All cultivations were performed with Aureobasidium pullulans NRRL 62042 (ARS Culture Collection of the United States Department of Agriculture, Washington D.C., United States). This strain was isolated from a leaf in Thailand (Patalung) (Manitchotpisit et al., 2014).

Previous works have produced polyol lipid in media that contain complex compounds. The complex medium optimized by Leathers et al. (2018) is the starting point for developing a minimal medium and consists of 1.5 g L⁻¹ peptone, 0.9 g L⁻¹ yeast extract, 1 g L⁻¹ NaCl, 1 g L⁻¹ K₂HPO₄, 0.8 g L⁻¹ MgSO₄ × 7 H₂O, and 120 g L⁻¹ sucrose. Yeast extract and peptone are replaced with a defined nitrogen source, a trace element, and a vitamin solution. The last two mentioned were previously used by Geiser et al. (2016) for cultivations with Ustilago maydis. The nitrogen content in yeast extract (0.13% w/w) and peptone (0.11% w/w) was estimated based on literature data (Garay et al., 2018) and replaced with 2.01 g L⁻¹ NaNO3, 1.56 g L⁻¹ (NH₄)₂SO₄, 1.27 g L⁻¹ NH₄Cl, or 0.95 g L⁻¹ NH₄NO₃. The vitamin solution contained per liter 0.05 g of D-biotin, 1 g of D-calcium pantothenate, 1 g of nicotinic acid, 25 g of myo-inositol, 1 g of thiamine hydrochloride, 1 g of pyridoxol hydrochloride, and 0.2 g of para-aminobenzoic acid. The trace element solution contained per liter 1.5 g of EDTA, 0.45 g of ZnSO₄ × 7 H₂O, 0.10 g of MnCl₂ × 4 H₂O, 0.03 g of CoCl₂ × 6 H₂O, 0.03 g of CuSO₄ × 5 H₂O, 0.04 g of Na₂MoO₄ × 2 H₂O, 0.45 g of CaCl₂ × 2 H₂O, 0.3 g of FeSO₄ × 7 H₂O, 0.10 g of H₃BO₃, and 0.01 g of KI. Sucrose is replaced with glucose because A. pullulans can form FOS, which complicates substrate depletion analysis. To avoid MgNH₄PO₄ × 6 H₂O formation, also known as struvite (Steimann et al., 2023), during medium preparation, MgSO₄ × 7 H₂O is added directly to the cultivation before inoculation to avoid precipitation.

After successfully establishing a minimal medium based on the medium from Leathers et al. (2018) the developed medium was optimized by a design of experiments (DoE) approach. This method uses the advancement of statistical techniques to increase efficiency by reducing the amount of work and incorporating the dependence of components on each other. The latter does not take place in the often-used one-factor-at-a-time method. In this case, only one factor is changed simultaneously without considering their interaction. Experiments were planned and evaluated with the software DesignExpert11 (Stat-Ease Inc., Minneapolis, Minnesota, United States).

For polyol lipid determination, culture broth was centrifuged (5 min, 17,000 ×g). The cell pellet and polyol lipid were resuspended in acetonitrile. After another centrifugation step (5 min, 17,000 ×g), the acetonitrile, including dissolved polyol lipid, is separated from the cell fragments. The determination is carried out gravimetrically by evaporating the acetonitrile in pre-weighed glass vials using a speed vac at medium temperature (Savant Instruments, Speed Vac SC-100, Farmingdale, New York, United States). All wells from cultivations in MTPs undergo complete processing, followed by an additional acetonitrile rinse. After rinsing, the acetonitrile was used to resuspend the cell pellet, and polyol lipid after the first centrifugation step. In the case of samples obtained from the bioreactor, 1 mL of culture broth is utilized for each sample, with an accompanying 1 mL of acetonitrile.

For cell dry weight (CDW) determination, culture broth was centrifuged (5 min, 17,000 ×g). The cell pellet and polyol lipid were resuspended in 50% ethanol (v_EtOH/v_H2O). After another centrifugation step (5 min, 17,000 × g), the supernatant, including dissolved polyol lipid, was discarded. The cell pellet was resuspended in ultrapure H₂O and transferred into pre-weighed glass vials. The determination is carried out gravimetrically by evaporating the H₂O for 2 days at 70°C. A correction factor of 1.52 was determined by the cell damage caused by ethanol. Therefore, a YEP preculture was cultivated for 12 h, the culture broth centrifuged (5 min, 17,000 × g), and the cell pellet resuspended with ultrapure H₂O or 50% ethanol (v_EtOH/v_H2O). All wells from cultivations in MTPs undergo complete processing, followed by an additional rinse with 50% ethanol. After rinsing, the 50% ethanol was used to resuspend the cell pellet and polyol lipid after the first centrifugation step. In the case of samples obtained from the bioreactor, 1 mL of culture broth is utilized for each sample, with an accompanying 1 mL of 50% ethanol (v_EtOH/v_H2O).

To determine pullulan concentrations, the culture broth was centrifuged (5 min, 17,000 × g), and the supernatant was mixed 1:1 (v/v) with ethanol. Precipitated pullulan was transferred after centrifugation (5 min, 17,000 × g) to a pre-weighed vial using ultrapure H₂O. The determination is carried out gravimetrically after drying at 70°C for 2 days.

Glucose, fructose, sucrose, 1-kestose, nystose, and fructofuranosylnystose were determined by high-performance liquid chromatography (HPLC). After centrifugation (5 min, 17,000 × g) of 1 mL culture broth, the supernatant was mixed 1:1 (v/v) with ethanol to precipitate pullulan. Samples were analyzed using an UltiMate 3000 HPLC System (Thermo Fisher Scientific, Waltham, Massachusetts, United States), including an Ultimate 3000 pump, Ultimate 3000 autosampler, Ultimate 3000 column oven (Thermo Fisher Scientific, Waltham, Massachusetts, United States) and a Knauer RI-Detector RefractoMax 521 (KNAUER Wissenschaftliche Geräte GmbH, Berlin, Germany). When glucose was used as a carbon source, substrate consumption was determined with a Metab-AAC column (BF-series, Ion exchange, 300 × 7.8 mm) from Isera GmbH (Düren, Germany) at 40 °C and a flow rate of 0.6 mL min^-1. The mobile phase consisted of 5 mM H₂SO₄. Using sucrose as a carbon source, glucose, fructose, and FOS concentration were measured with a NUCLEODUR HILIC column (Multospher APS HP–3 µm HILIC, 250 × 2 mm) from CS-Chromatographie Service GmbH (Langerwehe, Germany) at 50 °C and a flow rate of 0.5 mL min^-1. The mobile phase consisted of 80% acetonitrile (v_ACN/v_H2O). Measurements were analyzed using Chromeleon 7.2.10 software (Thermo Fisher Scientific, Waltham, Massachusetts, United States).

While using sucrose as carbon source, A. pullulans builds up and consumes fructooligosaccharides (FOS). FOS are polysaccharides consisting of one glucose and two or more fructose units. To streamline the process and simplify substrate uptake kinetics, sucrose is replaced by glucose as carbon source. To ensure that this does not lead to reduced production or impaired growth, the two carbon sources were compared with each other. For this purpose, cultivations with the same amount of carbon were performed in MTPs using SystemDuetz and the complex medium developed by Leathers et al. (2018).

Within the first 24 h, exponential growth occurred until 5.5–6.0 g L⁻¹ CDW (Figure 2A). After the exponential growth phase, the CDW increased to 17.3 g L⁻¹ using sucrose and 10.7 g L⁻¹ using glucose as a carbon source. This resulted in a final product-to-biomass yield (Y_P/X) from sucrose and glucose of 1.8 g g⁻¹ and 2.6 g g⁻¹, respectively. Analyzing off-gas data in bioreactor cultivations (data not shown), a substrate limitation except carbon was found, which could be attributed to a nitrogen limitation between 24 and 48 h. As nitrogen is essential for cell growth, biomass stagnation occurs in a limitation. In the absence of nitrogen, a constant biomass concentration should, therefore, be measured until cell death. A. pullulans form intracellular storage lipids under unbalanced cultivation conditions, like high carbon (C) and low nitrogen (N) concentrations (high C/N ratio). These lipids are still present in the biomass when determining CDW and are thus included in the weight (Ratledge and Wynn, 2002; Xue et al., 2018). This increase in biomass, without increasing the number of active cells, leads to a lower Y_P/X. With both carbon sources, maximal polyol lipid titers of approximately 30 g L⁻¹ were achieved (Figure 2A). For this reason, it is assumed that growth ends after 48 h and Y_P/X are calculated using the measured biomasses at this time. This resulted in a Y_P/X from sucrose and glucose of 4.8 g g⁻¹ and 4.0 g g⁻¹, respectively. At the end of the cultivation, 0.42 Cmol L⁻¹ glucose and 0.95 Cmol L⁻¹ fructose remained when sucrose was used as a carbon source (Figure 2B). During cultivation, almost all sucrose was converted to glucose, fructose, 1-kestose, and nystose within the first 8 h. After that, the glucose concentration stagnated markedly, indicating that glucose was metabolized. The 1-kestose concentration decreased after 8 h, while fructose, nystose, and fructofuranosylnystose accumulated. After 100 h, all sugars were converted into the monomers fructose and glucose and were metabolized simultaneously. When glucose was used as the initial carbon source, 1.0 Cmol L⁻¹ was left after 240 h (Figure 2C). Furthermore, Figure 2C shows that the consumption of the residual sugars is similar when glucose or sucrose is used as the initial carbon source. Substrate costs make up to 50% of the production costs (Akbari et al., 2023). Leftover substrate is discarded at the end of a batch process and thus lowers the economic efficiency. Therefore, the product-to-substrate yields (Y_P/S) calculated in this work refer to the initial carbon amount. Y_P/S is given in gram product per Cmol substrate to compare glucose and sucrose. Using glucose as carbon source resulted in a higher Y_P/S of 7.7 g Cmol_glucose ⁻¹ compared to 7.2 g Cmol_sucrose ⁻¹. In both cultivations, approximately 6 g L⁻¹ pullulan was produced, which appeared to be degraded toward the end (Figure 2D). Furthermore, the pH value in both cultivations decreased to approximately 3 within 24 h and remained below 3 (Figure 2D). This low pH value might be due to the production of PMA, whose malate monomers have a free acid group. Altogether, the only noticeable difference seems to be higher biomasses when using sucrose as carbon source after exponential growth, possibly due to the assumed increased formation of intracellular storage lipids using sucrose as carbon source. In addition, gravimetrical CDW determination after washing the cells with 50% ethanol could lead to certain deviations. Regarding growth behavior and polyol lipid production, A. pullulans NRRL 62042 did not show mentionable differences between glucose and sucrose. Having in mind that sucrose entails more complex carbon uptake kinetics, glucose was used in the following experiments.

Previously used media included complex nitrogen sources such as yeast extract and peptone or complex carbon, and nitrogen sources like corn steep liquor, xylose mother liquor, or gluconate mother liquor (Cheng et al., 2017; Leathers et al., 2018; Li et al., 2021). For the development of a minimal medium, complex components have to be replaced. Therefore, several defined nitrogen sources were tested: NH₄NO₃, NaNO₃, NH₄Cl, and (NH₄)₂SO₄. Yeast extract and peptone contain trace elements and vitamins; these were substituted using vitamin and trace-element solutions. Cultivations were performed in MTPs for 10 days in SystemDuetz.

Compared to the yeast extract- and peptone-containing medium, the media with defined nitrogen sources seemed to have a longer lag phase (Figure 3A), arguing for an adapted preculture preparation. When NaNO₃ was used, the lag phase was 8 h longer. This is probably because microorganisms first convert nitrate into ammonium, which requires energy and thus results in slow growth (Siverio, 2002). Cultivations with media containing ammonium salts (NH₄NO₃, NH₄Cl, and (NH₄)₂SO₄) showed an almost identical growth phase. Strikingly, nitrogen sources containing nitrate resulted in higher CDW in the first 120 h (Figure 3B). After the exponential growth phase, the biomass concentration using yeast extract- and peptone-containing medium was 6.3 g L⁻¹, comparable to those media with (NH₄)₂SO₄ and NH₄CI as nitrogen sources. With nitrate salts, CDWs from NH₄NO₃ and NaNO₃ of 9.1 g L⁻¹ and 10.7 g L⁻¹, respectively, were reached after 48 h. This could be due to the increased intracellular storage lipid formation. Polyol lipid production was highest using NH₄NO₃ with a titer of 32.4 g L⁻¹ (Figure 3C). However, taking into consideration the standard deviation, all cultivations are in a similar range concerning the maximal polyol lipid titers. NH₄Cl-containing medium achieved the highest Y_P/X with 4.5 g g⁻¹. The lowest was 2.9 g g⁻¹ by using NaNO₃. After the growth phase, glucose consumption was almost linear under all conditions (Figure 3D). In the cultivations where nitrate was used, glucose was depleted after 168 h (NaNO₃) and 192 h (NH₄NO₃). The others still contained at least 28 g L⁻¹ residual sugar, comparable to the reference with complex nitrogen sources. When glucose was depleted, pullulan concentrations reached their highest values with up to 11 g L⁻¹, a concentration that started to decrease afterward (Figure 3E). With NH₄Cl and (NH₄)₂SO₄, almost no pullulan was produced (c_pullulan,max < 1 g L⁻¹). Moreover, pullulan formation stagnated in these cultivations after the pH dropped below 2. When NH₄NO₃ was used as a nitrogen source, the pH value increased after 24 h (Figure 3F). This might be due to the diauxic metabolism of NH₄ ⁺ and NO₃ ⁻. While metabolizing the preferred NH₄ ⁺ (base), the counter ion NO₃ ⁻ (acid) remains in the culture broth, leading to its acidification (sharp drop in the pH value during the first 24 h). After NH₄ ⁺ is depleted, metabolizing NO₃ ⁻ removes the acid from the culture broth, which leads to alkalization (increase in pH value). The subsequent slow drop of the pH value is presumably associated with PMA production. Using nitrate-containing nitrogen sources tends to result in slightly higher pH values than pure ammonium-containing ones. Furthermore, dark coloration of the culture broth was observed after 48 h for the media containing NH₄Cl and (NH₄)₂SO₄, which changed to a deep black after 72 h, indicating melanin formation (Jiang et al., 2016).

Depending on the media composition, carbon remains present even after 10 days. The carbon consumption using the initial medium was not described by Leathers et al. (2018). A direct comparison is, therefore, not possible. It was also shown that the increased carbon consumption was due to increased pullulan formation. The strain A. pullulans NRRL50384 (equivalent to RSU29) used by Leathers et al. (2018) in their medium development is known as a strong pullulan producer, enabling the complete consumption of sucrose (Manitchotpisit et al., 2014). In cultivations in which lower pH values were reached, also low pullulan titers were obtained (Figure 3F). In batch fermentations with A. pullulans CCTCC 209298, where the pH was controlled, higher pullulan production was noted at pH 3.5 (26.3 g L⁻¹) and 4.5 (24.6 g L⁻¹), while the lowest titer (18.5 g L⁻¹) was achieved at pH 2.5 (Wang et al., 2013).

Altogether, sucrose, yeast extract, and peptone were successfully replaced. Although NH₄Cl and (NH₄)₂SO₄ offer the advantage of preventing pullulan production by a low pH and resulted in a better Y_P/S, using only NH₄ as a nitrogen source led to melanin formation. Melanin may stain the product and thus complicate purification, while pullulan remains in the aqueous phase and can thus be easily separated from polyol lipid. Cultivation with NH₄NO₃ as a nitrogen source showed a shorter lag phase and lower pullulan production than NaNO₃. The maximum polyol lipid titer and Y_P/S were comparable, while a higher Y_P/X was achieved using NH₄NO₃. Accordingly, NH₄NO₃ was selected as the nitrogen source for subsequent experiments.

After developing a defined minimal liamocin production medium (MLP), a DoE approach was used for medium optimization regarding polyol lipid production. First, medium components were investigated to identify factors significantly influencing polyol lipid production. Therefore, a two-level factorial design was used. Second, the concentrations of factors influencing polyol lipid titer were optimized by a central composite design. The results were assessed by analysis of variance (ANOVA) and summarized in tabular form. In ANOVA, the significance of an observed effect is assessed using the p-value. A significant effect can be assumed if the p-value is less than 5% (p < 0.05). The smaller the p-value, the more important the effect. To investigate whether a scatter is due to an effect or to noise, the variance between the groups (prediction variance) is compared with the variance within the groups (error variance) by calculating the quotient of the sums of squares, yielding the F-value. A group is a concentration of a factor. There are, therefore, three groups per factor (component), each of which is defined by the concentration of the factor. The larger the F-value, the greater the scatter between the individual groups compared to the error variance. The greater the variance (F-value) between the groups, the more likely there are significant differences between them. The model is the mathematical representation of the relationships between the individual factors. The lack of fit describes the extent to which the model’s prediction misses the observations or measured values. Therefore, a non-significant lack of fit means that the model can predict the measured values.

While MTP cultivations allow high-throughput investigation of cultivation conditions at low cost, bioreactor cultivations offer generation of precise online data and automated feeding controls. In bioprocess development, however, reproducibility is a well-known challenge when the process is transferred from a shaken to a stirred system (Büchs, 2001). For these reasons, the cultivation with the developed minimal medium was transferred from the 2 mL MTP scale into a 700 mL stirred-tank bioreactor cultivation. Further, the optimized minimal medium was compared with the synthetic starting medium. To enable a comparison between MTP and bioreactor, batch fermentations were performed as described in Chapter 2.4.4 without pH control. The process parameters are summarized in Table 7.

In the following, MLP and oMLP medium were compared at bioreactor scale (Figure 7) During the optimization process, glucose concentration was increased from 126 to 208 g L⁻¹, K₂HPO₄ concentration was increased from 1.0 to 6.3 g L⁻¹, and NH₄NO₃ concentration increased from 0.95 to 1.91 g L⁻¹. Overall, the polyol lipid titer in the bioreactor increased by almost 44% from 29 g L⁻¹ (MLP) to 41 g L⁻¹ (oMLP) (Figure 7A). Polyol lipid formation stagnated in both media, though at different timepoints, after 71 h (MLP) and 120 h (oMLP) and increased again later. Before the stagnation, the polyol lipid formation rate was 0.18–0.19 g L⁻¹ h⁻¹ for both. Medium optimization shortened the time of stagnation by approximately 24 h. Subsequently polyol lipid formation rate increased to 0.31 g L^-1 h⁻¹, while in MLP, it decreased to 0.12 g L^-1 h⁻¹. Using the oMLP medium, a higher final biomass of 27.0 g L⁻¹ than in the MLP cultivation (13 g L⁻¹) was reached (Figure 7B). Increasing carbon concentrations while maintaining a constant C/N ratio also implies proportionally rising nitrogen concentrations, leading to enhanced growth. The maximal growth rate was increased from 0.11 h^-1 to 0.18 h^-1. It should be noted that only a few samples were taken during the growth phase, and the growth rate could, therefore, be inaccurate. However, a faster growth using oMLP compared to MLP can be seen in Figure 7B. Figure 7C shows a peak in oxygen transfer rate (OTR) of 33 mmol L⁻¹ h⁻¹ (oMLP) and 17 mmol L⁻¹ h⁻¹ (MLP), followed by a constant value between 8 and 10 mmol L⁻¹ h⁻¹ until carbon depletion. This course suggests a substrate limitation except carbon (Anderlei and Büchs, 2001), which may be assigned to a nitrogen limitation based on CDW measurements and the high C/N ratio. The respiration quotient, calculated by dividing carbon dioxide by the oxygen transfer rate, was in the range of 1.5 after growth, which means that more CO₂ is produced than O₂ is consumed, indicating the production of a reduced product. Depending on the congener, the degree of reduction of polyol lipid is 5–5.2, so a respiration quotient greater than 1 could be attributed to polyol lipid production. The growth was terminated after approximately 50 h, and the following biomass increase might possibly be attributed to the formation of intracellular lipids. A renewed decrease in oxygen transfer rate suggests the consumption of glucose. After 240 h, the cultivation with the oMLP reached a biomass of 27 g L⁻¹, whereas after 50 h, only 17 g L⁻¹ was present. 10 g L⁻¹ was thus probably due to intracellular storage lipid formation. The Y_P/X was 1.5 g g⁻¹ and 2.4 g g⁻¹, respectively. Figure 7D shows that the substrate decrease in the stationary phase is linear. The glucose depletion rate in the stationary phase before the optimization was 0.6 g L⁻¹ h⁻¹ and increased to 1.0 g L⁻¹ h⁻¹ after optimization. After the consumption of glucose, pullulan was degraded (Figure 7E). The highest pullulan concentrations were 27.1 g L⁻¹ with the MLP and 46.1 g L⁻¹ with the oMLP medium. The pullulan-to-substrate yield was 0.22 g g⁻¹ for both. An increased glucose concentration led to the increased formation of storage substances such as pullulan, intracellular lipids, and polyol lipid. In both cultivations, the pH oscillated during the growth phase (Figure 7F). This was due to the metabolization of NH₄NO₃. Successively, the weak base (NH₄ ⁺) and the weak acid (NO₃ ⁻) were consumed. In addition, possibly released polymalate has a free acid group and thus might additionally lower the pH.

Comparing cultivations in MTPs and the bioreactor, growth rates, polyol lipid formation, Y_P/S, Y_P/X, and space-time yields (STY) were in a comparable range (Table 7). Regarding polyol lipid production, the upscaling had no effect when MLP was used and a slightly negative effect in the case of oMLP. It is striking that the switch from a shaken to a stirred system led to an almost threefold increase in the pullulan titer (9.3 g L⁻¹ versus 27.1 g L⁻¹). This effect has already been described in literature. Pullulan production can be influenced not only by the pH value but also by the agitation rate or energy input, whether positive or negative, depending on the strain. In a batch fermentation with A. pullulans ATTC9348, a tenfold increase in agitation rate led to a halving of the pullulan titer (Gibbs and Seviour, 1996). In contrast, in the fermentation of an A. pullulans strain that is not described in detail, the pullulan titer increased with increasing stirrer speed (McNeil and Kristiansen, 1987). In terms of biomass, comparing the cultivation in MTPs with the MLP medium in the bioreactor, nearly constant biomasses of 9.5–10 g L⁻¹ were reached. A decrease in Y_P/X was probably due to increased pullulan and storage lipid production.

All in all, the developed and optimized minimal medium was successfully transferred from MTPs to the bioreactor with a scaling factor of 350. This resulted in 41 g L⁻¹ polyol lipid, which is the highest polyol lipid titer reported with an A. pullulans wildtype strain using a laboratory-scale bioreactor.