Exploration of In Situ Extraction for Enhanced Triterpenoid Production by Saccharomyces cerevisiae Authors: Mariam Dianat, Sarah Straaten, Aldo Maritato, Daniel Wibberg, Tobias Busche, Lars M. Blank, Birgitta E. Ebert Published: 2024 ABSTRACT ABSTRACTPlant‐derived triterpenoids are in high demand due to their valuable applications in cosmetic, nutraceutical, and pharmaceutical industries. To meet this demand, microbial production of triterpenoids is being developed for large‐scale production. However, a prominent limitation of microbial synthesis is the intracellular accumulation, requiring cell disruption during downstream processing. Destroying the whole‐cell catalyst drives up production costs and limits productivity and product yield per cell. Here, in situ product extraction of triterpenoids into a second organic phase was researched to address this limitation. An organic solvent screening identified water‐immiscible isopropyl myristate as a suitable in situ extractant, enabling extraction of up to 90% of total triterpenoids from engineered Saccharomyces cerevisiae . Combining isopropyl myristate and β‐cyclodextrins improved extraction efficiency. In a first configuration, repeated batch fermentation with sequential product extraction and cell recycling resulted in 1.8 times higher production than a reference fermentation without in situ product extraction. In the second configuration, yeast cells were in contact with the second organic phase throughout a fed‐batch fermentation to continuously extract triterpenoids. This resulted in 90% product extraction and an extended production phase. Further improvement of triterpenoid production was not achieved due to microbial host limitations uncovered through omics analyses. Introduction Plant‐derived triterpenoids are a large group of natural products widely used in traditional medicine (Liu and Cheng 2012; Ren and Kinghorn 2019). Due to their broad applications, these compounds awakened huge interest in different sectors, such as the consumer care, nutritional, and pharmaceutical industries (Caputi and Aprea 2011; Bergman, Davis, and Phillips 2019; Jahangeer et al. 2021). Microbial‐based heterologous triterpenoid production is a promising alternative to exploiting plants or animals (Guo, Wang, and Huo 2020) or complex chemical synthesis (Wang and Hui 2021). Saccharomyces cerevisiae is a suitable, robust host for the heterologous synthesis of triterpenoids, which can be partially attributed to the native synthesis of 2,3‐oxidosqualene in the sterol pathway, the common precursor of many triterpenoids (Guo, Jacobsen, et al. 2022; Guo, Wang, et al. 2022). However, this promising microbial platform has one significant limitation: To date, no terpenoid transporters have been described for S. cerevisiae (Fang and Xiao 2021). While some studies have reported triterpenoids to be secreted into the medium (Zhou et al. 2016), in our work, triterpenoids were primarily detected intracellularly (Czarnotta et al. 2017). Hence, it is unclear whether the extracellular detection of triterpenoids was due to secretion or cell lysis‐related product release into the medium. Due to their hydrophobic structure, triterpenoids are generally assumed to accumulate in or on the ER membrane (Dianat et al. 2023), the subcellular compartment where triterpenoid biosynthesis is localized. Accumulation in lipid droplets and vesicular transport from ER to the plasma membrane has also been suggested (Fang and Xiao 2021). Intracellular triterpenoid accumulation constitutes a major constraint for microbial production. It prohibits prolonged production of triterpenoids by the same biocatalyst and necessitates cell disruption for product release in the downstream processing step, causing higher costs, more complexity, and the loss of the biocatalyst. Hence, designing microbial processes where the product is secreted is preferred to improve process economics. Based on the properties of triterpenoids, the application of a lipophilic solvent in a two‐phase liquid–liquid extraction system could be a suitable method to facilitate the extraction of triterpenoids from the cell. For instance, Sun, Kwak, and Jin (2019) applied a two‐phase in situ extraction with olive oil and dodecane to extract vitamin A from S. cerevisiae . Similarly, isopropyl myristate (IPM) was used as the organic phase for the extraction of triterpenoids and carotenoids from Yarrowia lipolytica (Zhang et al. 2020) and from S. cerevisiae , respectively (Yamada et al. 2022). Furthermore, other studies showed the successful extraction of triterpenoids with cyclodextrins (CDs), cyclic oligosaccharides consisting of glucose subunits (Moses et al. 2014; Lan and Xiao 2020). The spatial arrangement of the glucose molecules creates a hydrophobic inner cavity that enables the encapsulation of hydrophobic compounds. The outer part of cyclodextrins is hydrophilic due to the presence of hydroxyl groups (see Figure S1), which allows the interaction of cyclodextrins with aqueous systems. Hence, cyclodextrins can be applied in drug‐delivery systems to increase aqueous solubility of poorly soluble compounds (Carneiro et al. 2019). Although the detailed mechanism of CD‐mediated triterpenoid transport is not understood, studies imply CD interactions with the cell membrane. Motivated by these findings, we explored the potential advantages of in situ product extraction on the example of production of the pentacyclic triterpenoid betulinic acid in S. cerevisiae . A S. cerevisiae CENPK strain expressing a recombinant betulinic acid biosynthesis pathway (Figure 1) was used throughout this study. In addition to betulinic acid, the strain also accumulates the less oxidized intermediates betulin and betulin aldehyde of this triterpenoid because of suboptimal activity of the heterologous cytochrome P450 monooxygenase. The high intracellular accumulation of these triterpenoids at around 0.20 ± 0.02 g/gCDW (Dianat et al. 2023) makes this strain an ideal test bed for in situ product extraction. We performed solvent screening and applied the best‐performing solvent for in situ extraction to (1) evaluate the possible increase of the cells' capacity for production, (2) ease the downstream process by separating the product from the biomass, and (3) investigate the possibility of continuously utilizing the same biocatalyst. Schematic overview of cytosolic mevalonate (blue) and farnesyl diphosphate synthase (FPP) (red) pathways. FPP is converted to squalene by Erg9p, which is then converted into yeast native ergosterol pathway (yellow). 2,3‐Oxidosqualene also serves as the common precursor for heterologous biosynthesis of betulinic acid (green). CPR, cytochrome P450 reductase; CYP, cytochrome P450 monooxygenase; Erg10p, acetyl‐CoA C‐acetyltransferase; Erg11p, lanosterol 14‐α‐demethylase; Erg12p, mevalonate kinase; Erg13p, HMG‐CoA synthase; Erg19p, mevalonate pyrophosphate decarboxylase; Erg1p, squalene epoxidase; Erg20p, farnesyl pyrophosphate synthetase; Erg24p, C‐14 sterol reductase; Erg25p, sterol C‐4 methyloxydase; Erg26p, sterol C‐3 dehydrogenase; Erg27p, sterol C‐3 ketoreductase; Erg2p, sterol isomerase; Erg3p, sterol desaturase; Erg4p, sterol reductase; Erg5p, sterol desaturase; Erg6p, delta 24‐sterol C‐methyltransferase; Erg7p, lanosterol synthase; Erg8p, phosphomevalonate kinase; Erg9p, squalene synthase; Hmg1/2, HMG‐CoA reductase; OEW, lupeol synthase. Adapted and modified from Czarnotta et al. (2017). Material and Methods Strains and Media The strain Saccharomyces cerevisiae BA6∆ used in this study was derived from S. cerevisiae CENPK 102‐5B (MATa, ura3‐52, his3‐11, leu2‐3/112, TRP1, MAL2‐8c, and SUC2). It is based on a strain expressing a truncated form of HMG1 (Polakowski et al. 1998), a single copy of Arabidopsis thaliana ATR2 and multiple copies of Vitis vinifera CYP716A15 (with pTEF promotor and tADH terminator) and Olea europaea OEW (with pPGK1 promotor and tCYC1 terminator). All engineered and heterologous genes were integrated into the yeast's chromosomes. Lanosterol synthase Erg7p activity was downregulated by replacement of the native gene with an ERG7 mutant (Guo, Jacobsen, et al. 2022; Guo, Wang, et al. 2022). In addition to these modifications, S. cerevisiae BA6∆ was further improved for triterpenoid production by deleting PAH1 (YMR165C) through replacement with a hygromycin resistance cassette (YMR165CΔ:hphMX; Dianat et al. 2023). Cryogenic stocks were prepared by mixing equal volumes, 400 μL each, of a sterile glycerol (40% v/v) solution and a 40‐h culture in mineral salt medium and storing at −80°C. 500 μL of a single cryogenic stock was used to inoculate a 50 mL pre‐culture. All cultivations were performed in WM8+ (Czarnotta et al. 2017), a modified version of the mineral salt medium WM8 (Lang and Looman 1995), complemented with vitamin and trace elements of Verduyn minimal medium (Verduyn et al. 1992), resulting in the following composition: (NH4)H2PO4 0.25 g/L, NH4Cl 2.80 g/L, sodium glutamate 10 g/L, MgCl2·6H2O 0.25 g/L, CaCl2·2H2O 4.6 g/L, KH2PO4 2 g/L, MgSO4·7H2O 0.55 g/L, myoinositol 100 mg/L, ZnSO4·7H2O 6.25 mg/L, FeSO4·7H2O 3.50 mg/L, CuSO4·5H2O 0.4 mg/L, MnCl2·4H2O 0.1 mg/L, MnCl2·2H2O 1 mg/L, Na2MoO4·2H2O 0.5 mg/L, nicotinic acid 11 mg/L, pyridoxin‐HCl 26 mg/L, thiamine‐HCl 11 mg/L, biotin 2.55 mg/L, calcium pantothenate 51 mg/L, Na2EDTA 2.92 g/L, CoCl2.6H2O 0.3 mg/L, H3BO3 1 mg/L, KI 0.1 mg/L, and p‐aminobenzoic acid 0.2 mg/L. Depending on the specific experiment and detailed in the results, glucose or ethanol was used as the carbon source. Shake Flask Cultivation Batch cultivations in 500‐mL shake flasks were conducted with a filling volume of 20 mL. A pre‐culture was used to inoculate the main cultures to a starting optical density measured at 600 nm (OD600) of 0.1. The OD600 of the cultures was determined using a spectrophotometer (Ultrospec 10, Biochrom, Germany) with 0.5% phosphate‐buffered saline (PBS) used as blank. The cultivations took place in a rotary shaker (New Brunswick Innova 44R) at 30°C, 300 rpm, and 25 mm shaking diameter. Online Monitoring of Exhaust CO2 Online monitoring of CO2 in the headspace of batch shake flask experiments was performed with BlueSens CO2 sensors BCP‐CO2 screwed onto the shake flask (see Figure S2). The sealed shake flask had a total volume of 1300 mL and was filled with 50 mL medium and 2 mL (4 vol. %) organic solvent for extraction. Oxygen limitation in the sealed shake flask was avoided by calculating the total amount of oxygen available in the shake flask's headspace and adjusting glucose concentration to the total available oxygen, ensuring the possibility of complete oxidation (cf to Data S1: Methods for details). The cultivation took place at 30°C, 150 rpm, and 2.5 cm shaking diameter (Eppendorf, Germany) for 72 h, and the initial OD600 was set to 0.1. Solvent Screening For solvent screening, a batch culture in stationary phase was distributed onto 15‐mL reaction tubes, in which the extraction was conducted. The total volume was kept constant at 5 mL. For the initial screening, 20 vol.% of each solvent were tested (equals a volume of 1 mL solvent on 4‐mL cell culture). The solvent amount of 20 vol.% was chosen as this was in the range of 10–50 vol.%, which was applied in previous studies (Sun, Kwak, and Jin 2019; Zhang et al. 2020; Yamada et al. 2022). After incubation at 60 rpm for 2 h, reaction tubes were collected and centrifuged at 4000 rpm (3011 × g) for 5 min. The interphase between organic and aqueous phases was marked. The organic phase was discarded, and the cell pellet was washed with 0.5% PBS (v/v). After discarding the supernatant, the inner tube surface was cleaned with paper towels to remove solvent residues. The pellet was resuspended in 0.5% PBS to the original volume as marked on the tube to obtain the initial biomass concentration. Technical duplicates of 800 μL were taken from the cell suspension and stored at −20°C until further processing. Furthermore, technical triplicates of 1 mL were taken for cell viability measurements. For a schematic overview, see Figure S2. In Situ Extraction of Triterpenoids in Shake Flasks To perform in situ extraction in a shake flask using isopropyl myristate (IPM) as solvent, IPM was added to a stationary batch culture and incubated at 30°C,150 rpm, and 25 mm shaking diameter for 24 h if not otherwise stated. After incubation, the entire content of the shake flask was centrifuged for 15 min at RT at 4000 rpm (3011 × g). Organic and aqueous phases were separated, and the pellet was resuspended in 0.5% PBS (see section 3.4). From this mixture, samples were taken to determine the intracellular product concentration (c product in pellet), the OD600, cell number, and cell viability. The extraction efficiency was determined from the distribution of the product in the organic phase and pellet using Equation (1). (cf. to Data S1: Methods for details): (1) Extraction efficiency%=Product in organic phasemgTotal amount of productmg×100% Repeated Batch Cultivation With Sequential Product Extraction Repeated batch experiments with sequential two‐phase product extraction were carried out by the addition of IPM at the end of each batch cultivation, followed by cell harvest, phase separation by centrifugation (10 min, 4000 rpm (3011 g)), washing the pellet in 0.5% PBS and resuspension of the pellet in fresh medium. A glucose concentration of 50 g/L was used for the initial batch of the control experiment and the in situ extraction approach (for obtaining high biomass concentrations). After 72 h of the first batch, 150 vol. % IPM was added for in situ extraction, while nothing was added to the control. After 24 h of incubation, the cultures were transferred into a 50‐mL reaction tube, centrifuged, and the pellet was resuspended in fresh WM8+ medium with 16 g/L ethanol. After 24 h of cultivation, 150 vol. % IPM was added for extraction, and the culture was again incubated for 24 h. This procedure was repeated five times (n = 5). For a schematic overview, see Figure S3. Application of β‐Cyclodextrins for Triterpenoid Extraction Prior to addition to the microbial culture, a weighed amount of β‐cyclodextrins (≥ 97% purity, Sigma‐Aldrich) was sterilized by exposure to UV light for 30 min. After incubation at 300 rpm and 30°C, the culture was poured into a reaction tube and centrifuged, and the supernatant and pellet were separated. The pellet was resuspended in 0.5% PBS and adjusted to the initial volume. The supernatant was evaporated; the residues were dissolved in chloroform and analysed by high‐performance liquid chromatography (HPLC). Samples from the resuspended pellet fraction were analysed by HPLC‐CAD as well. Determination of Triterpenoid Solubilities in IPM To determine maximal solubilities of betulin and betulinic acid, supersaturated solutions of these molecules were prepared in IPM. Diluted samples of these supersaturated solutions were analysed by HPLC to determine the concentration in the saturated solution. Each compound in each solvent was measured twice (technical replicates). Fed‐Batch Cultivation in Bioreactor Bioreactor cultivations were conducted in 1.3‐L vessels (BioFlo 120, Eppendorf, Hamburg, Germany) with a working volume of 500 mL, an aeration rate of 1 vvm, 30°C, and a pH of 6 monitored online with an autoclavable pH probe (Mettler Toledo, Switzerland) and regulated with 30% (v/v) H2SO4 and 30% (v/v) NH4OH. The dissolved oxygen tension (DOT) was measured with a pO2 probe (Mettler Toledo, Switzerland) and kept above 30% using a stirring cascade (350–1200 rpm). The temperature was measured using a Pt100 probe and was automatically regulated via the heating jacket (Eppendorf, Germany) and the water‐cooling system. The setup of the fermenter was completed by a stirrer with two Rushton turbines, a gas sparger, and an exhaust cooler. The medium was inoculated with an OD600 set to 0.4 via a septum using a sterile syringe. Pulsed fed‐batch cultivations were performed using a DO‐triggered pump system. The carbon source was pulsed into the bioreactor using a syringe via the septum when the DO exceeded 70%, indicating carbon source depletion. 150 vol. % IPM was transferred into the bioreactor via a separate feed tube at the end of the batch phase for in situ extraction. With each new pulse of carbon source, trace elements and vitamins were replenished. Determination of Cell Viability and Cell Number Cell viability analysis and cell counts were performed using the electrochemical impedance spectroscopy system Ampha Z32 (Amphasys, Lucerne, Switzerland). The single cell's impedance provides information about the integrity of the cell membrane, which correlates with cell viability. For the measurements, a 50‐μm microchip was used. To prepare the samples, 100 μL of cell culture was centrifuged, and the pellet was resuspended and diluted in 0.5× PBS to obtain an OD600 of approx. 0.05. By diluting all samples in 0.5% PBS, the same conductivity was assured for all samples, and the impedance was only dependent on the cell membrane integrity. Viable and dead cells were distinguished by gating using a sample with heat‐inactivated dead yeast cells as control. Measurements were run at two frequencies of 2 and 12 MHz and a pump speed of 80 rpm. The trigger level was set to 0.008 V, and the measurement‐specific settings modulation, amplification, and demodulation were set to 5, 4, and 1, respectively. Quantification of Extracellular Metabolites, Triterpenoids, and Ergosterol Extracellular metabolites (glucose, ethanol, glycerol, and acetate) were measured as described in (Czarnotta et al. 2017). Analysis of intracellular triterpenoids (betulinic acid, betulin, betulin aldehyde, and lupeol) and squalene was performed as described by (Czarnotta et al. 2017) using a reverse‐phase HPLC system (Thermo Fisher UltiMate 3000 Dionex) with a C18 column (Merck Spherisorb ODS‐2 (5 μm), 250 × 4 mm, Merck KGaA, Germany), and a guard column (EC 4/3 NUCLEODUR C18 Gravity, 3 μm, Macherey‐Nagel GmbH & Co. KG Germany), using a charged aerosol detector (CAD, esa ERC GmBH). Ergosterol was analysed with the same systems but was detected using a UV detector (Dionex UniMate 3000) at a wavelength of 300 nm. To quantify ergosterol, serial dilutions of ergosterol standards (Sigma‐Aldrich, ≥ 95%) in chloroform were measured in a concentration range of 0.03–3.0 g/L. The concentration of triterpenoids and ergosterol in the organic phase (IPM) was determined by directly injecting the organic phase into the HPLC without pre‐processing (for HPLC chromatograms, see Figures S2 and S3). For metabolite quantification, serial dilutions of standards of betulinic acid (Sigma‐Aldrich, ≥ 98%), betulin (Sigma‐Aldrich, ≥ 98%), betulin aldehyde (Cfm Oskar Tropitzsch GmbH, ≥ 98%), lupeol (Sigma‐Aldrich, ≥ 94%), and ergosterol (Sigma‐Aldrich, ≥ 95%) in IPM were measured (concentration ranges: betulinic acid 0.01–0.5 g/L, betulin 0.01–1.0 g/L, betulin aldehyde 0.01–0.34 g/L, lupeol 0.01–0.92 g/L, ergosterol 0.01–1.3 g/L). Corresponding concentrations in the aqueous phase, c product_aq [g/L], were determined from the measured concentrations in the organic phase using the following Equation (2): (2) cproduct_aq=cproduct inOPg/L×VorgL/VaqL Nanopore Library Preparation and GridION Sequencing Samples of S. cerevisiae BA6∆ grown in both ethanol‐pulsed and glucose‐pulsed fed‐batch cultures, taken at the start (0 h) and end (approx.150 h), were used for whole‐genome DNA isolation and sequencing. Whole‐genome DNA was isolated and purified as described before (Schafhauser et al. 2022). In brief, the ‘NucleoBond HMW DNA’ kit from Macherey‐Nagel (Düren, Germany) was used for DNA isolation according to the manufacturer's instructions. The isolated DNA was used for sequencing when the required quality criteria were fulfilled and when a 0.5% agarose control gel confirmed the presence of large DNA fragments (>> 10 kb) and the absence of visible amounts of RNA. A sequencing library with genomic DNA from the different yeast strains was prepared using the Nanopore Rapid DNA Sequencing kit (SQK‐LSK106, Oxford Nanopore Technologies, Oxford, UK) according to the manufacturer's instructions as described recently (Ullmann et al. 2022). Sequencing was performed on an Oxford Nanopore GridION Mk1 sequencer using an R9.4.1 flow cell, which was prepared according to the manufacturer's instructions. Base Calling, Assembly, and Analysis MinKNOW v21.05.25 (Oxford Nanopore Technologies) was used to control the run using the 48 h sequencing run protocol; base calling was performed offline using guppy v5.0.16. The assemblies were performed using canu v2.1.1 (Koren et al. 2017) and flye v2.9 (Freire, Ladra, and Parama 2021). For more insight, genomic contigs resulting after the assembly were analysed for large local similarities using BLASTn (Altschul 1991) as described before (Schwarzhans et al. 2016). Each contig was compared to a local database, including the vector sequence. Hits with an e‐value > 1 × 10−20 and a sequence identity of 95% were analysed in detail. In addition, using an in silico‐based approach, the insertion site of each vector into the S. cerevisiae genome was identified. In addition, a mapping approach using a Burrow–Wheeler Aligner (BWA) of 0.7 (Li 2013) was performed to determine the coverage of the vector in each sample to predict the amount of inserted vectors in the genome. Quantitative Proteomic Analysis Analysis of the protein abundance in pulsed fed‐batch cultures was performed for cells after a cultivation time of 23 h (t start), 74 h (t mid), and 147 h (t end) for ethanol‐pulsed and 17 h (t start), 57 h (t mid), and 160 h (t end) for the glucose‐pulsed fed‐batch. Proteomics analysis was performed by the Queensland Metabolomics and Proteomics Facility at the University of Queensland. Biomass samples taken during bioreactor experiments (Figure S6) were freeze‐dried to avoid degradation during transport. The samples were solubilized in 50 μL 5% sodium dodecyl sulphate (SDS) lysis buffer. Then, the proteins were reduced by adding 4 μL aliquot of 500 mM of dithiothreitol (DTT) and heating for 60 min at 70°C. The alkylation of cysteines was conducted by addition of 8 μL (500 mM) of iodoacetamide (IAA) and incubated in the dark for 30 min. The proteins were acidified with 2.5 μL to a final concentration of 12% phosphoric acid and diluted with 165 μL S‐Trap binding buffer to form colloidal protein particulate. Then, the mix was transferred onto S‐Trap Micro Spin columns (Protifi), trapped in a filter by centrifugations at 4000 × g for 1 min, and washed thrice with 150 μL S‐Trap binding buffer. Proteins were digested into peptides by adding a digestion buffer containing 2 μg trypsin and 25 μL of 50 mM ammonium bicarbonate (pH = 8) on top of the S‐traps. Then, the S‐trap was incubated in an oven overnight at 37°C. The digested peptides were eluted with 40 μL of 5% acetonitrile (ACN) in 0.1% formic acid, followed by 50% ACN in 0.1% formic acid and 75% ACN in 0.1% formic acid. All elutions were dried and resuspended in 5% acetonitrile (ACN) + 0.1% formic acid (FA) prior to liquid chromatography. Peptides were analysed by liquid chromatography‐mass spectrometry (LCMS). The HPLC system was an Ultimate 3000 RSLCnano (ThermoFisher, Germany), and the mobile phase was H2O (A) and 80% ACN (B), both in 0.1% formic acid. An aliquot of 2 μL sample was injected onto a trap column used for online desalting and washed at 20 μL/min prior to a resolving C18 nanoEaseTM M/Z CSH C18 1.7 μm (150 μm × 100 mm) column (Waters) using a 60 min gradient of 8 to 95% acetonitrile in 0.1% formic acid at 0.5 μL/min flow rate. The eluted peptides were electro‐sprayed into a Thermo nano‐Exploris Mass Spectrometer (ThermoFisher, USA). For full MS, a resolution of 60,000 and a scan range of 400–1200 m/z were applied. For MS2, the isolation window was 1.6 m/z, and the orbitrap resolution was set to 15,000. The post‐analysis was performed against an in‐house S. cerevisiae protein database using ThermoFisher Spectronaut (version 15.0). Data are available via ProteomeXchange with identifier PXD057071. Results and Discussion Methyl Decanoate, Ethyl Decanoate, and Isopropyl Myristate Are Suitable Biocompatible Extraction Solvents In our quest for environmentally friendly and biocompatible organic solvents suitable for in situ extraction, we selected eight candidates such as ethyl acetate, butanol, hexanol, octene, methyl decanoate, ethyl decanoate, farnesene, and isopropyl myristate (IPM), for our first round of screening (for schematic overview of the procedure, see Figure S2). These solvents were chosen with regard to biocompatibility, often correlated with a logPow value > 5, as suggested by Bruce and Daugulis (1991), and product solubility. LogPow is the logarithm of the partition coefficient of a given component in the octanol–water two‐phase system (Reslow, Adlercreutz, and Mattiasson 1987). The logPow values of the candidates are shown in Table 1. TABLE 1LogPow [−] value of solvents selected for triterpenoid extraction screening.SolventlogPow [−]PubChem IdentifierReferenceEthyl acetate0.7CID 8857Information (2024g)Butanol0.9CID 263Information (2024a)Hexanol2.0CID 8103Information (2024e)Octene4.6CID 8125Information (2024f)Methyl decanoate4.7CID 8050Information (2024d)Ethyl decanoate4.6CID 8048Information (2024c)Farnesene6.1CID 5281516Information (2024h)Isopropyl myristate7.2CID 8042Information (2024b) LogPow [−] value of solvents selected for triterpenoid extraction screening. We added 20 vol. % of each solvent to a culture in the stationary phase, and after 2 h of incubation at RT, the concentration of triterpenoids in the cell pellet and the cell viability of the individual extractions were determined. Untreated cells from a stationary phase culture were used as a control. Cells from extractions with ethyl acetate, butanol, and hexanol showed considerably low cell viability (Figure 2B), and minimal content in the pellet (Figure 2A), likely because of solvent‐induced cell disruption and release of intracellular products into the medium. Despite the high product solubility, these solvent candidates had to be excluded from the screening due to the very poor biocompatibility. The cell viability in setups using octene, methyl decanoate, ethyl decanoate, farnesene, and IPM as extractants was high and in the same range as in the control, indicating high biocompatibility of these solvents (Figure 2B). Product extraction with these solvents showed significant differences. Almost 80% of the product remained in the cells when treated with octene, suggesting low extraction efficiency. At the same time, treatment with methyl decanoate resulted in a high release of the products into the organic phase, and only 20% remained in the pellet. In cultures treated with ethyl decanoate, farnesene, and IPM, 40%–50% of the product remained in the cells, indicating a moderate triterpenoid extraction capacity (Figure 2A). Due to complications encountered during HPLC measurements with farnesene, our subsequent analyses focused on methyl decanoate, ethyl decanoate, and IPM. Ratio of intracellular triterpenoids in treated cells and the untreated control [%] (A), and viability [%] of cells treated with solvents (B). All cell suspensions were incubated with 20 vol. % solvent for 2 h, except for the control. IPM Does Not Impact Cell Metabolism To investigate yeast metabolism in the presence of methyl decanoate, ethyl decanoate, and IPM (20 vol. %), we conducted online measurement of the CO2 produced by S. cerevisiae when exposed to these three solvents alongside a control experiment without any solvent (Figure 3). The course of CO2 volume concentration in the headspace was tracked as it reflects the metabolic activity and can be correlated with distinct fermentation stages. The control culture without solvent addition showed the expected CO2 growth profile characterized by two phases of exponential CO2 evolution during exponential growth on glucose and ethanol and termination of CO2 formation upon carbon source depletion and cessation of cell growth. Cultivations containing ethyl decanoate showed a CO2 evolution like that of the control. However, after ethanol depletion at approx. 40 h, a further increase in the CO2 partial pressure was observed, indicating metabolic activity of S. cerevisiae on ethyl decanoate. Because of the yeast's native ability to produce this molecule, it can likely also degrade this substance (Saerens et al. 2010). Experiments with methyl decanoate showed a slower CO2 production than the control, possibly due to slower cell growth caused by metabolic stress. This observation was also underlined by the cell viability of samples treated with methyl decanoate, which was 20% lower than the control. Methyl decanoate was henceforth excluded as a solvent candidate. In contrast, IPM addition had no pronounced effect on cell metabolism. IPM is commonly used in cosmetic and skincare industry, and regarded as safe (Vadgama, Odaneth, and Lali 2015; Zhao et al. 2016; Zhang et al. 2017; Dib et al. 2021). The results obtained in this study highlight its biocompatibility on yeast cells, which is also in accordance with the previously reported application of IPM in yeast cultures (Sun, Kwak, and Jin 2019; Zhang et al. 2020; Yamada et al. 2022). Headspace CO2 volume concentration during the cultivation of S. cerevisiae BA6∆ in the presence of (A) 4 vol. % ethyl decanoate (blue), (B) 4 vol. % methyl decanoate (red), and (C) 4 vol. % IPM (green). For each approach, a control batch without solvent (black) was carried out. In approach containing ethyl decanoate (A), CO2 increased after 40 h, indicating ethyl decanoate being used as carbon source by the cells. In approach containing methyl decanoate (B), cell growth is slower than the control approach, showing the negative impact of methyl decanoate on cell metabolism. In approach with IPM (C), the CO2 course is similar to the control, indicating no negative impact of IPM on cell metabolism. Representative data sets of two biological replicates are shown. IPM Addition Post‐Batch Cultivation Allowed Triterpenoid Extraction of More Than 80% Since our initial screening results showed that extraction with IPM was a suitable trade‐off between our two main criteria, biocompatibility, and product extraction capability, we evaluated in situ product extraction with this solvent. Guided by experimentally determined solubilities of the two main triterpenoids betulinic acid and betulin in IPM, which were 1.3 ± 0.1 and 2.1 ± 0.1 g/L, respectively, we estimated that a solvent‐to‐culture ratio of 1.5:1 (150 vol.% IPM) would provide sufficient capacity for complete extraction of all triterpenoids produced in standard batch cultivation with 50 g/L glucose. We determined the distribution of triterpenoids in the solvent and pellet after different time points to investigate the influence of extraction duration. The fraction of extracted triterpenoids in IPM increased steadily with increasing incubation time, reaching more than 80% after 24 h (Figure 4A). Combining 8 mM β‐CD and 150 vol. % IPM significantly improved the extraction efficiency. While treatment with IPM alone released 60% of the product within 2 h (Figure 4A), 80% was extracted in the same period when 8 mM β‐CD was added in addition to IPM (Figure 4B). The positive effect of β‐CD was expected, as β‐CD was shown to extract cholesterols and phospholipids from the cell membrane (Zidovetzki and Levitan 2007). Due to the structural similarity of sterols and triterpenoids, it is likely that triterpenoids accumulated in the plasma membrane can also be extracted by and form complexes with CDs. However, adding 8 mM β‐CD alone did not significantly enhance product secretion from the cells (Figure S7). Therefore, we propose that β‐CD facilitates extraction into the organic phase by functioning as a shuttle. Distribution of intracellular and extracted triterpenoids in [%] after 2, 5, 18, and 24 h extraction with 150 vol. % IPM (A) and 150 vol. % IPM plus 8 mM β‐CD (B), showing that addition of β‐CD allows a faster extraction of triterpenoid from the cell into the organic phase. Two biological replicates per each time point were measured (shown by (I) and (II)). Error bars refer to two technical replicates. Lan and Xiao reported that 2‐hydroxypropyl‐β‐cyclodextrin (HP‐β‐CD) or 2,6‐dimethyl‐β‐cyclodextrin (DM‐β‐CD) enabled fast extraction of the triterpenoid ganoderic acid 3‐hydroxy‐lanosta‐8,24‐dien‐26 oic acid (GA‐HLDOA) from S. cerevisiae (Lan and Xiao 2020). Failure to achieve similar results with β‐CDs could be due to differences in CD structure and polarity or different properties of betulinic acid compared to GA‐HLDOA. Sequential Batch Cultures With Intermittent Extraction by IPM Increased Triterpenoid Production To explore the potential enhancement of triterpenoid production by extracting these intracellular molecules between sequential batch experiments, we subjected a stationary‐phase‐batch culture grown on glucose to a 24‐h treatment with IPM. After extraction, the entire biomass was aseptically transferred into fresh medium for a new batch run with ethanol as carbon source. We chose ethanol, because it was shown to promote triterpenoid synthesis in yeast (Czarnotta et al. 2017). Moreover, cell growth on ethanol is slower than on glucose, leading to reduced oxygen demand and decreases the risk of oxygen limitation in shake flask. The second batch run was followed by a new extraction step at the end of the batch cultivation. Six rounds of this sequence were performed (for schematic overview, see Figure S3). Sequential batch cultivations with alternating incubations without product extraction served as a control. We measured the total product amount produced at the end of each batch round. For IPM‐treated cultures, this value comprised the triterpenoid amount in the pellet and the organic phase; triterpenoid amounts in the aqueous phase were insignificant (data not shown) and were not included. In the control experiment, the triterpenoid quantity in the entire cultivation broth was measured. To compare the triterpenoid production in both approaches, we determined the sum of total triterpenoids produced at the end of each batch as well as the total amount over all batches (Table S1). The total triterpenoid amount produced over all batch experiments was significantly increased in the experiment with IPM extraction compared to the control (Figure 5A), supporting our hypothesis that intracellular product accumulation limits productivity. Sequential batch fermentations with and without in situ extraction. Shown are the sum of total triterpenoids (intracellular (pellet) and extracellular (OP)) which is higher in the in situ extraction approach (grey), than in the control approach (black) (A), cell viability, which was similar in both in situ extraction approach (grey) and control (black) (B), and product yield in both approaches, which was slightly higher in in situ extraction approach than in the control (C) over six sequential batch runs. For the in situ extraction approach, the product distribution between pellet and OP [%] was determined (D). Error bars show the standard deviation of three biological replicates. Nevertheless, we also observed that the production rate decreased with increasing number of batches, apparent from the flattening of the plotted total triterpenoid amounts for both approaches (Figure 5A). While triterpenoid production increased in the first two batch rounds, after the third batch, the triterpenoid amount produced per batch decreased independent of intracellular product accumulation (Table S1). This observation lets us argue that other factors may limit triterpenoid synthesis in addition to intracellular product accumulation. Impedance flow cytometry found that the cell viability in both cultivation setups was comparable (Figure 5B), highlighting that treatment with IPM does not affect cell viability, in line with the measurement of CO2 evolution (see section 0). The drop in cell viability observed after the fifth batch could not be explained and might be a technical error. In approaches with product extraction, we further observed differences in intra‐ and extracellular product distribution (Figure 5D), specifically, a decline in the extracted product fraction from 90% to 30%. Since fresh solvent (150 vol. %) was used for every extraction, saturation of the extractant can be excluded as a reason for the reduced extraction efficiency. Instead, we hypothesize that the repeated exposure to the organic solvent induces changes in the cell membrane rigidity, which could impede the extraction of the hydrophobic products. Overall, this experiment showed that product extraction enables higher total production of triterpenoids as well as higher product yield on substrate in at least three sequential batch rounds (Figure 5C) while also pointing to additional hurdles limiting the reusability and extended productivity of the whole‐cell biocatalyst. In Situ Extraction Is Applicable for Pulsed Fed‐Batch Bioreactor Cultivation Sequential batch cultivation with intermittent product extraction showed a positive influence of in situ extraction on triterpenoid production in S. cerevisiae . However, this alternating process is practically complex and involves a risk of contamination. To address these concerns, we tested an alternative process configuration in which in situ extraction was continuously performed during fed‐batch cultivation in bioreactors. The fed‐batch process involved an initial batch culture on glucose followed by pulses of approximately 25 g/L ethanol injected upon carbon source depletion, as indicated by a rapid rise in the dissolved oxygen (DO) tension. For in situ extraction, 150 vol. % IPM and 8 mM β‐CD were added to the culture at the beginning of the fed‐batch phase simultaneously with the first ethanol pulse, whereas the control was run without addition of the extractant. We compared the total amounts of triterpenoids, squalene, and ergosterol in both approaches (Figure 6A,B). Interestingly, no significant differences were observed in the final product titres between the control and the in situ extraction approach (Table 2). After about 80 h, product formation slowed down, similar to our observations in the sequential batch experiments, where the production plateaued independently of product extraction. Ethanol‐pulsed fed‐batch fermentation: Concentration profiles of triterpene and sterol products (top) and extracellular substrates and products as well as cell viability (bottom) for the control (A, C) and the fermentation with in situ extraction (B, D) shows no difference in product titre. The dashed line indicates the start of in situ extraction. (A and B) Concentrations of betulinic acid (BA), betulin (B), betulin aldehyde (BetAld), lupeol (Lup), ergosterol (Erg), and squalene (SQ) and total triterpenoid concentration (B, BetAld, BA). (C and D) Concentrations of glucose, ethanol, glycerol, acetate, and CDW and cell viability. Data from a single experiment are shown. TABLE 2Final concentrations of betulinic acid (BA), betulin (B), betulinic acid (BetAld), lupeol (Lup), ergosterol (Erg), squalene (SQ), sum of triterpenoids (Triterps), and total product concentration including SQ and Erg in ethanol‐pulsed and glucose‐pulsed fed‐batch cultivations without (control) and with in situ product extraction (in situ). In all approaches, similar final concentrations were measured.ExperimentBA [g/L]B [g/L]BetAld [g/L]Lup [g/L]SQ [g/L]Erg [g/L]Triterps [g/L]Total [g/L]Ethanol (control)1.6 ± 01.5 ± 01.6 ± 00.2 ± 02.1 ± 00.1 ± 04.9 ± 07.1 ± 1Ethanol (in situ)1.8 ± 01.4 ± 01.0 ± 00.1 ± 02.1 ± 00.7 ± 04.3 ± 07.1 ± 0Glucose (control)1.6 ± 01.7 ± 01.7 ± 00.3 ± 01.7 ± 00.1 ± 05.2 ± 17.1 ± 2Glucose (in situ)1.8 ± 02.0 ± 00.9 ± 00.070.7 ± 00.4 ± 04.7 ± 05.8 ± 0 Final concentrations of betulinic acid (BA), betulin (B), betulinic acid (BetAld), lupeol (Lup), ergosterol (Erg), squalene (SQ), sum of triterpenoids (Triterps), and total product concentration including SQ and Erg in ethanol‐pulsed and glucose‐pulsed fed‐batch cultivations without (control) and with in situ product extraction (in situ). In all approaches, similar final concentrations were measured. The pTEF promoter regulates the expression of the gene encoding cytochrome P450 monooxygenase. As it has been shown that this promoter is more active on glucose (Peng et al. 2015), we decided to perform the pulsed fed‐batch experiments with glucose, to evaluate whether the limited triterpenoid production can be attributed to decreased pTEF promoter activity during growth on ethanol. Hence, we tested the production performance in glucose‐pulsed fed‐batch experiments, with and without in situ extraction (Figure 7). The results revealed no extended production compared to the ethanol fed‐batch. Like the ethanol‐pulsed fed‐batch cultivations, a plateau in the production of the oxyfunctionalized compounds occurred after 80 h of cultivation (Figure 7A,B). Glucose‐pulsed fed‐batch fermentation: Concentration profiles of triterpene and sterol products (top) and extracellular substrates and products as well as cell viability (bottom) for the control (A, C) and the fermentation with in situ extraction (B, D) shows no difference in product titre. The dashed line indicates the start of in situ extraction. (A and B) Concentrations of betulinic acid (BA), betulin (B), betulin aldehyde (BetAld), lupeol (Lup), ergosterol (Erg), and squalene (SQ) and the total triterpenoid concentration. (C and D) Concentrations of glucose, ethanol, glycerol, acetate, and CDW, and cell viability. Data from a single experiment are shown. Glucose pulses were added when a rapid increase in the DO signal (Figure S8) indicated carbon source depletion. This time point is displayed in Figures C and D with an assumed glucose concentration of zero. In contrast, the glucose concentration after pulse was determined by HPLC‐RI. The final concentration and yields in both ethanol‐pulsed and glucose‐pulsed fed‐batch cultivations were in a similar range, regardless of the application of in situ extraction (Table 3). In all four approaches, there was an increase in ergosterol concentration towards the end of the fermentation. The ergosterol concentration in cultivations with in situ extraction was higher than in the control (see Tables 1 and 2). However, since the ergosterol content in the control also increased over the course of the fermentation, this phenomenon seems to be unrelated to in situ extraction but might be due to the extended exposure to ethanol (Aguilera et al. 2006; Girardi Piva et al. 2022) or a decreased flux towards triterpenoid synthesis promoting ergosterol synthesis. In all, these results underpinned the presence of other constraints on product synthesis in addition to the accumulation of intracellular products and transcriptional regulation. TABLE 3Final yields of betulinic acid (BA), betulin (B), betulinic acid (BetAld), lupeol (Lup), ergosterol (Erg), squalene (SQ), sum of triterpenoids (Triterps) and total product concentration including SQ and Erg in ethanol‐pulsed and glucose‐pulsed fed‐batch cultivations without (control) and with in situ product extraction (in situ). In all approaches, similar final yields were obtained.ExperimentBA [mg/gCDW]B [mg/gCDW]BetAld [mg/gCDW]Lup [mg/gCDW]SQ [mg/gCDW]Erg [mg/gCDW]Triterps [mg/gCDW]Total [mg/gCDW]Ethanol (control)59.9 ± 658.6 ± 562.4 ± 35.7 ± 082.2 ± 43.6 ± 0186.6 ± 20272.4 ± 3Ethanol (in situ)66.2 ± 149.1 ± 137.3 ± 12.1 ± 176.6 ± 524.9 ± 2154.7 ± 5256.1 ± 2Glucose (control)58.4 ± 1560.8 ± 1161.8 ± 1110.0 ± 363.0 ± 154.6 ± 1191.0 ± 57258.7 ± 8Glucose (in situ)66.1 ± 274.6 ± 231.5 ± 32.5726.7 ± 215.3 ± 1174.7 ± 9216.7 ± 2 Final yields of betulinic acid (BA), betulin (B), betulinic acid (BetAld), lupeol (Lup), ergosterol (Erg), squalene (SQ), sum of triterpenoids (Triterps) and total product concentration including SQ and Erg in ethanol‐pulsed and glucose‐pulsed fed‐batch cultivations without (control) and with in situ product extraction (in situ). In all approaches, similar final yields were obtained. In contrast to the sequential batch cultivation in shake flask, pulsed‐fed‐batch fermentation with continuous product extraction did not improve triterpenoid production. Saturation of the organic phase over the course of the fed‐batch cultivation could be excluded, as intracellular triterpenoids stayed at a low level throughout the fed‐batch fermentation (Figure 8). The cell viability data also exclude possible negative impacts of IPM exposure. Moreover, in the sequential batch cultivations, we found a decline in extraction efficiency after the third round of extraction, with increased intracellular product accumulation (Figure 5). In contrast, in fed‐batch cultivations, the amount of product in the pellet consistently remained under 20% throughout (Figure 8). Further investigations are necessary to discern the principal differences between the two cultivation modes and the factors contributing to varying extraction efficiencies. Notably, both cultivation setups exhibited a flattened production pattern of oxyfunctionalized triterpenoids, not only in the control but also in systems with in situ extraction. This underscores the existence of factors beyond intracellular product accumulation, which impede extended triterpenoid production. Distribution of the triterpenoids betulinic acid (BA), betulin (B), and betulin aldehyde (BetAld) in pellet and organic phase in ethanol‐pulsed (A) and glucose‐pulsed (B) fed‐batch with application of 150 vol. % IPM and 8 mM β‐CD for in situ extraction. After start of in situ extraction, indicated by the arrows, the triterpenoids were detected in the organic phase, showing the successful extraction of triterpenoids into the organic phase. Data from a single experiment are shown, error bars represent the standard deviation of two technical replicates. Triterpenoid Production Is Limited on Proteomic and Genetic Level Since in situ extraction did not completely lift triterpenoid production limitations in S. cerevisiae BA6∆, we applied quantitative proteomics to determine the expression changes of essential triterpenoid biosynthesis enzymes at the early, mid, and late stages of the cultivation. Statistical significance of the differential expression of proteins was assessed by the p‐value. A p‐value less than 0.5 was considered significant, while a p‐value exceeding 0.5 was deemed insignificant (not shown). To filter out differences that are not only statistically significant but also biologically relevant, an absolute average log2 (fold change) cutoff value of 1 was used (Schork et al. 2021). Based on these criteria, differentially expressed proteins were categorized as significantly upregulated or downregulated (absolute average log2 (fold change) > 1, p‐value < 0.05), and weak/no effect proteins (absolute average log2 (fold change) < 1, p‐value < 0.05). Table 4 shows the protein abundances for proteins involved in triterpenoid synthesis, mevalonate and FPP biosynthesis, and ergosterol biosynthesis. Table S2 shows the results for protein abundances of proteins involved in glyoxylate shunt, lipid biosynthesis, cytoskeleton, nuclear core complex, and histones. The protein abundances for maintenance proteins belonging to cytoskeleton, nuclear pore complex, and histones were checked as a quality control of our data evaluation, as their expression was expected to be constant during the entire cultivation (Table S2). Despite a few changes related to measurement errors, mostly weak or no changes for these maintenance proteins were observed, as expected. TABLE 4Changes in protein abundance between three sampling times in pulsed fed‐batch cultures.Protein nameEthanol‐pulsed fed‐batchGlucose‐pulsed fed‐batchAverage log2 (fold change)(I)(II)(III)(I)(II)(III)Triterpenoid synthesisOEWLupeol synthase−1.38−0.840.53n.s.n.s.−0.60CYPP450 monooxygenasen.s.−0.62−0.44−1.64−1.020.62CPRCytochrome P450 reductase−1.30−1.73−0.43n.s.−1.07n.s.Mevalonate and FPP biosynthesisErg10pacetyl‐CoA C‐acetyltransferase0.08−0.06−0.14n.s.−0.92−0.66Erg13p3‐hydroxy‐3‐methylglutaryl coenzyme A synthasen.s.n.s.−0.42−0.40−0.40n.s.Hmg1HMG‐CoA reductasen.s.−1.10n.s.n.s.n.s.0.58Hmg2HMG‐CoA reductasen.s.1.271.890.57n.s.n.s.Erg12pMevalonate kinase−0.51−0.140.37−0.87n.s.n.s.Erg8pPhosphomevalonate kinasen.s.n.s.n.s.n.s.−1.85−0.76Erg19pMevalonate phosphate decarboxylase1.221.27n.s.1.401.03−0.37Ergosterol biosynthesisErg9pSqualene synthase−1.27−1.25n.s.−0.75n.s.0.77Erg1pSqualene epoxidase−2.52−2.67−0.15−3.80−2.791.02Erg7pLanosterol synthasen.s.n.s.n.s.n.s.n.s.n.s.Erg26pSterol reductase−0.98−0.94n.s.−1.18−0.370.81Erg27p3‐keto sterol reductase−2.53−2.87n.s.n.s.n.s.n.s.Erg6pDelta 24‐sterol C‐methyltransferasen.s.−0.46−0.13−0.90−0.65n.s.Erg2pSterol isomerase−0.58−0.45n.s.n.s.n.s.n.s.Erg3pSterol desaturase−1.40−1.83−0.43−1.44−1.55−0.11Erg5pSterol desaturase−2.23−1.89n.s.−0.66n.s.n.s.Erg4pSterol reductasen.s.n.s.n.s.−1.99n.s.n.s. Note: Average log2 (fold change) for t mid/t start (I), t end/t start (II), and t end/t mid (III) in ethanol‐pulsed and glucose‐pulsed fed‐batch cultures. Ethanol‐pulsed fed‐batch: t start: 23 h, t mid_ 74 h, and t end: 147 h. Glucose‐pulsed fed‐batch and t start: 17 h, t mid: 57 h, and t end: 160 h show a decrease in abundance for triterpenoid synthesis proteins. Weak/no biological change; Upregulated; Downregulated.Abbreviation: n.s., not significant. Changes in protein abundance between three sampling times in pulsed fed‐batch cultures. Note: Average log2 (fold change) for t mid/t start (I), t end/t start (II), and t end/t mid (III) in ethanol‐pulsed and glucose‐pulsed fed‐batch cultures. Ethanol‐pulsed fed‐batch: t start: 23 h, t mid_ 74 h, and t end: 147 h. Glucose‐pulsed fed‐batch and t start: 17 h, t mid: 57 h, and t end: 160 h show a decrease in abundance for triterpenoid synthesis proteins. Weak/no biological change; Upregulated; Downregulated. Abbreviation: n.s., not significant. For the enzymes OEW, CYP, and CPR catalysing the conversion of 2,3‐oxidosqualene into the target triterpenoid betulinic acid (Figure 1), we mainly observed negative average log2 (fold change) values among the significant and biologically relevant results, indicating the loss of translation of these heterologous proteins in both ethanol‐pulsed and glucose‐pulsed cultures over the course of the fermentation. In addition to the triterpenoid synthesis proteins, also native enzymes implicated in the mevalonate and upper sterol synthesis pathway (Figure 1) were investigated since a potential downregulation could limit the supply of the precursor 2,3‐oxidosqualene and triterpenoid production. Most enzymes involved in mevalonate and FPP biosynthesis showed no clear trend, except for mevalonate phosphate decarboxylase (Erg19p), which was upregulated in both approaches. Low coverage of pathway enzymes with statistically significantly differential values prevented further conclusions about the proteome level of the mevalonate pathway. For enzymes involved in upper ergosterol biosynthesis, squalene synthase (Erg9p), downregulation on ethanol, and squalene epoxidase (Erg1p), downregulation was observed on ethanol and glucose. However, this was only partially consistent in all measurements. A similar trend was observed for enzymes involved in converting 2,3‐oxidosqualene to ergosterol. This was surprising, as we observed an increased ergosterol content in the cultivations (Figures 6 and 7). The mutated lanosterol synthase Erg7p, which replaced the native enzyme in the engineered S. cerevisiae, showed low abundances and no significant change as expected (Guo, Jacobsen, et al. 2022; Guo, Wang, et al. 2022). The reduced abundance of heterologously expressed proteins, as indicated by the proteomic data, underlines a possible genomic instability of the strain. The multiple integrations of the heterologous genes encoding for cytochrome P450 monooxygenase (CYP) and lupeol synthase (OEW) into the genome of S. cerevisiae BA6∆ had been realized by insertion into retrotransposon sites TY4 (Figure 9A), reported to lack stability in cultivation over several generations (Maury et al. 2016). Hence, we performed whole‐genome sequencing on cells at the start and end of the cultivation to track a possible loss of heterologous genes. Although TY4 sites are distributed over nearly all chromosomes of S. cerevisiae (Maury et al. 2016), our results showed the presence of the genes of interest only on chromosome X. In all samples, the genes of interest were found to be located between NCA3 (193,860–194,873 bp) and ASF1 (196,287–197,126 bp; Figure 9). Since the sequencing data contained gaps, we cannot exclude the presence of additional gene copies in these uncharted regions. In the ethanol‐pulsed fed‐batch approach, one of the two copies of the expression cassette consisting of CYP and OEW together with the bidirectional pTEF1‐pPGK1 and the terminators tADH1 and tCYC1 (Figure 9B) found at the beginning of the cultivation (Figure 9B) was lost at the end of fermentation (Figure 9C). In contrast, genome sequencing found one copy of the genes OEW and CYP each at the beginning and the end of the glucose‐pulsed fed‐batch cultivation (Figure 9D,E). However, additional copies of the terminators tADH1 and/or tCYC1 indicated a loss of previously integrated expression cassettes. Schematic illustration of the gene cassette encoding cytochrome P450 monooxygenase (CYP) and lupeol synthase (OEW), with the respective promoters pTEF1 and pPGK1 and terminators tADH1 and tCYC1(A), the marker gene URA3 and the promoter pURA3 with the loxP sites, the TY sites TY4/5 and TY4/3. Annotation of genetic components integrated into chromosome X of S. cerevisiae BA6∆ at the beginning (B), and at the end (C) of ethanol‐pulsed fed‐batch cultivation as well as at the beginning (D) and end (E) of glucose‐pulsed fed‐batch based on results of the genome sequencing. In all approaches, only one or parts of the plasmid were found. The declining enzyme levels during fermentation can hence be attributed to inefficient multicopy gene integration. This insight provides an opportunity for creating strains with stabilized gene integration and protein expression. Potential gene loss through homologous recombination between TY sites (Qi et al. 2023) can be mitigated by employing more stable integration methods (Jessop‐Fabre et al. 2016) and the avoidance of reusing identical promoters and terminators to minimize the occurrence of homologous regions, for example artificial, sequence‐diverse promoters (Kotopka and Smolke 2020). Conclusion In this study, we addressed the challenge of intracellular triterpenoid accumulation in engineered S. cerevisiae by introducing in situ product extraction through the synergistic application of IPM and β‐CDs. Our proof‐of‐concept study demonstrated the potential for extracting intracellular triterpenoids from living cells into a biocompatible organic phase, simplifying downstream processing and allowing for extended use of the same biocatalyst. Through two distinct cultivation methods, a sequential batch approach in shake flasks and a pulsed fed‐batch approach in a bioreactor, we achieved up to 90% product extraction into the organic solvent IPM. Notably, we observed variances in extraction efficiencies between these cultivation modes, emphasizing the need for continued research to optimize this method. However, a fundamental limitation of IPM lies in its low triterpenoid solubility, necessitating the use of substantial volumes of the organic phase. Indeed, this study employed a solvent‐to‐culture ratio of 1.5–1, presenting a significant challenge for translating this method to large‐scale applications. Thus, future research should be directed towards the exploration of more efficient solvents or the design of co‐solvent systems that can address the solubility issue. Moreover, innovative techniques for separating the product from the solvent and enabling continuous reuse should be pursued. We also uncovered unexpected limitations related to unstable gene integration, resulting in decreased enzyme levels and production instabilities that metabolic engineering needs to address. In summary, our research highlights the promise of in situ extraction for enhancing microbial production of hydrophobic, intracellular products. However, for this specific case, fully realizing the benefits of in situ product extraction depends on the sustained activity of the triterpenoid synthesis pathway through metabolic engineering. This work encourages further exploration of in situ extraction strategies and provides valuable insights for those seeking to utilize this innovative approach in bioprocess engineering. Author Contributions Mariam Dianat: writing – original draft, investigation, methodology, project administration, visualization, writing – review and editing, conceptualization, formal analysis, data curation. Sarah Straaten: data curation. Aldo Maritato: data curation. Daniel Wibberg: software, data curation. Tobias Busche: software, data curation. Lars M. Blank: supervision, funding acquisition, writing – review and editing, resources, project administration, conceptualization, investigation. Birgitta E. Ebert: supervision, writing – review and editing, software, conceptualization, investigation, methodology, project administration. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Data S1.