher’s Note served across kingdoms. iScience. 2019;19:1133–44. https://doi.org/10.1016/j. Springer Nature remains neutral with regard to jurisdictional claims in isci.2019.08.047. published maps and institutional affiliations. 50. Tomàs-Gamisans M, Andrade CCP, Maresca F, Monforte S, Ferrer P, Albiol J. Redox engineering by ectopic overexpression of NADH kinase in