OR WAIT null SECS
Amit Banerjee, PhD, is a research fellow at Pfizer Global Research & Development
The authors describe challenges faced in transfer and scale-up of a fermentation process.
A fermentation process for the manufacture of virus like particles (VLPs) in Escherichia coli (E. coli) was transferred from an external collaborator and rapidly implemented in Pfizer's cGMP pilot plant. Challenges faced in the transfer were meeting the high oxygen demand of the original process, and attempting high density cultivation of E. coli in a bioreactor system primarily designed as a seed tank for larger-scale mammalian and microbial culture. These concerns were overcome by an approach that combined process and equipment characterization, allowing suitable adaptation of the process to fit the pilot facility.
Technology transfer of manufacturing processes for recombinant proteins frequently involves the reassessment of equipment capabilities and corresponding process requirements. In this context, this article describes the transfer of a fermentation process for the manufacture of Q-beta virus like particles (VLPs) in Escherichia coli (E. coli) from an external collaborator to Pfizer's cGMP pilot plant. In preparation for the technology transfer, establishment of appropriately organized teams as well as effective communication at all levels were key (see Figure 1). These early teams included both process development and analytical development scientists, since analytical assay support for an atypical product such as the Q-beta VLP was considered important for the success of this endeavor.
Figure 1: Technology transfer team structure. (ALL FIGURES ARE COURTESY OF THE AUTHORS)
The Q-beta VLP is an antigen delivery platform designed to serve as a key component of therapeutic and prophylactic vaccines (1–4). Each Q-beta VLP consists of 180 copies of a single coat protein from the Allolevivirus Q-beta, which together form an icosahedrally symmetric particle stabilized by disulfide bonds (5). The first step of Q-beta VLP production involves cytoplasmic expression of the coat protein (Q-beta monomer) in E. coli. Subsequent processing and purification steps ensure the formation and recovery of mature VLPs. Technology transfer challenges related to the Q-beta VLP manufacturing fermentation process included: a change in scale, significant equipment-related differences, and the high oxygen demand of the original process.
Initially, the VLP fermentation process was transferred and implemented at the laboratory scale, which allowed further process characterization data to be collected. Following this, equipment characterization was undertaken to assess the capabilities of the cGMP pilot plant bioreactor system and compare it with the laboratory-scale system. Based on this information, required process adaptations were reviewed with our collaborators, and the process was successfully implemented in the cGMP pilot plant.
Strain and expression system
The strain used for production of the Q-beta VLP was an E. coli K-12 derivative that constitutively overexpresses the lacI repressor. The target protein, Q-beta monomer, was expressed under the control of the hybrid tac promoter induced by lactose or isopropyl beta-D-1-thiogalactopyranoside (IPTG) in a high copy-number plasmid. Plasmid selection pressure was maintained by a kanamycin resistance gene (6). The E. coli production strain was transferred to Pfizer as vials from a cGMP working cell bank (WCB) containing a low-density cell suspension in culture medium with cryoprotectant.
The fermentation process consisted of a seed culture stage and a fed-batch fermentation stage. The same yeast extract-enriched medium containing glycerol as the primary carbon source was used for cell bank preparation, seed culture preparation, and the fermentation batch phase (see Table I) (6, 7). For the fermentation fed-batch phase, two different fermentation feed media were used; both were similar in composition to the batch medium, but contained either a high concentration of glycerol alone or glycerol and lactose (see Table I) (6). Lactose was used as a carbon source and as the inducer.
Table I: Fermentation Medium.
Samples were collected throughout the fermentation to assess culture growth and determine Q-beta monomer titer. Cell growth was examined by measuring absorbance at 600 nm (OD600), as well as determining dry cell weight with a 1 mL cell broth sample. Titer was determined by quantifying the amount of Q-beta monomer present in the soluble fraction of a lysed cell pellet using reverse phase-high performance liquid chromatography (RP-HPLC). Any fully or partially formed VLP present in the extract was first reduced to its monomer form by incubation with a reducing agent. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) analysis was also used to ascertain monomer identity based on comparison with a standard. Cell-free spent fermentation media samples were used to measure glycerol and lactose concentrations. Glycerol content was estimated using a Nova Bioprofile 300 Analyzer (Nova Biomedical, Waltham, MA) and lactose was measured using a YSI 2700 Biochemistry Analyzer (YSI Inc., Dayton, OH). Off-gas analysis was carried out during the fed-batch fermentation runs using Tandem off-gas analyzers by Magellan Instruments (Limpenhoe, UK) to obtain an estimate of the oxygen uptake rate (OUR). After subsequent processing and purification, the quality of the intact Q-beta VLP was assessed by several methods, including analytical size exclusion chromatography (SEC). The titer and product quality assays were transferred from our collaborator or developed in house.
The objective of first transferring this process into the laboratory was to enable characterization of the fermentation protocol and allow the production of Q-beta VLP test batches for analytical characterization or development of the VLP purification process. It was understood that the VLP fermentation process implemented in the laboratory might require further adaptation to fit into Pfizer's cGMP pilot plant, as determined by the available equipment.
Starting from a single cGMP WCB vial, a high-density research and development cell bank was quickly created for laboratory use (see Figure 2). Frozen cell suspension from the cGMP WCB vial was used to inoculate a small volume of medium in a 250 mL shake flask; this culture was then used to inoculate a 3-L shake flask. When the culture reached the desired OD600, the cells were harvested by centrifugation and resuspended at a high density in fresh medium containing glycerol (20% v/v) as a cryoprotectant. Aliquots were frozen at –80 °C for long term storage. High density cell banks with cell concentrations of the order of 1 x 109 colony forming units were used to minimize seed vial requirements per run and ensure consistency with respect to seed culture growth.
Figure 2: Preparation of laboratory (non-cGMP) and cGMP cell banks for Q-beta virus-like particle (VLP) manufacture. Lab cell bank was tested for bacteriophage and purity only; this cell bank was used at risk until test results were available. The cGMP cell bank preparation did not include the Shake Flask 1 stage.
Original fermentation process
The Q-beta VLP fermentation process to be transferred from our collaborator was a high-density fed-batch cultivation process that was inoculated using a shake flask seed culture. The fermentation itself consisted of three phases: a batch phase, an exponential fed-batch phase, and an induction fed-batch phase (see Figure 3). The batch phase lasted until glycerol in the batch medium was consumed, and was immediately followed by the exponential fed-batch phase, that involved the introduction of glycerol feed at a preset, exponentially increasing rate. Exponential feeding was continued for a set duration to achieve a target biomass concentration, at which time the glycerol feed medium was replaced with induction medium containing glycerol as well as lactose, and the induction fed-batch phase commenced. The induction medium was fed at a constant rate, equal to the feed rate at the end of the exponential feeding phase. The induction phase lasted a predetermined duration, after which the culture was harvested. This protocol was repeatedly shown to achieve a high titer of the Q-beta monomer along with a high biomass concentration (6). Significant biomass accumulation occurred during the induction phase despite the initiation of recombinant protein production. An oxygen-enriched air stream was used to satisfy the high oxygen demand of the culture during fermentation.
Figure 3: Feed profiles for Q-beta virus-like particle (VLP) fermentation. The solid line shows original feeding protocol. Feed I was the glycerol feed added at an exponential rate. Feed II was the glycerol/lactose feed added at a constant rate for induction. Dotted line shows the manner in which exponential feeding was extended for process characterization.
Process characterization and development
Laboratory scale fermentations were carried out in a 15-L working volume, computer-controlled Sartorius Biostat C DCU3 fermentation system. Operating conditions for the 15-L scale fermentations were derived from the original process conditions, taking equipment limitations and prior fermentation experience into account (6). For all laboratory fermentations, the dissolved oxygen (DO) set point was 20%, with cascade control by agitation (500–1200 rpm) and airflow (10–15 Lpm). The temperature and pH were controlled at 30 °C and 6.8 respectively. Operating pressure was increased with respect to the original process to increase oxygen solubility, because a pure oxygen stream to enrich the process air was unavailable.
A staged approach was followed for the technology transfer, which was intended to help understand the fermentation process requirements, such as oxygen transfer, and assess the laboratory fermentation system's ability to meet these requirements. This enabled adaptation of the original process for implementation in the laboratory without the use of oxygen, while maintaining productivity as close to the original process as possible. In practice, the staged approach was implemented by first performing uninduced batch fermentations, followed by uninduced fed-batch fermentations and finally induced fed-batch fermentations. Data from these runs drove discussions on process adaptation.
Batch fermentations were conducted to determine the biomass concentration after all the glycerol present in the batch medium was consumed. This data was required for calculation of the feeding profile. The batch fermentation medium contained 5 g/L glycerol as the main substrate for growth; however, additional substrate was also available due to the presence of yeast extract (6). At the 15-L laboratory scale, a starting batch volume of 10 L was used, and the batch was inoculated with 2.4% v/v shake flask preculture broth. During the batch fermentations, all glycerol in the medium was consumed in 7–8 h (data not shown).
Fed-batch fermentation with exponential feeding
Exponential fed-batch fermentations were designed to assess culture oxygen requirements and determine the maximum oxygen uptake rate that could be supported by the laboratory fermentation system. This would be observed on the DO profile as a sustained decrease in the DO value below the 20% set point. These fermentation runs included a batch phase as described earlier, followed by a fed-batch phase that involved the addition of glycerol feed medium at an exponentially increasing rate. The equation used to define the exponential feeding profile has been well documented in literature (6–9):
where, V(t) is volumetric flow rate of feed medium at time t (L/h), µ is specific growth rate (1/h), Vt,f is culture volume at feeding start (L), Xt,f is biomass (cell dry weight, g/L) at feed start (determined from batch fermentation), t is process time (h), tf is process time at feed start (7–8 h), Sf is concentration of glycerol in the feed medium (g/L), and YX/S is glycerol yield coefficient determined from batch fermentations (g dry cell weight / g of glycerol at the batch phase end).
It was observed that after about 8.5 h of feeding, to maintain the dissolved oxygen set point at 20%, the agitation and airflow reached maximum values of 1200 rpm and 15 Lpm respectively, and that beyond this time the set point could not be maintained (see Figure 4A). Taking this into account, the duration of the exponential feed phase was adjusted to 8 h for future laboratory runs, which was shorter than described in the original process but deemed satisfactory to allow sufficient biomass accumulation for induction. The OUR at this time was measured at approximately 230 mM/L–h.
Figure 4: Q-beta virus-like particle (VLP) fermentation at the lab scale. (A) Characterization experiments to observe process oxygen requirements - dissolved oxygen and oxygen uptake rate profiles for fermentations conducted with the extended exponential feeding. (B) Dissolved oxygen and oxygen uptake rate profiles for modified fermentation process with shortened exponential feed phase to decrease peak oxygen demand. Colors show dissolved oxygen profiles for different batches.
Fed-batch fermentation with induction
These fermentation runs were conducted to complete the laboratory-scale fermentation process transfer and assess final Q-beta VLP production with a shorter (8 h) exponential feeding phase prior to induction. The batch and fed-batch phases were conducted as described earlier. At the end of the exponential feed stage, glycerol feed medium was replaced with the induction medium containing glycerol and lactose. The induction medium was added at a constant feed rate equal to the feed rate reached at the end of the exponential feed phase. The induction phase duration was maintained at 5 h as described for the original process. The culture density increased significantly during the induction phase, and the cell density at harvest was more than twice that at induction. The OUR decreased after induction and remained at approximately 150 mM/L–h (see Figure 4B). Some lactose accumulated during the early part of the induction phase as the culture adapted to this new carbon source (data not shown). As expected, the harvest cell density and final titer were approximately 80% and 85% of the corresponding values for the original process, respectively (see Figure 5A). The specific productivity, calculated as grams Q-beta monomer per grams dry cell weight, corresponded very well with the original process, and product quality after purification was also satisfactory (see Figure 5B and 6A).
Figure 5: Culture growth and productivity data at the laboratory scale. (A) Growth for the adapted fermentation and original process. Original process is in red. (B) Productivity of the adapted process was comparable to the original process.
Fermentation process transfer to the laboratory was completed in approximately three weeks, including preparation of the development cell bank. A well-characterized small-scale fermentation process that fully utilized the capabilities of the 15-L laboratory fermentation system was now available to produce test batches of Q-beta VLP for development.
Figure 6: (A) Adapted process showed comparable product quality at the lab scale. Analytical size exclusion chromatography (SEC) was one of several methods used to test VLP quality. SDS-PAGE was used to observe the monomer (inset). (B) Adapted process showed comparable product quality at the pilot scale. Analytical SEC and SDS-PAGE were among the assays used to used to check VLP quality (inset).
The cGMP pilot plant fermentation system had a 100 L maximum working volume and had previously been used as a seed tank to support larger scale mammalian and microbial cultivation. Thus, it was expected that this system would have certain operating constraints that might require further adjustment of the laboratory-scale process prior to implementation. To prepare for the transfer, the pilot plant bioreactor was characterized and compared to the laboratory system.
The production of a high-density cGMP cell bank was initiated prior to the planned cGMP fermentation. The protocol used for creation of the cGMP cell bank was similar to that used in the laboratory. However, additional testing of the cell bank for viability, purity (bacterial and fungal contaminants), host strain identity, and absence of bacteriophage was undertaken to ensure compliance with existing guidelines.
Figure 7: Characterization strategy for fermentation equipment.
Equipment-related parameters for characterization were selected based on literature review, past experience with E. coli fermentation, and prior knowledge of the equipment (see Figure 7) (10–16). These included certain theoretical characteristics, as well as experimentally measured characteristics of the fermentation vessels.
Table II: Fermentation vessel specifications.
Characterization of the fermentation equipment was initiated with a review of system specifications (see Table II). Theoretical characteristics of the fermentation systems selected to serve as comparative metrics were geometric similarity, maximum impeller tip speed, maximum power input per unit volume, and superficial velocity at 1 vvm (see Table III) (12, 15, 17, 18). The difference in the geometric similarity parameter for the 100-L vessel and 15-L vessel reflected the difference in aspect ratio of the two tanks. Calculations for impeller tip speed indicated that higher tip speeds were achieved at the 100-L scale due to larger impeller size. To estimate power input per unit volume, a range from 5.2 to 6.5 was assumed for the power number (NP) based on the literature, and both values were used for calculations (17). Again, it was observed that for a given agitation rate, the power input per unit volume was higher for the 100-L fermenter, due to a larger impeller size. The superficial velocity was much lower for the 100-L vessel compared to the 15-L vessel for the same gas flow rate expressed in vessel volume per minute (vvm).
Table III: Summary of characterization results.
Liquid level with respect to internal features was measured with and without agitation or sparging for the two fermentation vessels. This was intended to help determine a starting batch volume that ensured complete immersion of impellers and hence optimum oxygen transfer (see Figure 8).
Figure 8: Bioreactor system characterization. (A) Liquid level with respect to internal features for the 15-L vessel. (B) Liquid level with respect to internal features for the 100-L vessel.
The liquid phase volumetric oxygen mass transfer coefficient (kLa h-1), was experimentally measured to provide a quantitative estimate of the oxygen transfer capability of the system. Since agitation rate and airflow rate are the two most common means of controlling oxygen transfer and hence DO, it was decided to estimate kLa under different combinations of agitation and airflow rates, while keeping all other parameters the same. In order to maintain simplicity and transferability of the test protocol, the static gassing-out method was selected to estimate kLa, with water as the test medium (13, 19, 20). This test involved filling a vessel with water and alternately sparging with nitrogen or air to deoxygenate and re-oxygenate the test medium, respectively. The rate of oxygen re-absorption was recorded with an oxygen electrode and provided an estimate of the kLa through the equation:
where, C is concentration of oxygen in the medium at time 't' (mM/L), t is time in h, kLa is liquid phase mass transfer coefficient (h-1 ), and C* is saturation/equilibrium oxygen concentration in medium under experimental conditions (mM/L).
At the 15-L scale, the experimental conditions selected were between 500–1200 rpm for agitation and 10–15 Lpm for airflow. At the 100-L scale, the experimental conditions selected were between 150–600 rpm for agitation and 60–90 Lpm for airflow. The temperature and pressure were set to operating values for the Q-beta VLP fermentation process. 11 L and 59 L of water were used in the laboratory vessel and pilot plant vessel respectively. Minitab 15 from Minitab Inc. (State College, PA) was used to select the experimental points at each scale and for subsequent analysis of the results. The kLa data at each scale was used to create a response surface with respect to agitation and airflow by employing a full quadratic model. Minitab was then able to calculate the optimum kLa for each response surface, which provided an estimate of the maximum kLa that could be achieved at a particular scale. The response surface model quality was analyzed by examining the normal plot of residuals, as well as the model fit for each point. Maximum kLa values obtained were 137 h-1 for the 15-L scale and 118 h-1 for the 100-L scale (see Figure 9, Table III). A lower kLa at the 100-L scale was not unexpected, as the pilot bioreactor was primarily designed as seed culture vessel rather than a high-density production culture vessel. A quadratic model was used for data interpretation because this was a convenient built in feature provided by Minitab. Other models would also be expected to give similar results, such as the commonly used correlation represented by kLa = k(P0/V)α (Us)β , where k, α and β are empirical constants.
Figure 9: (A) Bioreactor system characterization â determination of kLa in water for the 15-L scale lab vessel. Agitation and airflow were varied, temperature and pressure were set at 0.5 Bar and 30 Â°C respectively. The vessel contained 11 L of water, completely submerging two impellers. (B) Bioreactor system characterization â determination of kLa in water for the 100-L pilot scale vessel. Agitation and airflow were varied, temperature and pressure were set at 0.5 Bar and 30 Â°C respectively. The vessel contained 59 L of water, completely submerging two impellers.
Heat transfer and heat load
To estimate and compare the cooling capacity of the fermentation systems at 30 °C, each vessel was filled with water, and the temperature set-point was repeatedly switched between 40 °C and 20 °C. All other parameters were set to operating values for the Q-beta VLP fermentation process. The recorded temperature profiles allowed calculation of cooling capacity, assuming the specific heat capacity of water at 30 °C as 1 kCal/kg –°C. Cooling capacity of the 15-L fermentation system was estimated at 21 kCal/min using 10 L water with 10 Lpm air flow, 755 rpm agitation at 30 °C (108 kCal/min–m2 of heat transfer area) (see Table II). The cooling capacity of the 100-L fermentation system was estimated at 60 kCal/min using 59 L water with 59 Lpm air flow, and 415 rpm agitation at 30 °C (90 kCal/min–m2 of heat transfer area).
The theoretical heat load was estimated based on OUR (9, 14, 21). For a rapidly growing culture using glycerol, assuming complete oxidation of the carbon source, and that all oxygen transferred into the liquid phase is used for the reaction:
where QH is the metabolic heat released in kCal/L–h, OUR is expressed as mM/L–h, and ΔHGlycerol is heat of formation of glycerol (0.397 kCal/mM). The heat load corresponding to an OUR of 250 mM/L–h was estimated at 5 kCal/min for an 11-L batch and 28 kCal/min for a 59-L batch.
Estimating water loss was considered important because the 100-L vessel did not include an exhaust condenser. Maximum theoretical water loss was estimated using a mass balance, assuming air entering the vessel was bone dry and air exiting the vessel was fully saturated with water vapor. Calculations assumed 59 Lpm air flow at 30 °C and atmospheric pressure. The water loss was also experimentally determined after holding water in the vessel under operating conditions for 16 h (59 L water at 30 °C, 420 rpm agitation, and operating pressure with 59 Lpm air flow). Theoretical evaporative losses were estimated at 0.12 kg/h for the 100-L vessel and actual water loss was measured at 0.06 kg/h after a 16 h hold step (see Table III).
Since the 15-L vessel used an exhaust condenser during normal operation, water losses would be negligible; however, for comparison, a theoretical evaporation rate of 0.019 kg/hr was calculated for the 15-L fermenter (10 Lpm air flow at 30 °C and atmospheric pressure). Actual water loss without the use of the exhaust condenser was measured at 0.0096 kg/h after a 44 h hold step (11 L water at 30 °C, 800 rpm agitation, and operating pressure with 10 Lpm air flow).
Transfer of the adapted process
Based on the equipment characterization results, it was concluded that the pilot scale 100-L fermentation system would be suitable for high density cultivation as long as the oxygen transfer requirements could be met. Since the kLa measured in water for the pilot system was approximately 86% of that for the laboratory scale system, it was assumed that the peak OUR supported by this system would also be lower. Thus, it was decided to decrease the duration of the exponential feed phase to 7 h to lower oxygen requirements. Based on the laboratory scale OUR data, this would decrease the peak oxygen uptake rate to about 80% of that observed in the 15-L fermenter. Again, the induction medium would be added at a constant rate equal to the feed rate reached at the end of the exponential feed phase. Another change was to set the exponential feed start time to 7 h after inoculation. At the laboratory scale, the feed start time varied between 7 and 8 h based on a rapid increase in DO that indicated the complete consumption of glycerol in the batch medium.
Figure 10: Growth data at the pilot scale. Dotted line shows a typical lab scale fermentation. Solid lines show two batches completed at the 100-L pilot scale with reduced feeding time.
Q-beta VLP fermentation runs were successfully executed in the pilot plant after further adaptation of the laboratory-scale process as described above. Growth performance monitored by OD600 as well as other online parameters were as expected. The final cell density and titer were lower than the corresponding laboratory-scale values due to a shorter exponential feed stage; however, Q-beta monomer productivity (g/g DCW) was comparable (see Figure 10, Table IV). Given the program requirements for VLP material, the fermentation titer was deemed to be acceptable. Product quality after purification was tested with several assays, including those for SEC-HPLC, SDS-PAGE, peptide mapping, RNA content and host cell protein content. Sample SEC-HPLC profiles are shown in Figure 6B. Based on all the analytical results, the produced VLP was found to be within the desired specifications.
Table IV: Summary of fermentation data.
An approach combining process and equipment characterization was used to transfer a high titer, fed-batch E. coli fermentation process for the production of Q-beta VLP rapidly and successfully from a collaborator to Pfizer's cGMP pilot plant. The early assembly of an appropriately staffed and sized technology transfer team that enabled efficient communication with the collaborator was key to the success of this endeavor. Based on a review of the fermentation process, high oxygen demand during the fed-batch phase was identified as an important issue, especially if the fermentation was to be reproduced without oxygen supplementation. The original fermentation process was rapidly transferred to a 15-L laboratory-scale fermentation system, while simultaneously collecting process characterization data. Subsequently, equipment characterization of the cGMP pilot plant and laboratory fermentation systems was undertaken. Based on these results, the original fermentation feeding strategy was modified to decrease the duration of the pre-induction feed phase and lower peak oxygen demand. The resulting fermentation process took advantage of the maximum oxygen transfer rate achievable in the pilot-scale fermenter, and successfully produced Q-beta VLP at a sufficiently high titer without the need for oxygen enrichment of the process air stream.
We would like to thank Cytos Biotechnology, Pfizer Vaccines Research Unit, Pfizer Bioprocess R&D Manufacturing and Analytical R&D Group, Michael Dupuis, Aparna Deora, John Amery, David Steinmeyer and Tom Warren.
Shamik Sharma* is a principal scientist, Allison Whalley is a scientist, Joseph McLaughlin is an associate research fellow, Frank Brello is a senior scientist, Bruce Bishop is a an associate research fellow, and Amit Banerjee is a research fellow, all in the department of Biotherapeutics Pharmaceutical Sciences, Worldwide R&D at Pfizer Inc, Chesterfield MO and Andover MA. *To whom corresepondance should be addressed: Shamik.Sharma@pfizer.com.
Article submitted: Nov. 18, 2010.
Article accepted: Apr. 22, 2011.
1. G.T. Jennings and M.F. Bachmann, Biol. Chem. 389 (5), 521–536 (2008).
2. G. Spohn and M.F. Bachmann, Expert Rev. Vaccines 7 (1), 43–54 (2008).
3. M.F. Bachmann and G.T. Jennings, "Virus-like Particles: Combining Innate and Adaptive Immunity for Effective Vaccination." in Novel Vaccination Strategies, S.H.E. Kaufmann, Ed. (Weinheim: Wiley VCH Verlag GmbH & Co. KGaA, 2004) pp. 415–432.
4. C. Ludwig and R. Wagner, Curr. Opin. Biotechnol. 18 (6), 537–545 (2007).
5. R. Golmohammadi et al., Structure 4 (5), 543–554 (1996).
6. M. Emmerling et al., Cytos Biotechnology A.G., Switzerland, "Scalable fermentation process for expression of recombinant bacteriophage capsid protein in a bacterial host," WO patent 2006/125821 A2. 20060524, 2006.
7. S.Y. Lee, Trends Biotechnol. 14, 98–105 (1996).
8. D.J. Korz, et al., Deckwer. J. Biotechnol. 39 (1), 59–65 (1995).
9. M.L. Shuler and F. Kargi, Bioprocess Engineering —Basic Concepts (Prentice Hall P T R, New Jersey, 1st ed., 1992).
10. F.R. Schmidt, Appl. Microbiol. Biotechnol. 68, 425–435 (2005).
11. M. Hoare et al., Biotechnol. Prog. 21 (6), 1577–1592 (2005).
12. B.H. Junker, J. Biosci. Bioeng. 97 (6), 347–364 (2004).
13. F. Garcia-Ochoa and E. Gomez, Biotechnol. Adv. 27 (2), 153–176 (2009).
14. M. Berge et al., BioPharm Int. 21 (4), 48–52 (2008).
15. G.T. Benz, Chem. Eng. Progr. 104 (2), 32–34 (2008).
16. G.J.E. Baart et al., Vaccine 25 (34), 6399–6408 (2007).
17. B.H. Junker et al., Bioprocess Biosyst. Eng. 19 (6), 403–413 (1998).
18. B.H. Junker et al., Bioprocess Biosyst. Eng. 18 (6), 401–412 (1998).
19. B.W. Rainer, Chem Biochem Eng. 4 (4), 185–196 (1990).
20. V. Linek, J. Sinkule, and P. Benes, Biotechnol. Bioeng. 38 (4), 323–30 (1991).
21. C. Marvin and J. Wilson, "Fermentor Design," in Bioprocess Engineering: Systems, Equipment and Facilities, B.K. Lydersen, N. D'Elia, and K.L. Nelson, Eds. (John Wiley & Sons Hoboken NJ, 1st ed., 1996) pp. 3–67.