Evaluation of various bioreactor process systems for the production of induced pluripotent stem cells

Induced pluripotent stem cells (iPSCs) bear potential to change the paradigms of conventional medicine. However, producing sufficient numbers of pluripotent and homogenous iPSC for therapeutic applications is challenging, in particular when using standard adherent cell culture. Suspension systems are prone to automation and thus represent a suitable alternative. Moreover, several well-established suspension culture vessels are already available. To support decision-making when establishing a cell production process, different culture systems were compared regarding the ability to standardize, product quality, economic efficiency, and potential for process optimization. Standard adherent culture was considered as a reference. In all experiments, the previously published AR1034ZIMA human iPSC line (ARiPS) was used. Culture systems comprised simple devices such as Petri dishes, Erlenmeyer flaks, as well as a computationally modelled and optimized stirred bioreactor. Highest standardization and cell yield was achieved when using the stirred bioreactor system, whereas Erlenmeyer flasks provided a good economic efficiency. Cell quality was comparable for all investigated vessels. Results indicate that the culture system should be carefully selected according to the application-specific requirements. Further, we show that low-cost culture vessels facilitate the production of high cell yields exhibiting a reasonable cell quality.


Introduction
Stem cells exhibit a high potential in basic research, e.g. to understand diseases, in drug discovery, as well as in cell-based therapies. Due to their theoretically unlimited self-sustainability and differentiation capacity, stem cells may allow to change the paradigms of conventional medicine by pharmacologically reinforcing aged and diseased tissues. Currently, therapies for the treatment of diseases like type 1 diabetes, Parkinson's disease, spinal cord injury, heart disease, and blindness are under investigation [1]. models by utilizing patient-specific cells [3]. Diseased phenotypes are selectively generated, or cells with genetic defects may be reprogrammed with the aim to develop personalized therapies as well as novel drugs [4]. Although already tested in first clinical studies [5,6], producing feasible numbers of pluripotent and homogenous cell populations for Tissue Engineering and Regeneration remains a challenging task.
For cell-based therapeutic applications, 1×10 7 to 1×10 9 cells are required [7][8][9][10]. Afore mentioned cell numbers are beyond the capacity of adherent cell culture. T-flasks and multi-well plates, which are widely used in iPSC culture, entail extensive labor costs by binding experienced personnel [11]. For strictly adherent cultures, the adhesion surface area is commonly expanded by the use of micro-carriers in spinner flasks. However, recent experiments suggest that the fluid flow regime needs to be carefully characterized in order to avoid redundant physical stress since mechanical stimulation entails lineage commitment, and thus a loss of differentiation capacity [12,13]. Still, suspension cultures represent a considerable alternative culture strategy. Without micro-carriers, iPSCs are propagated as floating aggregates [14][15][16][17]. In contrast to standard adherent cell culture, suspension systems provide cost-efficient cell production [18]. However, transferring a specific iPSC line from standard adherent culture to suspension culture is challenging [19,20]. From an economic point of view, automated highend suspension culture systems allow for reduced labor costs whereas intensive financial investment is required.
Currently, suspension culture strategies of various working volumes, manual operation and complexity are available. Here, Petri dishes placed on an orbital shaker are considered as the least complex culturing system. Under these conditions, Petri dishes are characterized by a chaotic flow regime, limitations regarding monitoring of process parameters, and comparably small working volumes. An increased working volume is feasible by using Erlenmeyer flasks. Like Petri dishes, the chaotic flow regime in an Erlenmeyer flask on a shaker can hardly be standardized, resulting in non-reproducible mechanical stimulation of cultured cells. In contrast to Petri dishes, online monitoring is possible. Commonly, single use sensors are therefore introduced into the culture vessel [21]. To avoid uncertainties due to undefined flow conditions, controlled mixing in stirred bioreactors is harnessed. Stirred bioreactors cover a broad range of working volumes up to 10.000 liters, and represent the gold standard in many biotechnological applications [22]. For the culture of mammalian cells, a variety of automated devices is commercially available. In addition to several online monitoring options, impellers exhibiting optimized geometries ensure defined agitation to support standardization [23]. However, high levels of defined and reproducible culture conditions in combination with process monitoring entail an increased technical complexity. A major advantage of controllable flow conditions in a stirred bioreactor system is the possibility to characterize the flow regime employing computational modelling. Thereby, mechanical stimuli such as shear stress are investigated, facilitating the identification of impeller rotational speeds that maintain cell aggregates afloat while minimizing mechanically-induced stress levels. Furthermore, computational model-derived mechanical culture conditions bear potential to guide cell fate [12].
In our study, we compared different culture systems for their ability to be standardized, product quality, economic efficiency, and potential for process optimization to support decision-making when designing human iPSC (hiPSC)-based production processes ( Figure  1A). Characterized culture devices comprised agitated Petri dishes and Erlenmeyer flasks, and a computational fluid dynamics (CFD)-optimized stirred bioreactor. Adherent cell culture in well plates was used as a reference.

Adherent culture of human induced pluripotent stem cells
For our studies, we explicitly used hiPS cells of the ARiPS line. This cell line was generated from human adult fibroblasts using lentiviral vectors and subsequent TAT-Cre treatment for the removal of transgenes [24]. hiPSCs were maintained on hESC-qualified Matrigel ®coated (BD Biosciences #354277, Germany) 6-well plates (Nunc #140675) in mTeSR TM 1 medium (STEMCELL Technologies #05850, Germany). Cells were seeded at densities of 2×10 4 cells cm -2 for 6 days adherent culture and treated with 10 μM Y27632 ROCK-inhibitor (Sigma-Aldrich #Y0503-1MG, Germany) during the first 24 hours. Cell culture medium was replaced on a daily basis, and cell culture was performed in standard incubators at 37°C and 95% relative humidity and 5% CO 2 .

Suspension culture of human induced pluripotent stem cells
For suspension cultures, ARiPS were seeded at 1×10 5 cells mL -1 in suspension vessels and culturing was done for 6 days. Following, cells were treated with 10 μM Y27632 ROCK-inhibitor (Sigma-Aldrich) for 24 hours. To assess culture systems, hiPSC aggregate formation was compared for: (I) Petri dish culture, (II) Erlenmeyer flask, and (III) an in-house constructed stirred bioreactor. Petri dish and Erlenmeyer flask cultures were performed in standard incubators at 37°C, 95% relative humidity, and 5% CO 2 on orbital shakers (Edmund Bühler GmbH, Germany), whereas hiPSC bioreactor cultures were performed

Rheology
To enhance the precision and thus the prediction power of the CFD model, density and dynamic viscosity measurements of mTeSR TM 1 cell culture medium were performed. Measurements were carried out using a density meter DMA4100M (Anton-Paar GmbH, Germany) and viscosity meter Lovis 2000M (Anton-Paar GmbH, Germany), respectively, following the manufacturer's suggestions.

In silico particle sedimentation modelling and model validation
For modelling cell aggregate suspension, gelatine microspheres (Cultispher®, Sigma Aldrich, Seelze, Germany) of varying diameters (50-300 µm) were used as they showed comparable sedimentation behaviour to iPSC aggregates. In silico particles were generated employing COMSOL's particle tracing module and their intrinsic geometrical attributes were adjusted to match the gelatine microspheres. In silico particles were implemented into the established CFD model and coupled to the stationary solution of the fluid velocity field. The fluid drag force was set to drive particle settling or elevation, respectively. A particle inlet was defined for the bottom boundary of the vessel and time dependent computations for particle elevation were performed for increasing impeller speeds. Particles that remained on the model's bottom boundary after 3 seconds were counted and correlated to the corresponding impeller speed. For model validation, microspheres were homogenously distributed in the bioreactor without mixing until all microspheres had settled at the bottom of the vessel. Impellerdriven suspension of microspheres was visually observed for increasing agitation rates. Throughout validation experiments, the fluid was kept at 37°C to avoid changes of fluid density and viscosity.
in a tailor-made incubator at 37°C and 5% CO 2 . Where applicable, gas exchange was allowed by lid in place, comparable to a well plate. Gas exchange for the bioreactor was provided by two air filters (Sartorius Stedim Biotech GmbH, Germany). An overview of the different culture conditions and reaction volumes, rotational speeds and the initial hiPSC seeding densities is provided in table 1.
For ROCK-inhibitor removal, cell culture medium was replaced 24 hours after beginning the suspension culture. Following, 7/8 of the cell culture medium was replaced every 48 hours.

Stirred bioreactor
A custom-made stirred bioreactor was constructed from a glass vessel (Weckert, Germany), and a three-bladed impeller (Bohlender, Germany). The vessel was equipped with an ion-sensitive field-effect transistor (ISFET) pH sensor (Endress + Hauser GmbH, Germany) in addition to an optical oxygen probe (Presens, Germany). Mixing was achieved by connecting the agitator shaft to an electric drive (Faulhaber GmbH, Germany).

Fluid dynamics and shear stress estimation
To investigate the shear stress caused by agitation, a CFD model was established (COMSOL Multiphysics GmbH, Germany). Therefore, a three-dimensional (3D) to scale model of the fluid body was generated using computer-aided design (CAD) software (SolidWorks, Dassault Systemes, Germany) and imported into COMSOL. The model was composed of a rotating domain enclosing the impeller, and a static domain that portrayed the shape of the submerged pH probe. The two domains were linked via flow continuity boundary condition. An open boundary condition was defined at the air-liquid interface. Additionally, a pressure point constraint was defined to set ambient pressure conditions at the fluid surface. Gravity effects were considered by defining a volume force of according mass acceleration. For the subsequent numerical simulations, COMSOL's rotating machinery module coupled to a κ-ε turbulent flow model based on Reynoldsaveraged Navier-Stokes equations was employed (C ε1 =1.44, C ε2 =1.92, Values derived from density and dynamic viscosity measurements of mTeSR TM 1 medium were used to adjust material properties. All remaining parameters were considered identical to those of water at 37°C.

Flow cytometry analysis of stem cells
Flow cytometry was performed on single cells, treated with Accutase (Sigma-Aldrich #A6S64-100ML, Germany). If not noted separately, all steps for flow cytometry were performed at 4°C, centrifugation between washing steps and before replacing all solutions was done at 250 g for 5 minutes. Antibodies used for immunofluorescence stainings are listed in Table 3.
Briefly, cells were transferred in centrifugation tubes and washed with FACS buffer composed of 1% fetal calve serum (Bio & SELL #ADD.0500, Germany) in PBS supplemented with 2 mM EDTA (Sigma-Aldrich #E5134-1KG, Germany). To stain cell surface antigens, 1×10 5 cells were diluted in 80 μL FACS buffer containing 1 μL FcR blocking reagent (MACS Miltenyi Biotec, Germany). Blocking was maintained for 20 minutes. Antibodies were diluted in FACS buffer and incubated for 30 minutes.
After primary antibody incubation, cells were washed 3 times with FACS buffer (cell surface staining) or Perm/Wash TM buffer (intracellular antigens). Secondary antibodies were diluted 1:100 and incubated for 30 minutes. Quantification of labeled cells was achieved by flow cytometer analysis (FACSCalibur, BD, Germany).

Quantitative real-time polymerase chain reaction
Prior to isolating RNA for quantitative real-time polymerase chain reaction (qRT-PCR) analysis, cells were separated with Accutase (Sigma-Aldrich, Germany). 5×10 5 hiPSCs were washed with PBS and transferred to a 1.5 mL reaction tube occupied with 350 μL RLT buffer + 1% β-Mercaptoethanol (Sigma Aldrich #M3148-25ML, Germany). RNeasy Micro Kit (Quiagen #74004) was used according to the manufacturer's instructions. RNA content was determined by microplate reader Infinite M200 (TECAN). cDNA was generated from 0.5 μg RNA using iScript TM cDNA Synthesis Kit (BIO-RAD #1708891) in a Thermocycler 48 (SensoQuest GmbH, Germany). qRT-PCR was performed with 1 μL cDNA dilution using the SsoFast TM EvaGreen® Supermix (BIO-RAD #1725202). Primers were designed with NCBI primer depot software (https://primerdepot.nci.nih.gov) and used at a concentration of 4 pmol μL -1 . Sequences are listed in the table 4. PCR was performed using a SensQuest (BIO-RAD) thermocycler, where the annealing temperature was set to 60°C, and the elongation temperature to 72°C. Relative expression was quantified using the comparative cycle threshold (Ct) method, normalized to GAPDH and RPL4 expression. Fold difference calculation was done by ΔΔCt method.

Statistics
Quantitative data was analyzed for statistical differences using a one-way ANOVA employing Fisher's least significant difference test. For all statistical tests a p-value <0.05 was considered significant, n=3 for all experiments.

In silico flow regime and shear stress assessment
The estimation of occurring shear stress is a major advantage when using CFD. Since the flow behaviour of any Newtonian fluid, and thus shearing potential, are determined by its corresponding density and dynamic viscosity [25], we investigated the intrinsic properties of mTeSR TM 1 cell culture medium and considered the obtained values in the CFD model. A density of 9.93×10 2 kg m -3 was measured for mTeSR TM 1 ( Figure 1B). This density was equal to the density of water at 37°C [26]. In contrast, 0.765 ± 0.01 mPa s was measured for the dynamic viscosity of cell culture medium, which was significantly higher than the dynamic viscosity of 0.692 mPa s obtained for water ( Figure 1C). The CFD model depicted in Figure 2A visualizes the fluid flow that was caused by the impeller. The streamlines and red arrows indicate the fluid's flow path direction and magnitude, so that we can assume the medium to be principally accelerated in the lower region of the vessel. This is also where the fluid is calculated to reach its peak flow velocity. Once the medium hits the reactor wall, it is then redirected upwards along the flow barrier. Based on the CFD model, shear stress data was extracted when stirring at different rotational speeds. Therefore, in silico domain probes indicating the maximum shear stress within the model's fluid domains were deployed. For the investigation, a range of possible impeller rotational speeds ranging from 20-100 revolutions per minute (RPM) was covered. In consequence, the obtained shear stress levels ranged from 4 mPa up to 12.5 mPa in a non-linear fashion ( Figure 2B).

Particle sedimentation modelling
During suspension culture, it is commonly observed that hiPSCs form aggregates that are prone to sedimentation. By applying drag forces, for instance by mixing, sedimentation is delayed and, when surpassing a specific power input, sedimentation is prevented. In order to identify the critical impeller agitation rate that hampers aggregate sedimentation at the lowest shear stress input possible, we extended the existing CFD model by introducing in silico particles representing cellular aggregates. Therefore, we used COMSOL's particle tracing module to define spheres of sizes and density comparable to that of cellular aggregates generally found in suspension cultures. A species of n=1000 particles with normally distributed diameters (mean: 0.2 mm, standard deviation: ± 0.05 mm) was generated at the bottom boundary of the computational model and particles were elevated by the drag force of the flow field caused by the moving impeller. Time-dependent    studies of 3 seconds each were performed for a range of impeller speeds and the percentage of particles that remained on the model's bottom boundary was correlated to the respective stirring speed ( Figure 2D). Validation experiments revealed a high predictive power of the in silico model (Figure 3).

Process parameter monitoring
Medium conditions were monitored throughout the entire hiPSC suspension culture process for bioreactor experiments. Optical fibre probes were used to measure dissolved oxygen content in a noninvasive fashion. For pH value acquisition, continuous measurements were performed by using a conventional pH dipping probe. An overall decrease of both oxygen ( Figure 4A) and pH ( Figure 4B) was observed during cultivation. Medium exchange at indicated time points (dotted vertical lines) led to the replenishment of nutrients and dissolved oxygen. Aggregate diameter was qualitatively investigated during medium exchange steps.

Suspension culture of hiPSCs
Without any possibility to adhere, hiPSCs are characterized by spontaneous aggregate formation. For the ARiPS cell line, hiPSC aggregates were formed during the first 24 hours in all suspension vessels. After 6 days, the diameter of the hiPSC aggregates from all vessels was similar with a mean of about 260 μm. All aggregates showed a homogenous, spherical morphology with an even contour. In contrast, adherent hiPSCs formed compact colonies with defined borders. The stirred bioreactor's reaction volume of 200 mL resulted in a yield of up to 4.4×10 7 hiPSCs within 6 days ( Figure 5A). For a comparison, 7 6-well plates of adherently cultured hiPSCs were required to obtain the same amount (cell yield of 1×10 5 hiPSCs cm -2 ). A single Petri dish and Erlenmeyer flask culture facilitated a yield of 5.8×10 6 cells and 1.48×10 7 hiPSCs, respectively.   The depicted fluid velocity field (red arrows and streamlines) was caused by stirring at 30 revolutions per minute (RPM) and was obtained using finite element method (FEM) simulation software. The impeller geometry induced a radial flow field, which is guided upwards when the fluid hits the vessel jacket. The velocity magnitude is indicated as colour range. (B) Extracted data on maximum shear stress that was introduced into the fluid at distinct impeller rotational speeds. (C) Gelatine microspheres were used as model particles mimicking hiPSC aggregate sedimentation behaviour and were employed to identify the critical agitation rate at which particle suspension is ensured. (D) Percentage of suspended particles was simulated for an impeller rotational speed range of 10-80 RPM. Coupling of the CFD model with particle physics allowed the calculation of the critical rotation speed at which cell sedimentation is overcome. Continuous particle suspension was ensured at 50 RPM.

Characterization of pluripotent stem cells after suspension culture in different culture vessels
In addition to cell morphology and hiPSC yield, pluripotency was investigated. Therefore, cells from all vessel types were characterized and compared on protein and gene level. In this context, hiPSCs were analyzed by flow cytometry for characteristic pluripotency-associated markers comprising transcription factors SOX2 and NANOG, and surface antigen SSEA4 (Figures 5B and 6). Flow cytometry revealed that at least 95% hiPSCs were found positive for all investigated pluripotency markers. Protein expression in cells cultured in suspension vessels and the adherent control showed no significant difference. Pluripotency of hiPSCs was confirmed on gene level by qRT-PCR analysis. The relative gene expression of transcription factors SOX2, NANOG and POU5F1 (also known as OCT3/4) was compared between adherent control (100%) and suspension cultures ( Figure 5C). The variations in gene expression between hiPSCs cultured in the stirred bioreactor and hiPSC standard adherent cultures were lower than ± 27% and, consequently, had no biological relevance. The expression of NANOG, which is known to be distributed inhomogenously in pluripotent stem cell populations, was reduced by approximately 33% in Petri dish, and by 45% in Erlenmeyer flask cultures. However, the expression of POU5F1 was slightly reduced in both culture vessels. Moreover, stem cells of stirred bioreactor cultures showed lower SOX2 expression compared to the adherent hiPSC control. In contrast, SOX2 expression was increased by 40 % after Petri dish cultivation and by 60% after Erlenmeyer flask culture. Finally, when using the in-house constructed stirred bioreactor, the expression of pluripotency-associated markers of hiPSCs was preserved.

Differentiation capacity of hiPSCs after stirred tank bioreactor culture
In order to confirm the pluripotency of hiPSCs after 6 days of culture in the stirred bioreactor, in vitro differentiation studies were performed and results were compared to adherent control cells. The capacity of both hiPSC groups to differentiate into cells of the endo-, meso-and ectoderm germ layers was demonstrated by immunofluorescence stainings. Here, the expression of characteristic markers Alpha-1-Fetoprotein (AFP; Figures 7A and 7B), smooth muscle actin (SMA; Figures 7C and 7D) and Tubulin β 3 (TUBB3; Figures 7E and 7F) was investigated. Interestingly, a variation in TUBB3 expression was detected. TUBB3 antibodies normally bind to cells with neuronal, elongated outgrowths. This was found true for bioreactor cultured cells ( Figure 7F). Yet, TUBB3-positive differentiated cells that were derived from the adherent control culture ( Figure 7E), were mainly not characterized by neurite-like processes, indicating an immature stage of neural progenitor cells.

Discussion
The scale-up of bioreactors for mammalian cell culture represents a technological challenge and requires reliable and efficient scaling methodologies [19]. In particular, the successful transfer of static adherent monolayer cultures to a 3D suspension environment is vital, since dynamic conditions show a considerable impact on cell behavior [27]. Due to the mechanical sensitivity of mammalian cells and the influence of shear stress, mechanical stimuli on cell fate have  to be considered to ensure consistent product quality. In addition, homogenous oxygen and nutrient distribution must be controlled throughout the entire cell culture process. Thus, the way of introducing agitation and, consequently, the magnitude of induced mechanical stress should be carefully designed and implemented. Nevertheless, reproducible flow and standardized culture conditions entail a high degree of technical complexity, in particular when increasing the working volume for process scale up. In this context, we characterized different culture systems of varying working volumes for their capacity to be standardized ( Figure 1A). Therefore, we cultured a hiPSC line, which exhibited a promising potential for therapeutic applications [24]. Based on the culture process results, the employed culture vessels were assessed regarding their ability to be standardized, yield and product quality, economic efficiency, as well as potential for process optimization.

Standardization
The highest level of standardization was provided by our stirred bioreactor system. In contrast to chaotically-agitated culture vessels such as Petri dish and Erlenmeyer flasks, the flow conditions inside the stirred bioreactor could be described in a reasonable approximation, and facilitated the development of a fluidic in silico model. Although a variety of computational models for stirred bioreactor systems have already been published [28,29], most models lack a realistic parametrization. For instance, rheological fluid properties are often considered to be identical to those of water. A comparison of the dynamic viscosity of cell culture medium, revealed significant variations, resulting in higher shear stress levels. To ensure a sufficient predictive precision, the model was successfully validated, and flow conditions were correlated to distinct impeller speeds. The chaotic flow regimes present in Petri dishes and Erlenmeyer flasks are challenging to validate [30]. In addition to defined flow conditions, the stirred bioreactor allowed for continuous monitoring of oxygen and pH to improve inprocess monitoring. Systems exist that overcome carbon dioxide and oxygen variations during culture handling. Additionally, small scale culture can be performed at constant environmental conditions. Such cultivation devices offer a high level of standardization at considerable system complexity [31].

Yield and product quality
A successful scale-up increases cell yield while retaining consistent product quality [32]. When comparing cell yield, the stirred bioreactor facilitated the production of approximately 4×10 7 human iPSCs within 6 days of culture. The obtained cell numbers are in the lower range of those required for cell-based therapeutic applications [10]. Solely Erlenmeyer flask cultures reached comparable cell numbers. Nevertheless, stirred bioreactor cultures exhibited less variation in cell yield. A limitation of both systems is the high initial cell number that is required to start the culture process. Concerning the adherent culture process, the ratio between initial and final cell numbers was increased fivefold.
In comparison to the reference culture, all tested vessels exhibited consistent cell quality on both gene and protein level. Hereby, analyzed markers comprised the transcription factors NANOG and SOX2 as well as the membrane antigen SSEA4 on protein level. On gene level, the expression of NANOG, SOX2 and POU5F1 (Oct3/4) was confirmed. Despite the advantages of standard adherent hiPSC culture, such as gradient-free feeding of cells (monolayer vs. aggregate), and microscopic monitoring of morphology [33], we suggest a cascade bioreactor approach with increasing working volume. Thereby, low initial cell numbers are possible, and by transferring harvested cells to a greater working volume, large quantities of stem cells can be achieved. Nevertheless, consistent cell quality over several passages needs to be demonstrated.

Economic efficiency
When assuming an average cell number of 1×10 8 cells, which is within range of cell numbers required for therapeutic applications, approximately 18 6-well plates, 17 Petri dishes, 7 Erlenmeyer flasks, and 3 (2.5) stirred bioreactors are needed. This calculation is based on the cell yields obtained after 6 days of culture. While adherent hiPSC culture routinely requires a daily medium change, the medium of the suspension cultures was changed every other day. The lowest media consumption per culture process and vessel was required by Petri dish cultures (29 mL), followed by adherent cultures (72 mL), and Erlenmeyer cultures (181 mL). Bioreactor cultures consumed 725 mL mTeSR TM 1 within 6 days. With regard to the amount of medium that is required to generate 1×10 8 cells, costs become relativized due to the number of required vessels. Stated in numbers, the total media consumption starting with the lowest is ranked as follows: 499 mL (Petri dish), 1202 mL (adherent culture), 1233 mL (Erlenmeyer flask) and 1668 mL (stirred bioreactor). In contrast to material costs, expenses for laboratory space and labor are less for systems with higher working volume. The total time and space required to process 17 Petri dishes is much greater compared to a single bioreactor where medium is exchanged every second day. In particular the standard adherent culture in 18 6-well plates, demanding for daily medium exchange, is disproportional in effort compared to Erlenmeyer flasks and stirred bioreactors.

Potential for process optimization
A major advantage of stirred bioreactors is the use of advanced probing devices such as sensors for dissolved oxygen and pH. This is mainly limited by the dimension-driven possibilities of the selected culture vessel and, in addition, by the method of agitation. Whereas it is comparably tedious to collect online measurement data from Petri dishes and Erlenmeyer flasks while placed on orbital shakers, sensors are easily connected to stirred tanks. The same holds when opting for oxygen and pH control in the medium. The required control appliances are in most cases already incorporated in commercially available stirred tank solutions [34][35][36]. Here, cell expansion under set and stable oxygen content and pH represents the major advantage of aforementioned systems, comprising hypoxic culture conditions if required.
Since stirred bioreactor agitation is prone to computational modelling, it is possible to obtain reasonable data on shear stress which can be easily correlated to the impeller speed ( Figure 2B). However, while all of the above is merely feasible for more complex devices such as stirred tanks with suitable port inlets, oxygen and pH control in well plates, Petri dishes, and flasks is more elaborate and unsuitable for automation. Still, shake flask readers are available for measuring oxygen and pH in fluids while shaking [37]. On the other hand, the resulting increase of system complexity and material cost needs to be considered.

Conclusion
When designing a suspension culture process for iPSCs, the culture vessel should be carefully selected. Our study demonstrates that even simple vessels such as Erlenmeyer flasks allow for the expansion of iPSCs in a range that is required for therapeutic applications with acceptable effort. Hereby, sufficient cell quality comparable to adherent culture is possible. Nevertheless, controlled stirred bioreactor systems facilitate standardization, and thus the integration of a culture process into an automated environment.

Authorship and contributorship
Antje Appelt-Menzel: hIPSC culture, cell characterization, analysis of results, preparation of the manuscript; Ivo Schwedhelm: bioreactor development, computational modelling, bioreactor culture, analysis of results, preparation of the manuscript; Fabian Kühn: bioreactor development, hiPSC culture; Alevtina Cubukova: hiPSC culture, cell characterization; Frank Edenhofer: differentiation studies, preparation of the manuscript; Heike Walles: analysis of the results, preparation of the manuscript; Jan Hansmann: Study design, bioreactor development, analysis of the results, preparation of the manuscript.

Funding information
Our work was funded by the German Federal Ministry of Education and Research; program NanoMatFutur; grant agreement number 13N12971 -ETface.