Switching industrial production processes from complex to defined media: method development and case study using the example of Penicillium chrysogenum
- First Online:
Filamentous fungi are versatile cell factories and widely used for the production of antibiotics, organic acids, enzymes and other industrially relevant compounds at large scale. As a fact, industrial production processes employing filamentous fungi are commonly based on complex raw materials. However, considerable lot-to-lot variability of complex media ingredients not only demands for exhaustive incoming components inspection and quality control, but unavoidably affects process stability and performance. Thus, switching bioprocesses from complex to defined media is highly desirable.
This study presents a strategy for strain characterization of filamentous fungi on partly complex media using redundant mass balancing techniques. Applying the suggested method, interdependencies between specific biomass and side-product formation rates, production of fructooligosaccharides, specific complex media component uptake rates and fungal strains were revealed. A 2-fold increase of the overall penicillin space time yield and a 3-fold increase in the maximum specific penicillin formation rate were reached in defined media compared to complex media.
The newly developed methodology enabled fast characterization of two different industrial Penicillium chrysogenum candidate strains on complex media based on specific complex media component uptake kinetics and identification of the most promising strain for switching the process from complex to defined conditions. Characterization at different complex/defined media ratios using only a limited number of analytical methods allowed maximizing the overall industrial objectives of increasing both, method throughput and the generation of scientific process understanding.
KeywordsFilamentous fungi Complex media Defined media Stoichiometric mass balancing Fast strain characterization
Carbon dioxide evolution rate
Critical process parameters
Critical quality attributes
Degree of reduction
h-value, statistical test value
International conference on harmonization
Number of balances
Key performance parameters
Number of measured rates
Overall number of rates
Percentage of complex nitrogen
Quality by Design
Recovery coefficient for species n
Degree of redundancy
c-molar degree of reduction for biomass
c-molare nitrogen content for biomass
variance-covariance matrix of the residues
variance-covariance matrix of the rates.
Filamentous fungi have been used for the large scale production of antibiotics, organic acids, enzymes and other industrially relevant compounds for many decades [1, 2]. However, due to the increased complexity of filamentous fungal cultivation systems compared to unicellular yeast or bacterial systems, including enhanced media demands during spore germination, adherent wall growth, extracellular proteolytic activity, formation of various side-products and complex growth morphology, process engineers still commonly rely on extensive experience and empiricism during bioprocess development and control rather than using science-based approaches [3, 4].
Lacking scientific understanding about interdependencies of critical process parameters (CPP), critical quality attributes (CQA) and key performance parameters (KPP), manufacturers conventionally stick to well-established complex media fermentation protocols to ensure growth and productivity, and thus fail at demonstrating scientific understanding for their thoroughly, though mostly empirically optimized, high-yielding production processes. However, considerable lot-to-lot variability of complex media ingredients not only demands for exhaustive incoming components inspection and quality control, but unavoidably affects process stability and performance . Additionally, the use of complex media components prevents stoichiometric and physiological determination of the bioprocess, resulting in uncontrolled process conditions that inhibit timely accurate detection of physiological changes, as well as the application of appropriate control strategies. Unarguably, hardly controllable process conditions (whether or not due to complex raw materials) inevitably reduce process yields compared to optimally controlled conditions. Moreover, not only the variability of complex raw materials, but also regulatory guidelines in context with the Quality by Design (QbD) initiative issued by the U.S. Food and Drug Administration strongly encourage biopharmaceutical manufacturers to move away from traditional empirically based bioprocess development towards science-based process design and control . As a direct consequence, bioprocesses using complex media ingredients should be switched to defined media.
Up to date, proposed methods for switching microbial production processes to defined media include statistical experimental designs aiming at media optimization [7, 8] as well as detailed analysis of complex media components and substitution of the relevant key components in a chemically defined form . Another interesting approach suggests chemical defined media formulation based on flux analysis through the metabolic network reconstructed from the organisms genome sequence . However, these approaches were only described for simple unicellular organisms and not yet for the much more complex fungi which exhibit more complicated strain specific nutrient demands for growth, productivity and germination.
Unfortunately, the most straight-forward solution for switching fungal strains from complex to defined media, namely a simple parallelized strain characterization on defined media, would most likely not only result in a lack of growth for most of the screened fungal strains and thus only yield limited scientific process understanding, but also in highly increased time demands compared to characterization procedures on complex media. Besides, characterization on fully complex media can only be accomplished using exhaustive, time-consuming offline analysis procedures including acid hydrolysis of insoluble nitrogen species as well as macromolecular soluble species . Moreover, having inoculated the fermentation process, insoluble media components cannot be separated from biomass elements and are therefore not quantifiable. Encouraged by these inherent problems in bioprocess analysis for complex, insoluble media compounds, we developed a simple strategy to infer strain specific complex media components uptake kinetics by applying a combined complex/defined media fermentation strategy.
The approach of investigating macroscopic balances for stoichiometric bioprocess modeling was presented already over 30 years ago [12, 13]. Over the last decade, we could successfully demonstrate general method applicability for batch, fed-batch and continuous cultivations of bacterial and yeast systems [14, 15, 16]. So far however, method dissemination to filamentous fungi and complex media fermentation has been hampered by the difficulty to accurately determine the biomass dry weight concentration for adherently growing organisms in cultivations containing insoluble media components, and the increased complexity of filamentous fungal systems.
In this work, we present a methodology for fast strain characterization of filamentous fungi in batch cultivations on combined complex/defined media. This approach allows the identification of the most promising candidate strain(s) for switching the bioprocess from complex to fully defined media. In the context of this study, method applicability was demonstrated by characterizing two different industrial Penicillium chrysogenum candidate strains. The developed methodology is based on minimal analytical needs and the indirect determination of specific uptake rates of complex media components by applying statistically verified redundant mass balancing techniques for carbon, nitrogen and electron balances. Moreover, we propose a strategy facilitating accurate determination of biomass dry weight concentrations throughout the process by successfully preventing adherent wall growth via headspace cooling and keeping the liquid level constant by refilling the withdrawn volume during sampling.
Using the suggested approach for strain characterization in batch cultivations on partly complex media and elucidation of strain specific physiological parameters for complex media components, we could demonstrate accurate and feasible bioprocess characterization for industrial applications at reduced analytical demands in accordance with QbD principles.
Results and discussion
Theory and modeling
As concentrations, uptake rates and yields for defined media components can be determined by standard analysis methods, uptake rates and yields for non-measurable media components may be inferred from mass balancing according to the following equations. Calculations of carbon, degree of reduction and nitrogen balances as well as recovery coefficients are described in equations 3 to 11.
Degree of reduction balance
Equation 19 shows the calculation of residuals using actual and thus inherently noisy measurements, while for each individual rate the expected error is specified in the variance-covariance matrix Ψ of the rates which are assumed to be non-correlated. The statistical test parameter h is calculated according to equations 20 and 21 with Φ as the variance-covariance matrix of the residuals. Generally, the hypothesis of having accurately determined the system needs to be rejected if the h-value is higher than 2.71 (according to the chi-square distribution) with a degree of freedom of 1 (S = 1) and a confidence interval of 90% and 4.61 with a degree of freedom of 2 and 6.25 with a redundancy of 3, respectively. Thus, the above outlined approach allows judging raw data quality. Unless the test parameter h exceeds its threshold acceptance value, it can be assumed that the system was accurately determined. All data discussed in this study met the criteria assuming a maximum relative error of 5% during the exponential growth phase. This assumed maximum error was specified in the variance-covariance matrix Ψ for the rate vector.
For a detailed and exhaustive, step by step description of theoretical mathematical aspects of elemental balancing and methods for gross error checking, the reader is referred to benchmark studies within this field [18, 19, 20]. Finally, using the above outlined approach analytical demands were strongly reduced by inferring uptake kinetics of complex nitrogen components from mass balancing. Moreover, accurate and feasible strain and process characterization were facilitated.
Summary of Key Performance Parameters for all batch characterization studies
Complex media fraction/Key performance parameter
Strain characterization on partly complex media
Verification on defined Media
Duration Lag-Phase [h]
max. specific formation rate for gluconic acid [g·g-1·h-1]
max. specific formation rate for mannitol [g·g-1·h-1]
Formation of fructooligosaccharides
Percentage of nitrogen content at the process end [%]
Cultivation of BCB1 on 100% complex media
As shown in Figure 2c, during the initial lag-phase, production of GF2 occurred up to an observed maximum yield of 0.7 Cmol·Cmol-1 initial substrate. The fructooligosaccharide was subsequently taken up during the exponential growth phase, whereas the gluconic acid yield increased until the end of the cultivation reaching a maximum of 0.4 Cmol·Cmol-1initial substrate. Mannitol on the other hand was produced to a minor extent with a significantly lower maximum yield compared to gluconic acid of approx. 0.08 Cmol·Cmol-1substrate, an observation which held true for all cultivations. Thus, strain characterization performed in this study focused on gluconic acid as the major side-metabolite. Generally, avoiding fructooligosaccharides and gluconic acid as unwanted side-products would be highly desired as their formation reduces the biomass space time yield and directly influences process performance. During the exponential growth phase the maximum specific growth rate was ~ 0.18 h-1 and the maximum formation rate of the major side-metabolite gluconic acid was approx. 0.18 Cmol·Cmol-1·h-1 (Figure 2d). The decrease in the specific formation rate for gluconic acid towards the end of the cultivation correlated with the increase of the respiratory quotient (RQ) and thus could be observed online in real-time (Figure 2c and 2d). The decrease in the specific growth rate from −10 h onwards (Figure 2d) might be explained by fungal biomass pellets having reached substrate diffusion limiting diameters and thus no longer exhibiting exponential growth characteristics .
Starting from cultivation on 100% complex media, the complex media fraction was reduced in the following experiments (vide infra). Consequently, strain characterization and process identification on fully complex media serves as benchmark analysis yielding the maximum rate vector x (Eq. 13), which is then used in mass balancing and data evaluation (Eq. 2–21).
Cultivation of BCB1 on 25% complex media
Kinetics of gluconic acid formation and complex nitrogen uptake showed highly similar trends suggesting a negative correlation between the production of the overflow metabolite and complex growth conditions (Figure 3d). Furthermore, the online detectable parameter of the RQ reflected kinetics of gluconic acid formation as it approached 1 as soon as production of gluconic acid was ceased (Figure 3c and 3d). As shown in Figure 3d, the specific uptake rate of complex nitrogen sources surprisingly increased again at the very end of the cultivation. We believe that due to the depletion of sucrose as available C-source (Figures 3a and 3b), the fungus resorted to the amino acids (i.e. complex N-sources) as alternative C-source.
Due to a prolonged duration of the lag-phase (23 hours compared to 17 hours for the cultivation on fully complex media; Table 1), the overall duration of the process was increased from 39.3 to 49 h, while the maximum specific growth rate was in a similar range (i.e. 0.18 h-1 and 0.15 h-1, respectively). As for the cultivation on fully complex media, the reduced specific growth rate at cultivation end compared to −10 h (Figure 3d) may be explained by fungal biomass pellets having reached substrate diffusion limiting diameters.
Cultivation of BCB1 on fully defined media
Comparing strains BCB1 and P2
Within this study, both P. chrysogenum candidate strains BCB1 and P2 were characterized on 100% and 25% complex media in order to infer process performance on fully defined media. In addition, verification cultivations were performed on purely defined media for both fungal strains BCB1 and P2. Results from all cultivations are compared in Table 1.
Formation of unwanted side-products
the specific production rate of gluconic acid can be minimized for strain BCB1 by reducing the fraction of complex media
formation of oligosaccharides can be prevented for strain BCB1 by reducing the fraction of complex media
Hence, starting from two candidate strains, the bioprocess could be successfully transferred to defined media conditions by selecting the more promising strain BCB1 using the outlined methodology. Finally, for the defined media cultivation of BCB1, the highest specific growth rate (0.19 g·g-1·h-1) was observed, while only very limited production of mannitol as the only side-product occurred (0.06 g·g-1·h-1).
For both strains, process duration increased with decreasing fractions of complex media. Considerably prolonged lag-phases were identified as the major causes for the increase in the overall batch duration (Table 1). In general, during germination the culture exhibits extended nutrient demands and thus benefits from complex media ingredients to a larger extent. However, we believe that the reduced demands for raw material quality control, as well as controllable, defined cultivation conditions outweigh the increase of the batch cultivation time on fully defined media. The actual process duration may not be influenced by the choice for defined media, because the process is at least a 2 stage process, for which equipment utilization and overall process productivity can be optimized by time and motion studies. Moreover, if, despite the possibility of timely staggering the usage of batch and fed-batch reactors and thus absorbing increased process duration for batch cultivations, the increased duration time of the batch process is a major issue, we suggest to implement a 3-stage process consisting of a complex media germination and a defined media batch and fed-batch stage.
Assimilation of the defined nitrogen source
The cumulated nitrogen fraction in the biomass at cultivation end originating from complex media constituents was considerably lower for BCB1 than for P2 when cultivated on 25% complex media (POCN of 23 and 33%, respectively). A consistent trend could be observed for the novel strain specific key parameter POCN (Eq.1) introduced in this study for both strains when cultivated on 25% complex and the verification runs on defined media. Therefore, the suggested key parameter allowed quantification of the strain specific affinity towards defined nitrogen sources and thus an easy identification of the most promising strain for switching the process from complex to defined cultivation conditions. The increased affinity of strain BCB1 to defined nitrogen sources on partly complex media, together with the reduced formation of oligosaccharides (Table 1 and Figure 3c and 4c) identified BCB1 to be superior over P2 for switching the process to defined cultivation conditions. For this reason, a fully defined 2-stage fed-batch production process was performed with strain BCB1.
Fully defined 2-stage production process for strain BCB1
The applied strategy for strain characterization of filamentous fungi on partly complex media at different complex/defined media ratios using redundant mass balancing techniques facilitated the elucidation of physiological kinetics for complex nitrogen species uptake mechanisms and growth characteristics on complex media. Interdependencies of specific biomass and side-metabolite production rates, formation of fructooligosaccharides, specific complex media components uptake rates and fungal strains were revealed. The novel strain specific key parameter “percentage of complex nitrogen of the overall nitrogen uptake (POCN)” enabled straight-forward identification of the most promising strain for switching the process from complex to defined conditions. Strain characterization on combined complex/defined media only asked for limited analytical methods and thus allowed maximizing the overall industrial objectives of increasing both, method throughput and the generation of scientific process understanding.
Strain characterization and process identification on fully complex media for benchmark analysis of the system’s maximum complexity. Cultivations should be conducted with top aeration during germination as well as cooling tubes at the reactor headspace and refilling sampling volumes to reliably determine the fungal biomass dry weight. Subsequent and easy-to-do mass balancing based on the maximum rate vector x identified in this step then allows the elucidation of complex nitrogen components uptake as well as side-product formation kinetics.
Batch cultivations on 25% complex media to determine strain specific uptake rates for complex and defined nitrogen compounds and the percentage of complex nitrogen of the overall nitrogen uptake (POCN). The lower the identified POCN, the more suitable is the strain for switching the process from complex to defined media.
Switching the process to a defined cultivation strategy with the most promising fungal strain.
Applying this approach, we obtained a 2-fold increase of the overall penicillin space time yield and a 3-fold increase in the maximum specific penicillin formation rate with the candidate strain BCB1 on fully defined media formulation compared to a 2 stage complex/defined fermentation strategy. Consequently, this approach is not only particularly suitable for industrial strain characterization applications where throughput and time are of major concern, but can also lead to more efficient bioprocesses.
Spore suspensions of two P. chrysogenum candidate strains for penicillin production (code BCB1 and P2), which are proprietary, engineered strains, were kindly provided by Sandoz GmbH (Kundl, Austria).
The investigated bioprocess displayed an industrial scale penicillin production process and comprised a batch fermentation on complex media with corn steep liquor as nitrogen source followed by a fed-batch process for penicillin production using a defined media feeding strategy and ammonium sulfate as defined nitrogen source.
The complex media was based on the media used in  and consisted of: Sucrose 18 g·l-1, Glucose 3 g·l-1, Corn steep liquor 26 g·l-1, Silicone Oil 1 ml·l-1. Composition for the defined media was as follows: Glucose 30 g·l-1, (NH4)2SO4 8.75 g·l-1, KH2PO4 1.6 g·l-1, NaNO3 0.2 g·l-1, KCl 0.5 g·l-1, CaCl2 × 2H2O 0.067 g·l-1, MgSO4 × 7H2O 0.5 g·l-1, TES-Stock 10 ml·l-1, Silicone Oil ml·l-1. TES-Stock consisted of: EDTA 14 g·l-1, CuSO4 × 5H2O 0.5 g·l-1, ZnSO4 × 7H2O 2 g·l-1, MnSO4 x H2O 2 g·l-1, FeSO4 × 7H2O 4 g·l-1. For both strains, characterization was performed in combined complex/defined batch cultivations at 100%, 25% and 0% complex media content. With a determined amino acid content of approx. 25% [w·w-1 for the corn steep liquor used in this work the C-molarity of batch media was always in the range of 0.85 to 1 Cmol·l-1.
Batch fermentations were performed in a 1.8-l stirred bioreactor (Applikon, The Netherlands), with the actual working volume of 1.5 liters. Cultures were aerated through a standard single port sparger located below the stirrer at a constant aeration rate of 0.8 vvm. Aeration air was sterilized through PTFE air filters of 0.2 μm pore size (Whatman, UK). Aeration flow was kept constant using the mass-flow-controller system 2Proc (Aalborg, USA). To prevent blow-out of spores, mode of aeration was switched from headspace to submerse after completed germination only. Water stripping from the reactor was prevented by the use of an off-gas condenser kept at a temperature of 8°C by means of a cryostat. The temperature of the culture was kept at 25°C by an external heat jacket, which was heated/cooled using a thermo circulator. The pH of the culture was strictly kept at 6.5 +/− 0.1 by addition of 2.5 N NaOH. Dissolved oxygen tension and pH were measured using a sterilizable pO2 and pH probe, respectively (both Mettler-Toledo, USA). Agitation was performed using three six-bladed Rushton turbine impellers. Rotation speed was controlled in order to guarantee sufficient dissolved oxygen tension. Throughout all experiments, the dissolved oxygen tension was controlled at 40% by adjusting the agitator speed. All process control measures were performed by the integrated process control and management system Lucullus (Biospectra AG, Switzerland). Fermentations were inoculated with spores from rice cultures. For all cultivations inoculum concentration of living spores was 1.2 × 109 l-1. Foaming was prevented by addition of small quantities of PPG 2000 (up to 1 ml).
The investigated complex media can be considered a typical empirically optimized, industrially used batch media for Penicillium spp. Addition of corn steep liquor accelerates germination of the fungus and thus reduces the lag-phase. Glucose was replaced with sucrose as the major carbon source in order to suppress significant formation of gluconic acid , however presence of sucrose is known to cause oligosaccharide formation [29, 30].
Fed-batch cultivations were carried out in a 7.5-l stirred bioreactor in defined growth media (Infors, Switzerland). Media composition was as follows and based on the media used in : Glucose 1 g·l-1, (NH4)2SO4 7 g·l-1, KH2PO4 1.6 g·l-1, FeCl3.6H2O 0.02 g·l-1, MgSO4.7H2O 0.1 g·l-1, KCl 0.5 g·l-1, PPG 2000 0.05 ml·l-1, CuSO4.5H2O 5.5 mg·l-1, ZnSO4.7H2O 35.6 mg·l-1, MnSO4.H2O 29.5 mg·l-1, CaCl2.2H2O 65 mg·l-1. Glucose was fed at 2.65 ml·l-1·h-1with a feed concentration of 500 g·l-1. Feeding of a 200 g·l-1 (NH4)2SO4 solution, representing the N-source in defined media, was started after 5 h with a rate of 0.9 ml·l-1·h-1, was subsequently set to 1.5 ml·l-1·h-1 from 15–50 h, to 1.35 ml·l-1·h-1 from 50–70 h and to 1.2 ml·l-1·h-1from 70 h until process end. Sodiumphenoxyacetate as Penicillin V precursor was fed at 0.4 ml·l-1·h-1 with a feed concentration of 160 g·l-1. Process duration was 72 hours. Fed-batch processes were inoculated with 160 ml/l cultivation broth from batch processes as soon as carbohydrate exhaustion was indicated by an pH increase of 0.5. Cultures were aerated through a standard multiport sparger located below the stirrer at a constant aeration rate of 1.35 vvm (with respect to the starting volume). The pH of the culture was kept at 6.5 +/− 0.1 by addition of 5 N NaOH and 15% H2SO4; otherwise fermentation set-up and control was identical to batch cultivations (vide supra).
Determination of biomass dry weight was performed in duplicates by pressure aided filtering of 10 ml of culture broth on pre-dried and pre-weighted Pall glass fiber filters type A/E (Pall, USA). Prior to drying, filters were washed twice with 20 ml deionized H2O for cultivations on defined media, or, when beneficial for removal of insoluble media constituents, with 5 ml 2% HCl and 5 ml Acetone first, secondly with 5 ml Acetone and thirdly with 10 ml H2O.
Analysis of glucose, TCA-intermediates, organic acids and sugar alcohols was performed by high pressure liquid chromatography, using an Agilent Technologies Series 1100 HPLC with RI and DAD detector (Agilent, USA) via isocratic elution with 0.1% H3PO4 on a SUPELCOGEL™ C-610 H column (Supelco, Sigma-Aldrich, USA).
Detailed quantitative analysis and identification of organic components present in the fermentation broth (including organic acids, carbohydrates and di/tri-saccharides) was performed by an Agilent Technologies 7890a gas-chromatography system coupled to a 5975 C XL MSD mass-spectrometer (Agilent, USA). Samples were dried in the vacuum oven for 20 min at 40°C. 200 μl pyridine were added for quantitative dissolution. Samples were silylated with 150 μl hexamethyldisilazan (HMDS) and 70 μl trimetyl-chlorsilan (TMCS). Samples were then shaken for dissolution while bigger aggregates were split in an ultra-sonic bath. Derivatization was performed at least for 4 hours, preferably overnight. 100–150 μl were subsequently transferred to GC vials using insert tubes. Analysis was performed on a Helium Column: HP-5, length: 30 m, diameter: 0.32 mm, 0.25 μm film thickness using the following temperature program: 150°C for 1 min; increase to 220°C at 4°C·min-1, then to 320°C at 20°C·min-1. Hold at 320°C for 6.5 min. Injector 260°C, Detector 300°C. Split 10:1. Injection volume 1 μl. Average velocity: 30 cm·sec-1. Oligosaccharides were also analyzed as their monomeric constituents using the above described method after acid hydrolysis in 4 M H2SO4 at 100°C for 4 hours.
Analysis of penicillin V and phenoxyacetic acid was performed by high pressure liquid chromatography using a ZORBAX C-18 Agilent Column (Agilent Technologies, USA) and 28% ACN, 6 mM H3PO4, 5 mM KH2PO4 as elution buffer.
Quantitative enzymatic analysis of ammonia was performed using a commercially available assay (Randox, USA) on the Cubian XC enzymatic robot (Innovatis, Roche Diagnostics, Switzerland).
Biomass elemental composition needed for stoichiometric balancing was analyzed at two time points for all cultivations by determining C, H, N and O content at the Microanalytical Laboratory of the University of Vienna (Austria).
For all offline analyses, samples were immediately put on ice after withdrawal from the reactor and stored at - 20°C until further analysis for a maximum duration of two weeks.
We want to thank Sandoz GmbH (Kundl, Austria) for kindly providing the strains used in this study and Vienna University of Technology (VUT) for partial funding of A.E. Posch in course of the Applied Bioscience Technologies PhD program.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.