4 Discussion
In this work, we demonstrated that mixing time can be used as a scaling parameter for HEK293T stable producer cell line culture between 24-DSW plates (operated with a large shaking diameter, ds = 25 mm) and 2 L STRs. Based on the published literature and previous works’ results, for geometrically similar reactors, a constant power per unit volume is generally the more preferred scaling option compared to mixing time. This is due to a potentially prohibitive power input requirement for large scale bioreactors in order to achieve short mixing times (Diaz & Acevedo, 1999; Schmidt, 2005; Yang et al., 2007). Larger fluid volumes result in longer flow paths for bulk circulation, hence higher fluid velocities are required to achieve similar mixing times (A. W. Nienow, 1998; Omar. Al Ramadhani, 2015). As the reactor scale increases, more power input is required if higher agitation rates are used in order to achieve higher velocities (A. W. Nienow, 1998). For scale-translation at small scale, the power input requirement is not limiting as successfully demonstrated by Sani and Baganz, 2016 who used mixing time as a basis to scale between 0.5 L micro-bioreactor (MBR) and 5 L STR for a GS-CHO cell line (Sani & Baganz, 2016).
In this study scale translation was required between small scale vessels having very different geometries and flow patterns (i.e. shaken microwells to bench-scale STRs) and matched mixing time was considered as a possible scaling method. In addition, the use of P/V was limited by the lack of accurate methods to measure the power input in microwell plates. Current methods only estimate the P/V using CFD predictions of the mean energy dissipation as demonstrated by Zhang et al , 2008 (Zhang et al., 2008). Barrett et al ., 2010 used the P/V estimates from Zhang et al ., 2008 as the basis for scaling a hybridoma cell culture process from a shaken 24-SRW system (800 µL) to a 250 mL shake flask culture (Barrett et al., 2010; Zhang et al., 2008). The data was also indicative of results obtained in the 5 L STR albeit at unmatched P/V. The STR could not be matched at constant P/V due to an unfeasible operating stirrer speed in excess of 300 rpm, at which point excessive foaming and vortexes formed in the vessel. The P/V estimates followed a counter-intuitive trend whereby microwell power consumption decreased with increasing shaking speeds. As there are currently no direct methods of measuring P/V in microwells, the accuracy of the CFD predictions are yet to be validated. Hence, it cannot be certain whether scaling in Barret et al ., 2010’s study was actually achieved at constant P/V. On the other hand recent developments in imaging techniques and the availability of high speed cameras and automated image processing methods allow accurate mixing time measurements in very different configurations.
The 24-DSW model for LVV production developed in this work offers great scale-down potential for carrying out initial high-throughput screening and optimisation studies. Compared with other scale-down systems used in early stage bioprocess development, this 24-DSW model offers several benefits in terms of cost, flexibility and ease of use. Micro-bioreactors (MBR) are frequently used in upstream processing for cell culture process development work. Applikon’s micro-Matrix (Applikon Biotechnology B.V, Delft, Netherlands) is a MBR based on the 24-DSW platform where in each well is individually controlled for pH, DO and temperature (Wiegmann, Martinez, & Baganz, 2020). The micro-Matrix’s measurement and control capabilities offer a better insight into the bioprocess compared to the 24-DSW model from this study. The incorporation of pH control also means that similar metabolite profiles to the STRs could be also be achieved with the micro-Matrix. However, the investment and operating costs are significantly higher than the more simplistic 24-DSW model. In terms of ease of use and flexibility, the 24-DSW model is also advantageous compared to MBR platforms, as these require time-consuming set-up with steps such as sterilisation and probe calibration. Furthermore, MBRs are often limited in terms of the number of vessels/wells that can be used for a particular run. In contrast, the metal clamp system (EnzyScreen BV, AL Leiden, Netherlands) holding the 24-DSW system accommodates up to four plates giving a total of 96 wells. Hence, the 24-DSW model offers a higher degree of parallelisation making it more suitable for high-throughput screening work. Other microwell based models have also been developed using the 24-SRW plate as the model system (Barrett et al., 2010; Guy et al., 2013; Silk et al., 2010). In comparison to the 24-SRW plate, this 24-DSW scale-down model offers over 6-fold greater working volume which increases scope for analytics. In addition, it also offers a higher experimental throughput since the larger working volume means several wells do not have to be sacrificed for a single condition.