Eliminating viscosity challenges in continuous cultivation of yeast producing a GLP-1 like peptide | Microbial Cell Factories | Full Text

Microbial Cell Factories volume 24, Article number: 130 (2025) Cite this article

422 Accesses

Metrics details

The emergence of GLP-1s for the treatment of diabetes, obesity and other diseases has led to increased focus on finding efficient ways to produce the peptides in sufficient amounts to satisfy the ever-increasing demand. Although the use of microbial hosts constitutes the cheapest, easiest and safest way to produce these peptides in high volumes, process challenges still exist that reduce the production capacity. One of the main production challenges is the high viscosity of cultivation broths, which reduces the mass and oxygen transfer, thereby creating substrate and oxygen gradients that potentially lead to unwanted secondary metabolism and eventually compromises capacity.

The methodology used to identify the underlying factors of highly viscous broths during the recombinant production of GLP-1 precursors in S. cerevisiae in continuous cultivation is presented. Two root causes leading to highly viscous broths were uncovered and solutions identified. The first one is found in the soluble fraction of the broth and relates to the aggregation of GLP-1 precursor molecules that leads to highly viscous, shear thinning cultivation broths. The cultivation conditions under which the aggregation occurs and the consequences for both cultivation and product recovery are discussed. The second source of viscosity is found in the insoluble fraction of the cultivation broth and relates to cell aggregation due to Amn1p dependent incomplete separation of mother and daughter cells. This type of cell aggregation causes formation of cell clumps and leads to high viscosity cultivation broths with mild shear thickening properties.

To eliminate the GLP-1 peptide related viscosity, a new generation of yeast host strains that tolerates cultivation at increased pH values, above those that cause GLP-1 precursor aggregation, were utilized. In the case of the cell derived viscosity, yeast strains carrying either a deletion of the AMN1 gene or integration of the non-clumping AMN1D368V gene variant were employed. The implementation of these changes led to a scalable cultivation process characterized by a significant improved oxygen mass transfer attributed to the low viscosity and Newtonian behaviour of the cultivation broth.

The global burden of metabolic disorders, particularly type 2 diabetes and obesity has reached significant levels necessitating the development of effective therapeutic strategies. In this context, glucagon-like peptide-1 receptor agonists (GLP-1RAs) have emerged as a revolutionary class of pharmacological agents, offering significant benefits in the management of these conditions. GLP-1RAs, including semaglutide, tirzepatide, and liraglutide, have gathered considerable attention due to their favorable pharmacological properties, such as enhanced glucose-dependent insulin secretion, weight reduction and cardiovascular benefits [1]. GLP-1RAs are also currently being investigated for other indications- including Alzheimer and Parkinson disease [2, 3].

Despite the success of GLP-1RAs, challenges remain in optimizing their production to keep pace with an ever-increasing demand for these therapeutic agents. Efforts are also ongoing to develop efficient delivery systems, such as oral formulations to improve patients’ comfort and safety. However, the low bioavailability of the oral formulation exacerbates the need for production of even larger amounts of GLP-1RAs to cover the lower bioavailability of the oral formulations. The increased demand may be met by the construction of additional or larger production facilities. However, this approach on its own is costly and non-sustainable as more resources are used and the carbon footprint is increased. A more prudent approach is to increase capacity through optimizing the production processes and making them less resource intensive. Currently, GLP1-RAs are produced synthetically or in a semi-recombinant fashion where a precursor or part of the molecule is produced in microbial hosts such as Escherichia coli (E.coli), typically relying on intracellular accumulation of the product in inclusion bodies, and Saccharomyces cerevisiae (S. cerevisiae) that supports secretion of a soluble product out of the cell. Processed precursors are converted into the active pharmaceutical GLP-1RAs by chemical modification such as acylation for prolonged pharmacokinetics or addition of non-natural amino acids for improved peptide stability [4].

To facilitate efficient production of peptides like insulin- and GLP-1RA- precursors in S. cerevisiae the peptides are expressed as fusion proteins, N-terminally linked to a yeast MFalpha1 pre-pro sequence or a sequence with similar function. This sequence directs the protein into and through the secretory pathway of the yeast cell. Processing of the fusion protein is accomplished in vivo by first the signal peptidase, that removes the pre-peptide and later by the endogenous yeast Kex2p endopeptidase, which cleaves off the pro-peptide, resulting in secretion of correctly processed peptide precursor into the culture broth (see [5]).

Cultivation conditions and performance are critical parameters for efficient recombinant production of peptides and proteins in yeast. Both the cultivation process and primary recovery of secreted peptide precursor from the cultivation fluid can be hampered by high broth viscosities, which lead to reduced yields, increased cost of operation, and an overall compromised capacity.

Cultivation broths are very complex matrices consisting of a huge array of components in soluble, colloidal or suspended state. Each type of particle can be of different shape, size, structure and concentration. The complexity of the cultivation broth matrix arises from the interactions between these individual components and affects the rheological behaviour of the fluid. Unfavorable rheological properties can dramatically influence the efficiency of bioreactor operations, as they may negatively impact various factors such as gas–liquid mass transfer, heat transfer, distribution of feeds, local concentration of substrate and oxygen, power consumption for agitation, the efficient dissipation of metabolic and mechanically induced heat, and thus affecting the cooling requirements.

According to Buckland [6] the oxygen transfer coefficient (KLa) decreases in proportion with the square root of the broth viscosity (Eq. 1). Hindrance in oxygen transfer leads to low levels of dissolved oxygen available for the microorganism. The consequence of this oxygen limitation is a shift towards fermentative metabolism, leading to the production of undesirable by-products and lower energy generation, ultimately culminating in a substantial decrease in overall yield.

Another negative effect of highly viscous broths could be the often-seen increased gas hold-up. Highly viscous broths lead to an increase in the drag force on rising bubbles, causing them to rise more slowly compared to bubbles in low-viscosity fluids. The slower rise velocity can increase the residence time of bubbles in the liquid, potentially increasing gas hold-up. High gas hold-up can have multiple consequences. While it can contribute to increased oxygen transfer due to higher residence time (however, often counterbalanced by the larger bubble size in high viscosity broths), it also increases the apparent volume of the broth, thereby reducing the effective fermentation capacity. Elevated viscosity can also help stabilize the foam by increasing the viscosity of the thin liquid film (lamella) separating the bubbles [7].

The most popular broth components increasing the viscosity described in the literature are (a) filamentous growing cells where entanglement of the hyphae can result in highly viscous suspensions [8, 9]; (b) polymers such as polysaccharides [9] and nucleic acids [10, 11] and; finally (c) smaller molecules such as sugars in high concentrations.

In the case of polymer production, the broths become highly viscous and of pseudoplastic nature.Footnote 1 This happens due to the intermolecular entanglement of polymer chains in a solution. At rest or low shear rates, the polymer chains have time to find their lowest-energy structure, where polymer chains form intermolecular entanglements, which makes it more difficult for them to move past each other. This behaviour results in a higher resistance to flow and thus a higher viscosity. The degree of entanglement between molecules can be influenced by factors such as molecular weight, degree of branching, concentration, and temperature. At high shear rates, polymers are entirely disentangled or have enough kinetic energy to disentangle and adopt a state of higher conformational energy state, where the stretched chains slide over each other and show lower viscosity and NewtonianFootnote 2 behaviour [12,13,14].

Similarly, hyphal growth of filamentous microorganisms results in high viscosity pseudoplastic broths due to the hyphae being entangled [8]. High density fermentation broths of filamentous fungi can reach flow behaviour index (n)Footnote 3 values as low as 0.2–0.4 [15, 16]. In case of pelleted growth of filamentous organisms, the viscosity is significantly lower due to increased compactness [17]. That is, pellets have a lower surface area-to-volume ratio compared to filaments, which means there is less interaction with the surrounding components, compared to the extensive network of filaments, leading to lower resistance to flow.

In this work, the rheological characteristics of cultivations broths of GLP-1 peptide precursor expressing S. cerevisiae strains were evaluated. The root causes of the observed high broth viscosity were identified and effectively addressed through the implementation of targeted genetic modifications in the host organism and the optimization of cultivation conditions.

Restriction enzymes and Phusion HF DNA polymerase used for polymerase chain reaction (PCR) were purchased from Thermo Fisher Scientific (Rødovre, Denmark). Single stranded DNA oligonucleotides used as primers in PCR reactions were purchased from Eurofins Genomics (Ebersberg, Germany).

Geneticin™, nourseothricin and Zeocin™ selection antibiotic was obtained from Thermo Fisher Scientific (Rødovre, Denmark).

Plasmids for heterologous expression of extended GLP-1 peptide precursor and extended GIP-GLP1 peptide precursor in yeast are S. cerevisiae-E. coli shuttle vectors that contain a DNA sequence encoding the S. cerevisiae MFalpha pre-pro sequence (to direct the peptides through the yeast secretory pathway) fused to an extended peptide precursor inserted between the TDH3 promoter and TPI1 terminator of S. cerevisiae. Plasmids harbor a POT selection marker to allow selection of plasmid transformants in yeast strains that carry a disruption of the endogenous TPI1 gene on glucose containing media [18].

Plasmid pSUJ2894 is used for genomic integration of a non-clumping allele, AMN1D368V, and was constructed by DNA synthesis (GeneArt, Life Technologies, Regensburg, Germany). The plasmid contains an integration cassette flanked by unique SapI restriction sites consisting of 80 bp of DNA sequence upstream of the AMN1 coding DNA sequence (CDS), the AMN1 nucleotide sequence with GAT to GTT mutation at amino acid position 368 (D368V), a lox71-TEFagprom-KanMX-TEFag term-lox66 selection marker and 70 bp of DNA sequence downstream of the AMN1 CDS. The upstream and downstream AMN1 sequences enable integration of the AMN1D368V DNA sequence at the endogenous AMN1 gene by homologous recombination.

The GLP-1 precursor is an N-terminal extended GLP-1 peptide with a molecular weight of approximately 4.8 kDa. The amino acid sequence of the GLP-1 peptide is: Glu–Gly–Thr–Phe–Thr–Ser–Asp–Val–Ser–Ser–Tyr–Leu–Glu–Gly–Gln–Ala–Ala–Lys–Glu–Phe–Ile–Ala–Trp–Leu–Val–Arg–Gly–Arg–Gly.

Deletion of FLO8 (YER109 C locus; Gene ID: 856845), FLO11 (YIR019 C locus; Gene ID: 854,836) and AMN1 (YBR158 W locus, Gene ID: 852,455) genes in yeast strains was achieved by homologous recombination using conventional molecular biology and yeast genetic methods [19]. Yeast transformation was performed using the LiAc method described by Gietz et al. [20]. Genomic DNA was extracted from yeast for PCR verification using the method described by Looke et al. [21].

Deletion of FLO8 in yNN2104 (host lineage B; see Table 1) to generate yNN2912 was performed by transforming yNN2104 with a PCR generated DNA fragment comprising the KanMX selection marker (lox71-TEF1prom-KanMX-TEF1 term-lox66) flanked with 40 bp sequence upstream and downstream of FLO8 CDS. Transformants were isolated on YPD media supplemented with 200µg/mL G418 and re-streaked twice on selective media. Transformants carrying deletion of FLO8 were identified by PCR on genomic DNA using primer pairs flanking the FLO8 CDS (5′ GTAAGTCACTGAGGCTAT 3′/5′ AGGAATCGAATGCAACCG 3′ and 5′ GTATTTCGTCTCGCTCAGG 3′/5′ GCGAAATGTCAGATACGTA 3′). yNN2912 is a representative flo8 deletion strain in the yNN2104 strain background.

Deletion of FLO11 in yNN2104 (host lineage B) to generate yNN2915 was performed by transforming yNN2104 with a PCR generated DNA fragment comprising the KanMX selection marker (lox71-TEF1prom-KanMX-TEF1 term-lox66) flanked with 40 bp sequence upstream and downstream of FLO11 CDS. Transformants isolated on YPD media supplemented with 200µg/mLG418 were re-streaked twice on selective media. Transformants carrying deletion of FLO11 were identified by PCR on genomic DNA and FLO11 CDS flanking primer pairs (5′ CCTATTCATCAGTTATACTCCCT 3′/5′ AGGAATCGAATGCAACCG 3′ and 5′ GTATTTCGTCTCGCTCAGG 3′/5′ AGTAAAGAGGGTATCGAAATA 3′). yNN2915 is a representative flo11 deletion strain in the yNN2104 strain background.

Deletion of AMN1 in yNN2104 and yNN2912 (host lineage B) was performed by transformation with a PCR generated DNA fragment comprising the NatMX selection marker (lox71-TEF1prom-NatMX-TEF1 term-lox66) flanked with 40 bp sequence upstream and downstream of AMN1 CDS. Transformants isolated on YPD media supplemented with 100 µg/mL nourseothricin were re-streaked twice on selective media. Deletion of AMN1 was verified by genomic DNA PCR using AMN1 CDS flanking primer pairs (5′ TTGGTTTAATATCCCTTTTTGGTT 3′/5′ ATGTCCTCGACGGTCAG 3′ and 5′ AGGTCACCAACGTCAAC 3′/5′ CAGAATAGTAAAATGCTAGTTAAA 3′). yNN2993 and yNN2998 are representative amn1 deletion strains in yNN2104 and yNN2912 backgrounds, respectively.

To generate an AMN1D368V strain, yNN2993 was transformed with the AMN1D365V integration fragment obtained by SapI digestion of pSUJ2894 and transformants isolated on YPD media supplemented with 200 µg/mL G418. Transformants were re-streaked twice on YPD + 200 µg/mLG418 agar plates for purity. Genomic DNA was extracted from transformants and PCR with primer pairs (5′ TTGGTTTAATATCCCTTTTTGGTT 3′/5′ AGGAATCGAATGCAACCG 3′, 5′GTATTTCGTCTCGCTCAGG 3′/5′ CAGAATAGTAAAATGCTAGTTAAA 3′ and 5′ TTGGTTTAATATCCCTTTTTGGTT 3′/5′ CAGAATAGTAAAATGCTAGTTAAA 3′) performed to verify replacement of the amn1::lox71-TEF1prom-NatMX-TEF1 term-lox66 cassette with the AMN1D365V::lox71-TEF1prom-KanMX-TEF1 term-lox66 cassette. DNA sequencing of genomic DNA isolated from transformants was performed to confirm the presence of the AMN1D368V allele sequence at the endogenous AMN1 locus. A representative AMN1D368V strain, yNN3073 (host lineage B) was selected for continuous cultivation.

The bioreactors employed for all experiments were Sartorius BIOSTAT B plus (Sartorius AG, Goettingen, Germany), each featuring a 2 L vessel. The working volume for the experiments was set to 1.2 L. The vessel had a diameter of 130 mm, yielding an aspect ratio of 2.2:1. The stirrer shaft was equipped with two 6-blade Rushton impellers, each with a diameter of 60 mm and blades measuring 20 mm by 16 mm. The impellers were spaced 45 mm apart, with the lower impeller positioned 70 mm from the bottom of the bioreactor. The setup also included four baffles, each measuring 220 mm in length and 125 mm in width. Fermentations were performed in continuous mode where cells initially were cultivated in batch mode followed by a fed-batch phase triggered by the decrease of the carbon dioxide signal in the off-gas measurement when glucose was depleted during the batch phase. During the fed-batch phase, glucose (50% w/v), was fed into the bioreactor using an exponential feed profile (0.08 h−1). Maintenance metabolism was approximately accounted for in the calculation of the exponential feed by utilizing empirical data on the net cell yield on the substrate (YX/S) across various feed rates. When the desired biomass was reached, the fed-batch phase was followed by a chemostat or "continuous phase", in which “full medium” (i.e., carbon source, salts, vitamins, and trace metals) was fed into the bioreactor. The dilution rate of the chemostat phase was set at 0.08 h−1 and the culture volume controlled via a gravimetric feed. Moreover, base (NH4OH, 10%) was added into the bioreactor as needed to maintain a constant pH but also to provide a non-limiting nitrogen source throughout the process. The gas flow rate was maintained at 2.5 vvm using pure air with an oxygen concentration of 21%. Initially, the stirring rate was set at 1000 rpm. Upon reaching the maximum biomass level in continuous phase, the stirring rate was increased to 1300 rpm.

The pH setpoint was maintained at 5.3 for the yNN2071 strain (host lineage A) and, unless otherwise stated, a pH setpoint from 6.2 to 6.5 was used for the strains of the new host cell lineage B (see Table 1).

The exhaust gas from the cultivation processes was analyzed online using a Thermo Scientific™ Prima BT Mass Spectrometer to calculate the oxygen uptake, carbon dioxide evolution rate, and ethanol emission. The respiratory quotient (RQ), defined as the ratio of carbon dioxide evolution rate to oxygen uptake rate, along with ethanol emission, acted as indicators of fermentative metabolism, which may result from either oxygen deficiency or overflow metabolism.

Rheological characterisation was performed on broth samples withdrawn from the bioreactor (~ 5 ml). The samples were left at room temperature for 20 min and then stirred gently by hand to ensure homogeneity, unless otherwise specified. The viscosity of the samples was measured using a microfluidic rheometer (microVisc™, Rheosense) equipped with a 50µm path chip. The microVisc measured the pressure drop along an array of sensors when a liquid passes through the cell. The slope (and corresponding R2 of the fit) was calculated from the pressure drop as a function of distance (essentially shear stress versus shear rate) and used to calculate the viscosity. The measurements were performed at a setpoint temperature of 22 °C and at different shear rates (4000, 2000, 1000, 500, 250, 150, 80 s−1 where applicable). Using an in-house CFD application, the range of the shear rate values was selected to represent those typically encountered in bioreactors, ranging from small to industrial scale under various conditions. This selection also included shear gradients from low to high within individual bioreactors. Although the estimated shear rate values range from 50 to 5000 s⁻1, the nature of the sample and equipment limitations may prevent viscosity measurements at certain shear rates.

When specific shear rates are presented against cultivation time, a shear rate of 2000 s⁻1 is used to approximate the average shear rate of the bulk broth in the BIOSTAT B plus setup utilized in this study, whereas a shear rate of 250 s⁻1 approximates the average shear rate of the bulk broth in a pilot-scale bioreactor.

For the rheological analysis of the supernatant and the resuspended pellet, the broth sample was centrifuged at a relative centrifugal force of 2250 × g for 10 min to reach pellet-liquid-separation. The supernatant was recovered and the viscosity of the supernatant at different shear rates was measured as described above. The remaining pellet was washed twice with de-ionized water and brought to the initial weight. The viscosity of the so-obtained re-suspended pellet in water was measured at different shear rates as described above.

The power law model was used to calculate the flow consistency index (K) and flow behaviour index (n) of a fluid (Eq. 2). To implement the power law model, the shear stress (τ) versus shear rate (γ) data was plotted on a log–log scale, and the slope and intercept of the linear portion of the resulting curve were used to determine n and K, respectively. The average flow behaviour index (n) values were determined by integrating the flow behaviour index (n) over the respective time intervals using the composite trapezoidal rule (trapz function from NumPy v1.26) and then dividing it by the duration of those intervals.

Dilution method: 0.1 mL of culture was added to 2.7 mL de-ionized water and mixed thoroughly prior to centrifugation. 1 mL of the supernatant was filtered using a 0.2 μm PVDF filter into an HPLC vial. The dilution method was used to extract the non-aggregated target peptide from the cultivation broth after which the concentration of target peptide was determined by UPLC.

Alkalic extraction: 1 mL of culture was added to 4 mL 0.075M phosphate buffer pH 12 and mixed thoroughly before centrifugation. Supernatant was weighed, diluted 1:1 with 4 N acetic acid and sterile filtered using a 0.2 μm PVDF filter into an HPLC vial. The alkalic extraction method was used to extract the total target peptide (both monomeric and aggregated) from the cultivation broth after which the concentration of total target peptide was determined by UPLC.

UPLC-MS analysis: The GLP-1 precursor was separated and analysed using a reverse phase UPLC-MS (Agilent Technologies 6200 (ESI-TOF LC–MS) on a Poroshell 120 SB-C8 2.1 × 50 mm column (Agilent), operating at 40 °C and at a flow rate of 0.5 mL/min. Buffer A was composed of 8.8 mM ammonium Formate + 0.1% formic acid in milliQ water, and buffer B was composed of 0.1% formic acid in acetonitrile, with an elution gradient of 26–43% B between 0.21 and 4 min, 80% B between 4.01 and 5 min and then return to the initial conditions of 26% B at 5.01 min. The UV-detection was performed at 214 nm, and the concentration was determined using a GLP-1 reference. The correct mass of the GLP-1 precursor was confirmed by MS analysis. Values were normalised to the peptide concentration of a high yielding standard process set at 100%.

2 mL of the cell culture was pipetted into a pre-weighed centrifuge tube, centrifuged at 4400 rpm for 5 min after which the supernatant was discarded by decanting. The cell pellet was subsequently washed in 2 mL saline to rid the sample of medium components and centrifuged, after which the wash was discarded. The wet biomass sample was dried at 105 °C for at least 24 h (h) and successively cooled down in a desiccator for 1 h before weighing. Values were normalised to the dry cell weight of a high yielding standard process set at 100%.

An S. cerevisiae strain (yNN2071) expressing a GLP-1 precursor peptide that is directed through the secretory pathway into the cultivation broth was cultivated in a bioreactor at a pH setpoint of 5.3, according to the method described in the Method section. When steady state was reached after the onset of the continuous phase, samples were removed to determine the GLP-1 precursor concentration and to characterize the rheological properties of the broth. Two methods were utilized to extract the GLP-1 precursor from the cultivation broth prior to peptide quantification by UPLC. The dilution method extracts the native, non-aggregated GLP-1 from the cultivation broth, whereas the alkalic extraction method employs a high pH strength buffer to de-aggregate GLP-1 aggregates and thereby extracts the total amount of GLP-1 precursor (both monomeric and aggregated) from the cultivation broth. As shown in Fig. 1A, the GLP-1 precursor was not fully recovered from the culture broth when the dilution method was applied for extraction and the recovery of precursor decreased over cultivation time suggesting that the target peptide has a strong tendency to aggregate. It has been described in the literature that GLP-1 peptides tend to aggregate and fibrillate in acidic and near neutral pH environments, while by increasing the pH the phenomenon is reversed [22, 23]. A rheogram of the collected yNN2071 cultivation samples is shown in Fig. 1B, where the viscosity at various shear rates was analysed for the intact broth, the supernatant (soluble fraction), and the twice washed pellet (solid fraction) resuspended in water. The results clearly show that the high viscosity in the cultivation broth originates from the soluble fraction (supernatant) and not the solid fraction (pellet). Moreover, the viscosity was reduced with increasing shear rates pointing towards a pseudoplastic nature of viscosity. In addition to pseudoplasticity, the broth sample also exhibited a time-dependent shear thinning or thixotropic behaviour as prolonged shear exposure by vortexing reduced the viscosity (Fig. 1C). Note the logarithmic scale used for the viscosity axis in Fig. 1C. This was implemented to effectively display the wide range of viscosity values (very high at low shear rate, low at high shear rate) for the undisturbed sample, as compared to the vortexed sample.

Results from a pH 5.3 cultivation of the yNN2071 strain. A Cultivation profiles of GLP-1 precursor concentration as measured by a) the dilution (●) and b) alkalic extraction method (○) to de-aggregate the precursor. B Example of rheological characterisation of a broth sample where viscosity at different shear rates of the fermentation broth , fermentation supernatant and re-suspended washed pellet were analysed. C The viscosity of a broth sample before and after vortexing for 2 min (◆). D Correlation between the relative concentration of the aggregated peptide (calculated by subtracting the relative concentration derived by the dilution method from the relative concentration derived by the alkalic method) and broth flow behaviour index (n). The dotted line represents the best fit linear regression line

The concentration of aggregated/fibrillated GLP-1 precursor was calculated indirectly as the difference in precursor concentration obtained from the alkalic method and the dilution method (Calkalic–Cdilution). When the relative concentration of aggregated/fibrillated precursor is plotted against the flow behavioural index (n), a clear negative linear correlation is observed showing that an increase in aggregated GLP-1 precursor leads to a higher pseudoplasticity (and thus viscosity) of the cultivation broth (Fig. 1D). This strongly indicates that the aggregated GLP-1 precursor is the actual cause behind the high viscosity observed. As the cause of the elevated viscosity and pseudoplasticity is present in the supernatant and correlates to the aggregated precursor concentration it can be inferred that the aggregated precursor resides in the soluble fraction of the cultivation broth.

This type of viscous and pseudoplastic fluid is of particular concern as scaling-up becomes challenging due to the exacerbated gradients in the larger-scale bioreactors. Beyond a certain distance from the impeller, the shear rate is very low and, as a result the viscosity of a highly pseudoplastic fluid will be very high compared to areas close to the tip of the impeller where the viscosity is minimal (due to the very high shear rate). Therefore, the viscosity differences within the bioreactor are expected to increase with the size of the bioreactor, leading to uneven distribution of nutrients and oxygen, which can have a negative impact on cells, as they are faced with different micro-environments with varying levels of nutrients and oxygen. This may result in cells going into an overflow metabolism if exposed to high levels of nutrients and cells going into a starvation mode if nutrient levels are low. In both cases, the growth and productivity of the cells is negatively affected. According to Gernaey [24], a factor contributing to poor scalability in industrial fermentation processes is the lack of accurate estimation of shear rates across different scales, which leads to incorrect predictions of viscosity in viscous non-Newtonian broths.

To confirm the relationship between viscous pseudoplastic broths and GLP-1 peptide aggregation, broth of yNN2071 cultivated at pH 5.3 was collected and subjected to gradual pH increase prior to sample analysis. As shown in Fig. 2A increasing the pH of the cultivation broth resulted in a higher recovery and concentration of the GLP-1 precursor, a reduced flow consistency index (K) and increased flow behaviour index (n) (i.e. less pseudoplastic and more Newtonian behaviour). Figure 2A also shows that a broth pH of around 8 is sufficient to de-aggregate the GLP-1 precursor, as the measured concentration reaches its maximum and plateaus at this pH value. This observation supports the hypothesis that the GLP-1 precursor aggregation is responsible for the observed high broth viscosity. This finding, along with the fact that the peptide, when aggregated, remains in the soluble fraction of the cultivation broth, has product recovery implications. Namely, for a low cultivation pH, there is no need to adjust the broth pH prior to clarification. Instead, the broth can be clarified at the low cultivation pH, and the resulting supernatant adjusted to higher pH to de-aggregate the product.

A Correlation between cultivation broth pH and relative GLP-1 precursor concentration (●), viscosity at 2000 s−1, flow behaviour index (n) , flow consistency index (K) . B GLP-1 precursor concentration with and without alkalic treatment of fermentation broth supernatants clarified at the respective cultivation pH for the lineage host A strain yNN2071 (broth pH 5.3) and for lineage host B strains yNN2101 and yNN2104 (broth pH 6.2)

To further investigate the effect of pH on the recovery of GLP-1 precursor from cultivation broth three GLP-1 precursor expression strains derived from two different host strains (A and B; see Table 1) were cultivated at different pH and their respective fermentation broths analysed. Figure 2B shows the GLP-1 precursor concentration of the supernatant clarified at cultivation pH from fermentation broths of yNN2071 (host strain A) cultivated at pH 5.3 and of yNN2101 and yNN2104 (host strain B) cultivated at pH 6.2. The native GLP-1 peptide concentration, obtained by the dilution method, is significantly higher for the supernatants from the broths obtained at cultivation pH 6.2 compared to the broth obtained at cultivation pH 5.3. In the case of the low cultivation pH broth, when the pH of the supernatant is increased (alkalic treatment) the GLP-1 peptide is immediately recovered. For the other two fermentation broths cultivated at pH 6.2, no significant gain in GLP-1 precursor concentration was achieved upon increase of the supernatant pH. This confirms that the aggregated GLP-1 peptide remains soluble and shows the possibility of de-aggregating the GLP-1 peptide by applying alkalic treatment in the supernatant instead of the fermentation broth. Such an approach would avoid cell lysis and the concomitant increased release of impurities such as host cell proteins leading to a leaner primary recovery process.

Rekhi et al. [25] reported that protein aggregation could lead to liquid–liquid phase separation with droplet formation, which increases the viscosity of the fluid, the magnitude of which depends on the interactions between the diverse residue pairs. In the current work, no visible signs of liquid–liquid phase separation was observed by optical microscopy or turbidity measurements. Presumably, the GLP-1 peptide forms soluble aggregates through weak, non-covalent interactions that do not cause phase separation but increase the liquid viscosity via the intermolecular entanglement and formation network-like structures, which can hinder the flow of the liquid.

It is worth noting that a viscosity derived from product aggregation does not exclude the possibility that other molecules in the cultivation broth may play an important role in triggering and/or contributing to the phenomenon. Zapatka et al. [22] suggested multiple steps and more than one pathway towards the formation of insoluble aggregates/fibrils in the case of the human GLP-1 molecule. Moreover, there has also been evidence of an"off-pathway"oligomeric species that compete with the fibrillation process [26]. In the current work, we observed that supernatants centrifuged at high G forces containing GLP-1 precursor peptides can jellify even at higher pH ranges in the presence of the broth pellet, while in the absence of it, they exhibit prolonged stability. This potentially shows that in our case, the soluble aggregated GLP-1 precursor peptides causing increased broth viscosity are rather a precursor of fibrillation than an “off-pathway”. Srivastava et al., (2019) [27] summarized causes/triggers of β-Amyloid aggregation and they range from lipids surfaces interactions, acidic pH, presence of some bivalent and trivalent metal ions. Therefore, the gelation we observed may be explained by interactions of the GLP-1 precursor peptide with membrane components such as ergosterol which can induce nucleation of GLP-1 peptide and lead to aggregation. In addition, the presence of high molecular weight compounds in the cultivation broth increases the actual concentration of the product by occupying volume and therefore increasing the chances of the product molecules to interact with each other and form aggregates and/or fibrils. This is in agreement with similar observations in the literature, where for example macromolecules increase the reactivity of enzymes [28] or more specifically, ‘macromolecular crowding’ may lead to a dramatic acceleration in the rate of protein aggregation and formation of amyloid fibrils by volume exclusion, as in the case of Parkinson's disease [29], which in turn have been reported to increase the fluid’s viscoelasticity [30].

To conclude, we demonstrate that the broth viscosity caused by GLP-1 precursor aggregation can be prevented by utilizing yeast expression strains that support cultivation at high pH (≥ 6.2). If such strains are not available, de-aggregation of the product can be performed by increasing the pH of the cultivation broth to 8. This will reduce the product-related viscosity and allow for full recovery of the GLP-1 precursor. Alternatively, due to the soluble state of the aggregates, the pH adjustment can be performed after clarification to prevent cell lysis during the alkalic treatment and facilitate a more efficient recovery process.

As shown in the previous section to avoid the formation of GLP-1 precursor peptide aggregates, it is imperative to increase the cultivation pH setpoint to above 6.2. To enable this, we explored a new generation of GLP-1 precursor expression strains derived from host strain B (Table 1), including yNN2101 and yNN2104 described above, which can grow and express recombinant peptides more efficiently at a higher pH range. As shown in Fig. 3A, identical GLP-1 peptide concentrations were obtained from supernatant derived from yNN2104 cultivation at pH 6.2 using the dilution and alkalic extraction method, showing the absence of GLP-1 peptide aggregation. Interestingly, we observed a reduction in the dissolved oxygen tension (DO2) of the culture after 300 h of continuous cultivation. The drop in DO2 was not due to increased oxygen uptake rate (OUR) but correlated to an increase in viscosity as depicted in Fig. 3B. Analysis of broth samples revealed that the cause of the viscosity was not present in the supernatant, as was the case with the yNN2071 cultivations (product-related viscosity),—but instead associated with the cell pellet (Fig. 4A). In addition, the viscosity was not of a pseudoplastic nature but rather of a slightly shear thickening nature. Furthermore, cultivation broth from a dummy strain (yNN1953) that does not express the GLP-1 precursor exhibited the same shear-thickening behaviour dictated by the solid fraction of the broth (Fig. 4B). Hence, it can be concluded that for the yNN2104 and yNN1953 strains derived from the host strain B, the solid fraction of the broth (pellet) is responsible for a new type of “cell-derived” viscosity.

Results from a pH 6.2 cultivation of the host lineage B strain yNN2104. A Cultivation profiles of GLP-1 precursor concentration as measured by i) the dilution (●) and ii) the alkalic extraction method (○) to de-aggregate the precursor. B. Cultivation profiles of dissolved oxygen tension (DO2) (―), oxygen uptake rate (OUR) and viscosity measured at 2000 s−1

Example of a rheological characterisation of broth sample from the host lineage B strains. A yNN2104 expressing the GLP-1 precursor and B yNN1953 expressing no precursor. The viscosity at different shear rates of the fermentation broth , broth supernatant , and re-suspended washed pellet was analysed

This type of shear thickening fluid is expected to have a milder negative effect on the formation of gradients during scaling-up compared to the pseudoplastic broth caused by the GLP-1 peptide aggregation described in the previous section. This is because its flow behaviour index (n) is closer to unity than that of the GLP-1 peptide aggregated broth. However, the shear thickening properties of this broth would bear a higher burden on the impeller shaft, and thus increase the power consumption, as the very high shear conditions around the impellers will increase the broth viscosity and consequently the resistance to the impeller movement.

Microscopic as well as macroscopic analysis revealed a cell clumping morphology of the cells in the cultivation broth in addition to wall growth in the bioreactors (supplement). To identify the root cause behind this type of viscosity coming from the pellet (solid fraction) of the broth, the clumping behaviour of the cells was investigated. Initially, we examined whether the clumping phenotype was linked to cell flocculation. Flocculation in yeast is a reversible process where cells aggregate through adhesion mediated by surface proteins known as flocculins. These proteins facilitate the binding to mannose residues on the cell wall polysaccharides of neighboring yeast cells, enabling cell–cell adhesion [31]. To assess the relevance of flocculation for the cell-derived viscosity, the yeast strain yNN2201 (host B lineage) expressing a GLP1-GIP fusion precursor was cultivated in continuous cultivation at pH 6.5. The pellet fraction isolated from broth samples were washed twice in water before being resuspended in (a) water; (b) EDTA solution; or (c) Ca2+ solution. Calcium ions enable the cell surface flocculins to achieve their active conformation and bind the mannans in neighboring cells [32], while EDTA, by sequestering any residual Ca2+ from the cell pellet, ensures that the flocculins are inactive. The rheogram in Fig. 5 shows that addition of calcium dramatically reduced and flattened the viscosity profile signifying a Newtonian fluid behaviour. In contrast, sequestering calcium by the addition of EDTA, resulted in an even higher viscosity level and significantly more shear thickening behaviour at shear rates higher than 500 s−1. Therefore, Ca2+ seems to have a significant effect on reducing broth viscosity by increasing cell flocculation and, presumably, particle compactness, which in turn leads to lower broth viscosity [15].

The effect of cell flocculation induced by Ca2+ on the rheological properties of yeast cell suspensions of strain yNN2201. Fermentation broth samples were a double-washed and resuspended pellet in de-ionized water , b double-washed and resuspended pellet in a 0.02M EDTA solution , c double-washed and resuspended pellet in a 25% (w/v) CaCl2 solution

A potential solution to the viscosity issue identified in yeast expression strains of the host strain B lineage (Table 1) could, therefore, be to increase the Ca2+ concentration in the cultivation medium. Figure 6A–D shows the rheological characterization of cultivation broth from continuous cultivation of GLP-1 peptide expression strains yNN2548, yNN2101 and yNN2104 and of dummy strain yNN1953 before and after the continuous addition of CaCl2 (0.05M). The addition of Ca2+ caused a significant decrease in the flow behavioural index (n) and an increase in the flow consistency index (K) of the broth, signifying a substantial increase in the pseudoplasticity (Fig. 6C). The viscosity of the broth, when measured at high shear rates, such as the one prevailing in the vessel of the bioreactor (Fig. 6A) dropped, while the flow consistency index (K) increased (Fig. 6B). This clearly shows that the addition of Ca2+ results in a decrease in the shear thickening type of viscosity derived from the cells and an increase in the shear thinning behaviour. The origin of the shear thinning behaviour can be readily elucidated by examining the GLP-1 peptide concentrations obtained by the dilution method and alkalic treatment both pre- and post- the introduction of Ca2+ as depicted in Fig. 6E–G. Consequently, it is evident that the addition of calcium causes the GLP-1 precursor to aggregate. This observation is consistent with prior research on aggregation and fibrillation of sensitive proteins and peptides, which demonstrates that Ca2+ and high ionic strength markedly increases aggregation and fibrillation [33, 34]. The effect of Ca2+ on GLP-1 peptide aggregation leading to increased pseudoplastic viscosity is supported by the fact that the addition of calcium to the dummy strain yNN1953 cultivation has no significant effect on the flow behaviour index (n), as it is already close to 1, nor on the flow consistency index (K) (Fig. 6B–D). In contrast, the addition of Ca2+ to the dummy strain cultivation has substantial effect on the high shear viscosity, measured at 2000 s−1, (viscosity derived from the cells), which is significantly reduced due to the increased compactness of the cells clumps as a result of the induced flocculation (Fig. 6A).

Cultivation profiles of the rheological behaviour of broth and relative GLP-1 precursor concentration determined by the dilution (●) and alkalic extraction (○) method for 3 GLP-1 producing strains (yNN2548 yNN2101 , yNN2104  and one non-producing (dummy) strain (yNN1953 ) before and after the addition of Ca2+ (in the form of CaCl2). A Viscosity at 2000 s−1, B Flow consistency index (K), C. Flow behaviour index (n), D Average flow behaviour index (n). E–G Relative GLP-1 precursor concentration

The results from these experiments suggest that flocculation is not the cause of the cell derived viscosity but, on the contrary, leads to a reduction in cell-derived viscosity. Since there is no microscopic evidence of filamentous growth (data not shown), an alternative explanation for the cell clumping behaviour and broth viscosity could involve biofilm formation. It has been reported that adherence to surfaces during biofilm formation is mediated by the yeast genes FLO11, FLO10, and FIG2 [35]. FLO8 is a gene involved in the transcriptional activation of FLO11, which in turn plays a role in biofilm formation [36]. To explore if the FLO genes are involved in the cell-derived viscosity, yeast GLP-1 precursor expression strains harboring deletion of the FLO11 gene or FLO8 gene were constructed and cultivated in continuous cultivation prior to rheological characterization of the cultivation broths. As shown in Fig. 7, only negligible improvement in the viscosity at shear rate of 2000 s−1 was observed in broths derived from Δflo11 and Δflo8 strains compared to wildtype counterpart and the flow behaviour index (n) of the broths were not affected. In addition, the clumping behaviour of the cells persisted (data not shown). Collectively, this shows that the FLO8 and FLO11 genes and biofilm formation is not involved in the cell-derived viscosity observed in our cultivations.

Comparison of parental strain (yNN2104, WT) and derivative strains with deletion of FLO8 (yNN2912) and FLO11 genes (yNN2915) . Cultivation profiles of the A viscosity at shear rate of 2000 s−1, and B flow behaviour index (n)

AMN1, a gene involved in mitotic exit regulation [37] and post-mitotic cell separation in yeast, has been linked to a cell clumping phenotype in certain cell types and yeast strains [38]. A natural occurring AMN1D368 allele was shown to be associated with a clumping cell morphology due to its inhibitory action on post-mitotic cell separation. In contrast, strains with the natural AMN1V368 allele, or amn1 deletion strains, were shown to be proficient in mother-daughter cell separation leading to formation of uniform cells. As the GLP-1 precursor expression strains derived from host strain B in this study indeed harbor the AMN1D368 allele in the genome, confirmed by NGS sequencing (data not shown), it raised the possibility that Amn1-dependent cell separation might be the cause of the cell clumping morphology and potentially the increased viscosity during cultivation described here. To assess this, we constructed GLP-1 precursor expression strains with an amn1 gene deletion or with the AMN1D368V non-clumping allele.

Indeed, when the AMN1 knockout strain yNN2993 was cultivated in continuous cultivation, we observed that not only the clumping cell phenotype was eliminated (Fig. 8E), but also the viscosity was significantly reduced, showing no pronounced shear thickening behaviour (Fig. 8A, B). Consequently, the cultivation experienced no oxygen mass transfer challenges, as the DO2 was maintained at constant levels throughout the extended cultivation (Fig. 8C), while the oxygen uptake rate was unaffected (Fig. 8D).

Comparison of yNN2104 (AMN1 WT) versus yNN2993 (ΔAMN1) . Cultivation profiles of A broth viscosity at 250 s−1, B flow behaviour index (n), C dissolved oxygen tension, and D oxygen uptake rate, E Microscopic image (630 × magnification) of the cells

To establish the link between the clumping allele AMN1D368 activity and the high viscosity, shear- thickening broths, the cultivations of yNN2104 (AMN1D368) and yNN2993 (ΔAMN1) were repeated together with yNN3073 harboring the non-clumping allele AMN1D368V. During continuous cultivation of the AMN1D368 strain (yNN2104), limited oxygen mass transfer was evident, as demonstrated by the persistently low levels of DO2 observed after 240 h, despite maintaining a stable oxygen uptake rate (OUR) throughout the continuous process. Conversely, both the amn1 deletion strain and AMN1D368V strain (yNN2993 and yNN307 respectively) demonstrated significantly enhanced oxygen mass transfer capabilities throughout the entire process characterized by consistently higher levels of DO2 compared to yNN2104 (Fig. 9A and B). The rheological data of Fig. 9D and E clearly show a substantially reduced viscosity of the non-clumping strains yNN2993 and yNN3073 compared to yNN2104, while as expected, the flow behaviour index (n) of the non-clumping strains were close to unity signifying Newtonian behaviour. In addition, growth in the bioreactor walls was eliminated in the case of the non-clumping strains (see supplement). It is important to emphasize that the observed differences in viscosity are not attributable to variations in biomass levels, as the biomass concentrations were identical between the strains throughout the process (Fig. 9C). Interestingly, we only find one reference in literature describing a cell suspension or fermentation fluid as having shear thickening properties namely that of highly concentrated S. cerevisiae cell suspension [39]. In this paper however, no explanation or etiology about the phenomenon was given.

Comparison of yNN2104 (AMN1 WT) , yNN2993 (ΔAMN1) and yNN3073 (AMN1D368V) . Cultivation profiles of A dissolved oxygen tension, B oxygen uptake rate, C relative dry cell weight, D viscosity at 2000 s−1, E flow behaviour index (n) and F relative GLP-1 peptide concentration

We assume that the incomplete mother-to-daughter separation of AMN1D368 strains leads to fluffy clumps that increase the viscosity of the cultivation broth. This may be due to the irregularities or roughness of the clumps’ surface and their increased specific surface area, which contribute to stronger particle–particle interactions and greater disruption of solvent flow, particularly at high particle concentrations and high shear rates. In contrast, in the case of flocculation, the cells form larger, more compact clumps with smoother surfaces and lower specific surface areas (due to the strong attractive interactions of the cells mediated by the action of flocculins) leading to significantly lower viscosity and Newtonian behaviour.

The cultivation experiment in Fig. 10 aims to separate the cell-derived (AMN1 dependent) viscosity from the GLP-1 precursor (i.e. product)-derived viscosity described in this study and to confirm that the latter is independent of the host strain, but strongly dependent on the cultivation pH. For this purpose, the amn1 deletion strain yNN2993 derived from host strain B, which can be cultivated at a broad range of pH values, was chosen for continuous cultivation to avoid interference in the rheological characterization by the cell-derived viscosity. The cultivation pH was reduced from 6.4 to 5.3 at 427 h. The reduction in cultivation pH caused an instantaneous increase in viscosity, and a concomitant reduction in the flow behavioural index (n) indicating that the broth pseudoplasticity increases. As expected, the ratio of GLP-1 precursor concentration measured by the alkalic extraction method over the dilution method (called aggregational index in Fig. 10) correlated well with the increase in both viscosity and pseudoplasticity. This demonstrates that the cultivation pH determines the extent of GLP-1 precursor aggregation and, consequently, viscosity. A high cultivation pH (> 6.4) eliminated GLP-1 peptide aggregation and thus product-related viscosity. Rheological characterization of the broth and its fractions before and after the change of pH clearly confirmed that the increase in viscosity and pseudoplasticity was derived from the soluble fraction of the broth (Fig. 10).

Effect of cultivation pH on yNN2104 (ΔAMN1). Cultivation profiles of broth viscosity at shear rate 250 s−1, flow behaviour index (n) , aggregation index which is the concentration as determined by the alkalic extraction method divided by the dilution extraction method, and cultivation pH (―). The viscosity at different shear rates of the fermentation broth , supernatant post-centrifugation , and washed, resuspended pellet was measured before and after the change of cultivation pH from 6.4 to 5.3

Controlling viscosity in industrial fermentation processes is critical, as it directly impacts mass transfer efficiency, gas hold-up, and overall bioreactor performance. Dissecting and optimizing viscosity-related parameters are essential for enhancing microbial productivity and to ensure the economic viability of large-scale fermentation operations. In this study, we identified two root causes of high viscosity in the industrial production of a critical pharmaceutical precursor and presented solutions to efficiently eliminate this. Specifically, we found that AMN1 gene activity is the cause of the cell-derived viscosity increase of cultivation broth from yeast continuous cultivations, while the expression of the GLP-1 precursor (product-related viscosity) at certain cultivation conditions induces a high viscosity and pseudoplastic behaviour of the broth. Figure 11 summarizes the main conclusions of this study showing that by combining strain engineering to eliminate Amn1p dependent cell-derived viscosity with cultivation conditions at higher pH to eliminate product-related viscosity (demonstrated by strain yNN2993), the cultivation broth can be transformed into a “cultivation friendly” low viscosity, Newtonian fluid, which can easily be scaled up to large bioreactor tanks.

Summary comparison for the host lineage A strain yNN2071 cultivated at pH 5.3 , host lineage B strains yNN2104 (AMN1D368) cultivated at pH 6.2 and yNN2993 (ΔAMN1) cultivated at pH 6.5 . Cultivation profiles of A broth viscosity at shear rate 250 s−1 and B flow behavioural index (n)

No datasets were generated or analysed during the current study.

A pseudoplastic fluid is a type of non-Newtonian fluid that decreases in viscosity with increasing shear rate and thereby exhibits shear-thinning behaviour.

A Newtonian fluid is defined as a fluid with a viscosity that remains constant regardless of the applied shear and thus exhibits a linear relationship between shear stress and shear rate.

The flow behavior index (n) is a dimensionless parameter that characterizes the type of flow exhibited by a fluid. When n = 1, the fluid has a constant viscosity regardless of the shear rate (Newtonian). For n < 1, the fluid's viscosity decreases as the shear rate increases (shear-thinning or pseudoplastic behaviour). Conversely, for n > 1, the fluid's viscosity increases with increasing shear rate (shear-thickening behavior). See Rheological characterization in Methods.

Drucker DJ. Efficacy and safety of GLP-1 medicines for type 2 diabetes and obesity. Diabetes Care. 2024;47:1873–88.

Article CAS PubMed Google Scholar

Tang H, Lu Y, Okun MS, Donahoo WT, Ramirez-Zamora A, Wang F, Huang Y, Armstrong M, Svensson M, Virnig BA, et al. Glucagon-like peptide-1 receptor agonists and risk of Parkinson’s disease in patients with type 2 diabetes: a population-based cohort study. Mov Disord. 2024;39:1960–70.

Article CAS PubMed Google Scholar

Hölscher C. Glucagon-like peptide-1 class drugs show clear protective effects in Parkinson’s and Alzheimer’s disease clinical trials: a revolution in the making? Neuropharmacology. 2024;253: 109952.

Article PubMed Google Scholar

Lau JF, Andersen AS, Bloch P, Lau J, Garibay PW, Kruse T, Noerby ISN, Jessen CU, Christensen C, Norrild JC. Semi-recombinant preparation of glp-1 analogues. WO-2009083549-A1. 2009.

Kjeldsen T, Andersen AS, Hubálek F, Johansson E, Kreiner FF, Schluckebier G, Kurtzhals P. Molecular engineering of insulin for recombinant expression in yeast. Trends Biotechnol. 2024;42:464–78.

Article CAS PubMed Google Scholar

Buckland BC, Gbewonyo K, Jain D, Glazomitsky K, Hunt G, Drew SW. Oxygen transfer efficiency of hydrofoil impellers in both 800L and 1900L fermenters. In: King R, editor. Second international conference on bioreactor fluid dynamics. London: Elsevier; 1988. p. 1–16.

Google Scholar

Tiso T, Demling P, Karmainski T, Oraby A, Eiken J, Liu L, Bongartz P, Wessling M, Desmond P, Schmitz S, et al. Foam control in biotechnological processes—challenges and opportunities. Discover Chemical Engineering. 2024;4:2.

Article Google Scholar

Goudar CT, Strevett KA, Shah SN. Influence of microbial concentration on the rheology of non-Newtonian fermentation broths. Appl Microbiol Biotechnol. 1999;51:310–5.

Article CAS PubMed Google Scholar

Seviour RJ, McNeil B, Fazenda ML, Harvey LM. Operating bioreactors for microbial exopolysaccharide production. Crit Rev Biotechnol. 2011;31:170–85.

Article CAS PubMed Google Scholar

Voulgaris I, Finka G, Uden M, Hoare M. Enhancing the selective extracellular location of a recombinant E. coli domain antibody by management of fermentation conditions. Appl Microbiol Biotechnol. 2015;99:8441–53.

Article CAS PubMed PubMed Central Google Scholar

Minero G, Antonio S, Møllebjerg A, Thiesen C, Johansen Mikkel I, Jørgensen Nis P, Birkedal V, Otzen DE, Meyer Rikke L. Extracellular G-quadruplexes and Z-DNA protect biofilms from DNase I, and G-quadruplexes form a DNAzyme with peroxidase activity. Nucleic Acids Res. 2024;52:1575–90.

Article CAS PubMed PubMed Central Google Scholar

Dobson CM. Protein folding and misfolding. Nature. 2003;426:884–90.

Article CAS PubMed Google Scholar

Bueche F. Viscosity of entangled polymers; theory of variation with shear rate. J Chem Phys. 1968;48:4781–4.

Article CAS Google Scholar

Larson RG. The structure and rheology of complex fluids. New York: Oxford University Press; 1999.

Google Scholar

Bliatsiou C, Schrinner K, Waldherr P, Tesche S, Böhm L, Kraume M, Krull R. Rheological characteristics of filamentous cultivation broths and suitable model fluids. Biochem Eng J. 2020;163: 107746.

Article CAS Google Scholar

Albaek MO, Gernaey KV, Hansen MS, Stocks SM. Modeling enzyme production with Aspergillus oryzae in pilot scale vessels with different agitation, aeration, and agitator types. Biotechnol Bioeng. 2011;108:1828–40.

Article CAS PubMed Google Scholar

Wucherpfennig T, Kiep KA, Driouch H, Wittmann C, Krull R. Chapter 4—morphology and rheology in filamentous cultivations. In: Laskin AI, Sariaslani S, Gadd GM, editors. Advances in applied microbiology, vol. 72. Amsterdam: Academic Press; 2010. p. 89–136.

Google Scholar

Kjeldsen T, Brandt J, Andersen AS, Egel-Mitani M, Hach M, Pettersson AF, Vad K. A removable spacer peptide in an α-factor-leader/insulin precursor fusion protein improves processing and concomitant yield of the insulin precursor in Saccharomyces cerevisiae. Gene. 1996;170:107–12.

Article CAS PubMed Google Scholar

Gardner JM, Jaspersen SL. Manipulating the yeast genome: deletion, mutation, and tagging by PCR. In: Smith JS, Burke DJ, editors. Yeast genetics: methods and protocols. New York: Springer; 2014. p. 45–78.

Chapter Google Scholar

Gietz RD, Schiestl RH. High-efficiency yeast transformation using the LiAc/SS carrier DNA/PEG method. Nat Protoc. 2007;2:31–4.

Article CAS PubMed Google Scholar

Lõoke M, Kristjuhan K, Kristjuhan A. Extraction of genomic DNA from yeasts for PCR-based applications. Biotechniques. 2011;50:325–8.

Article PubMed PubMed Central Google Scholar

Zapadka KL, Becher FJ, Uddin S, Varley PG, Bishop S, dos Gomes AL, Jackson SE. A pH-induced switch in human glucagon-like peptide-1 aggregation kinetics. J Am Chem Soc. 2016;138:16259–65.

Article CAS PubMed Google Scholar

Egbu R, van der Walle CF, Brocchini S, Williams GR. Inhibiting the fibrillation of a GLP-1-like peptide. Int J Pharm. 2020;574: 118923.

Article CAS PubMed Google Scholar

Gernaey KV. A perspective on PSE in fermentation process development and operation. Comput Aided Chem Eng. 2015;37:123–30.

Article CAS Google Scholar

Rekhi S, Garcia CG, Barai M, Rizuan A, Schuster BS, Kiick KL, Mittal J. Expanding the molecular language of protein liquid–liquid phase separation. Nat Chem. 2024;16:1113–24.

Article CAS PubMed PubMed Central Google Scholar

Přáda Brichtová E, Krupová M, Bouř P, Lindo V, dos Gomes Santos A, Jackson SE. Glucagon-like peptide 1 aggregates into low-molecular-weight oligomers off-pathway to fibrillation. Biophys J. 2023;122:2475–88.

Article PubMed PubMed Central Google Scholar

Srivastava AK, Pittman JM, Zerweck J, Venkata BS, Moore PC, Sachleben JR, Meredith SC. β-Amyloid aggregation and heterogeneous nucleation. Protein Sci. 2019;28:1567–81.

Article CAS PubMed PubMed Central Google Scholar

Banani SF, Lee HO, Hyman AA, Rosen MK. Biomolecular condensates: organizers of cellular biochemistry. Nat Rev Mol Cell Biol. 2017;18:285–98.

Article CAS PubMed PubMed Central Google Scholar

Munishkina LA, Cooper EM, Uversky VN, Fink AL. The effect of macromolecular crowding on protein aggregation and amyloid fibril formation. J Mol Recognit. 2004;17:456–64.

Article CAS PubMed Google Scholar

Ray S, Singh N, Kumar R, Patel K, Pandey S, Datta D, Mahato J, Panigrahi R, Navalkar A, Mehra S, et al. α-Synuclein aggregation nucleates through liquid–liquid phase separation. Nat Chem. 2020;12:705–16.

Article CAS PubMed Google Scholar

Lipke PN, Ragonis-Bachar P. Sticking to the subject: multifunctionality in microbial adhesins. J Fungi (Basel). 2023;9:419.

Article CAS PubMed Google Scholar

Soares EV. Flocculation in Saccharomyces cerevisiae: a review. J Appl Microbiol. 2011;110:1–18.

Article CAS PubMed Google Scholar

Urrutia R, Cruciani RA, Barker JL, Kachar B. Spontaneous polymerization of the antibiotic peptide magainin 2. FEBS Lett. 1989;247:17–21.

Article CAS PubMed Google Scholar

Huang S, Xu B, Liu Y. Calcium promotes α-synuclein liquid-liquid phase separation to accelerate amyloid aggregation. Biochem Biophys Res Commun. 2022;603:13–20.

Article CAS PubMed Google Scholar

Vallejo JA, Sánchez-Pérez A, Martínez JP, Villa TG. Cell aggregations in yeasts and their applications. Appl Microbiol Biotechnol. 2013;97:2305–18.

Article CAS PubMed Google Scholar

Wang Z, Xu W, Gao Y, Zha M, Zhang D, Peng X, Zhang H, Wang C, Xu C, Zhou T, et al. Engineering Saccharomyces cerevisiae for improved biofilm formation and ethanol production in continuous fermentation. Biotechnol Biofuels Bioprod. 2023;16:119.

Article CAS PubMed PubMed Central Google Scholar

Wang Y, Shirogane T, Liu D, Harper JW, Elledge SJ. Exit from exit: resetting the cell cycle through Amn1 inhibition of G protein signaling. Cell. 2003;112:697–709.

Article CAS PubMed Google Scholar

Fang O, Hu X, Wang L, Jiang N, Yang J, Li B, Luo Z. Amn1 governs post-mitotic cell separation in Saccharomyces cerevisiae. PLoS Genet. 2018;14: e1007691.

Article PubMed PubMed Central Google Scholar

Bryjak M, Meissner W. Concentration of Baker’s yeast suspension by means of membrane enhanced drying—rheological approach. Chem Pap. 2003;57:11–5.

CAS Google Scholar

Download references

We would like to thank Asser Sloth Andersen, Jakob Brandt, Peter Boldsen Knudsen and Vera Schramke for their careful input to the manuscript.

The authors did not receive funding from any organization for the submitted work.

Recombinant Drug Research, Novo Nordisk A/S, Novo Park Alle, Novo Nordisk Park 1, 2760, Måløv, Denmark

Ioannis Voulgaris, Anders Nygaard Nielsen, Tine Petersen & Sanne Jensen

Search author on:PubMed Google Scholar

Search author on:PubMed Google Scholar

Search author on:PubMed Google Scholar

Search author on:PubMed Google Scholar

IV conceptualized and designed the experiments regarding the viscosity related to the product, IV and SJ conceptualized and designed the experiments regarding the viscosity derived from the AMN1 expression. SJ designed the strains. TP conducted the molecular biology work. ANN conducted the fermentations and the associated analysis. ANN and IV conducted the rheological analysis. IV analysed the data. IV and SJ wrote the manuscript. All authors read and approved the final manuscript.

Correspondence to Ioannis Voulgaris.

The authors declare no competing interests. This work is part of a patent application.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, 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 you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. 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-nc-nd/4.0/.

Reprints and permissions

Voulgaris, I., Nielsen, A.N., Petersen, T. et al. Eliminating viscosity challenges in continuous cultivation of yeast producing a GLP-1 like peptide. Microb Cell Fact 24, 130 (2025). https://doi.org/10.1186/s12934-025-02745-6

Download citation

Received: 13 March 2025

Accepted: 07 May 2025

Published: 05 June 2025

DOI: https://doi.org/10.1186/s12934-025-02745-6

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative