Preparing for cell culture scale-out: establishing parity of bioreactor- and flask-expanded mesenchymal stromal cell cultures

Cell-based therapies have the potential to become treatment options for many diseases [1], but efficient scale-out of these therapies has proven to be a major hurdle [2, 3]. The production of adherent cells for therapeutic purposes is particularly difficult, due to the need for a large surface area to expand cells to clinically relevant numbers. Traditionally, adherent cell expansion for clinical application involves cell culture in large numbers of flasks in incubators installed in cleanroom facilities. This method requires heavy operator involvement as the culture needs to be seeded, refreshed, passaged and harvested individually and manually. The high costs of cell production in flasks poses a risk to the commercial viability of these cell-based therapies by jeopardizing the positive outcome of a cost–benefit assessment, which is necessary for therapy reimbursement [4]. Consequently, the need for more efficient culture methods that would enable a widespread deployment of cell therapies is well established [5].

Several alternatives that aim to address the specific problems associated with flask culture have come to market. These include stacked and multi-layered flask systems, as well as various bioreactor systems.

The use of bioreactors reduces the amount of labour and allows more space-effective production. Moreover, the use of closed systems enables cell therapy medicinal product (CTMP) manufacturing in environments with less stringent good manufacturing practise (GMP) classification thereby reducing production costs, compared to expansion using flasks. However, changing the culture methods can introduce changes to the characteristics and functionality of the cell products [6], as the properties of the cell product may change upon minor changes in cell manipulation. Therefore, it is crucial to investigate whether proposed new culture methods result in a product that is comparable in terms of identity, safety and potency [7].

Mesenchymal stromal cells (MSCs) are currently investigated for clinical application in numerous trials rendering this adherent cell population highly relevant for cell therapy manufacturing using bioreactors. Here we investigate the possibility to efficiently generate MSCs using a novel, closed bioreactor system (Scinus), designed for automated cell expansion. We compared the characteristics of MSC populations produced with three different culture methods. We expanded MSCs on dissolvable microcarriers in the Scinus and in spinner flasks—which are a down-scaled model for bioreactor culture- and compared this to the traditional flask-based MSC production method. The comparison included MSC expansion potential and key attributes of the resulting cell populations. The expansion potential was assessed in MSC cultures where the MSCs were expanded to therapeutic relevant numbers (typically multiple infusions at a dose range of 1–2 millions of cells per kg of recipient’s body weight, amounting to several hundred millions of cells). MSCs have been attributed many functional properties [8]. They can modulate immune responses [9], secrete various factors contributing to tissue repair [10] and can be differentiated towards relevant cell types for tissue replacement [11]. Since the immunomodulatory properties of MSCs are widely recognised and are frequently included in the proposed mechanisms of action for various therapeutic effects [12], we focus on these immunomodulatory properties and the responsiveness to inflammatory cytokines as indicators for the therapeutic potential of the MSCs.

For one CTMP manufacturing process to be replaced by another, parity between the cell products from both processes needs to be established [7]. We demonstrate that these bioreactor-expanded MSC populations are phenotypically similar to flask- and spinner-expanded MSC populations and are functionally equivalent. Expansion in the bioreactor yields clinically relevant cell numbers.

Bone marrow processing, MSC pre-culture, parallel expansion in flasks, spinners and in the Scinus bioreactor, cell harvest and microscopic inspection of the expanded cells was performed at Scinus Cell Expansion BV. All cellular and molecular analyses of the resulting MSC populations were independently performed at the Leiden University Medical Centre.

In this paper we compared key properties of MSC populations generated with three different culture methods, including a novel closed bioreactor system. The goal was to investigate the potential of the bioreactor for future scale-out of MSC production. For successful scale-out of CTMP manufacturing the following aspects are of importance: (1) (high) cell numbers are reliably obtained, (2) product quality is assured and parity is maintained compared to the original process, and (3) production is efficient. Our comparative study shows evidence that the microcarrier-based bioreactor system meets at least the first two criteria. MSCs were expanded in the bioreactor to therapeutically relevant cell numbers and the resulting MSC populations demonstrated equality to MSC populations generated with a conventional flask-based process with respect to identity, safety and immunomodulatory properties.

A novel clinical expansion method has to reliably achieve clinically relevant cell numbers. In many current clinical MSC applications 1–2 × 10⁶ cells per kilogram recipient body weight are infused. We therefore required that a minimum of 200 × 10⁶ cells was reliably achieved. This minimum was reached in all bioreactor cultures (range 222–510 million). As the cells were still steadily expanding at the day of harvest, cell yield might be increased with prolonged cell culture. Factors that will ultimately limit the maximum yield are the maximum concentration of microcarriers, which imposes the maximum available surface area, and the maximum perfusion speed, which determines the rate of nutrient delivery and waste removal. It should also be noted that the donor material is obtained from relatively old patients (average age 65.4 years) with a diseased joint. While these donors were chosen because they represent an easily accessible donor pool with limited ethical restrictions, they are not necessarily representative of the donor populations used for clinical applications, especially in allogeneic therapies where MSCs are likely obtained from younger and healthy donors.

Continuous growth and high cell numbers were achieved using both microcarrier-based methods, but traditional flask-based culture displayed slightly faster growth. It has been described that expansion of MSCs on microcarriers results in a prolonged lag phase compared to monolayer culture [19, 20], which could have attributed to the observed difference in PDL. Another potential explanation for the difference in observed PDL is the lower seeding density that was used for monolayer culture (400 cells/cm² instead of 1000 cells/cm²) to prevent over confluence at the time of harvest. Seeding density can have a measurable effect on growth kinetics, with lower seeding densities resulting in faster proliferation [21]. Interestingly, the microcarrier culture in the Scinus bioreactor resulted in faster growth than culture in the spinners. This difference is most likely attributed to the continuously controlled culture environment in the bioreactor (maintaining pre-set values for DO, pH and temperature). Also, shear forces in the Scinus system are low compared to impeller-based bioreactors (manuscript submitted) and MSCs are known to be sensitive to shear stress, which can affect population doubling times [22].

A recognised issue in microcarrier-based cell expansion is aggregate formation, which may compromise bead-to-bead transfer which is important in the cell expansion phase. Microcarrier aggregates may also impair live cell harvest efficiency in the downstream processing. In this study we obtained single cell suspensions at the end of culture, indicating that harvest efficiency was not compromised. Minimal aggregation was observed which, combined with a straightforward harvesting procedure, resulted in easy cell collection.

MSCs have been defined by the presence and absence of a specific set of surface markers, their ability to adhere to plastic and a tri-lineage differentiation potential, which is based on MSC expansion in tissue culture flasks [23]. MSCs expanded using the three culture methods were morphologically similar, adhered to plastic and expressed CD73, CD90 and CD105, and maintained their differentiation potential in accordance with these criteria. The cells did not display a method-related qualitative difference in differentiation capacity between culture conditions. We demonstrated robust differentiation along the chondrogenic and adipogenic lineage for all donors and all conditions. Osteogenic differentiation was only clearly observed for Donor 1. Other donors displayed minimal to no osteogenic differentiation, but again no difference was observed between the culture conditions. Although expression of CD73 and CD90 was similar between the culture methods, individual cell preparations harvested from the Scinus bioreactor displayed a reduced frequency of CD105⁺ cells. Reduced expression of CD105 in bioreactor cultures has been reported before [24, 25]. We demonstrated that CD105 is susceptible to degradation by the enzyme used in the harvesting process, explaining the lower CD105 expression levels. Indeed, CD105 expression levels were restored after subsequent expansion and harvesting from flasks (data not shown). Thus far, there is no clinical relevance to the expression of CD73, CD90 and CD105 on MSCs, and the expression has rather been used as a characteristic phenotype of the cells. Therefore, it seems arguable to adhere to this criterion to define MSCs. The MSC criteria previously defined have served a good purpose in the flask-based expansion era, but need to be revisited now that more efficient selection and expansion methods become available.

The cell populations cultured with either method did not show differences with respect to the fraction or composition of cellular impurities. Also the karyotypic analyses did not show a compromised genetic profile for either of the culture methods. Therefore, the data collected in this comparative study, indicate a comparable purity and safety profile for the three production methods.

MSCs cultured with the bioreactor system showed equality to flask-expanded cells with respect to their immunomodulatory properties. IFNγ-induced expression of CD274 (PD-L1) and IDO, molecules that are associated with functional inhibition of immune responses [15, 26], was observed for all cultures. Previously, inter-donor variation in the levels of such upregulation has been reported [27]. This phenomenon was also observed in our study, both in the flask cultures as well as in the microcarrier-based cultures. Our IFNγ-induced immunomodulatory molecule expression data indicate that attributes of MSC populations are more influenced by the starting material than by the culture method and indicate parity between the MSC populations resulting from flask and microcarrier-based culture methods.

For logistical reasons our IFNγ-stimulation assay was preceded by an extra plating and culturing step on a flat surface. Although all induction attributes were therefore measured on flat surface-expanded cells, our results indicate that the preceding expansion on microcarriers does not lead to irreversible changes, whereas donor-dependent differences seem to be preserved.

Inhibition of T cell proliferation is the most commonly tested MSC function relevant for the proposed mechanisms of action of MSCs in various clinical trials aiming at immunoregulation [28]. The inter-donor variation in IFNγ-induced expression levels did not translate into significantly different functionality in an in vitro assay measuring the capacity to inhibit T cell proliferation. Our data, therefore, do not support a direct link between responsiveness of CD274 or IDO expression to IFNγ and the capacity to inhibit T cell proliferation.

Manufacturing efficiency is the third crucial aspect for the viability of cell therapies as a whole. Previous pioneering therapies involving expensive production processes have failed due to unfavorable cost–benefit assessments and lack of reimbursement [29]. Therefore, there is a clear need for alternatives to traditional labor-intensive flask-based cell expansion processes. The use of closed, automated and controlled devices is seen as an important step that facilitates affordable cell expansion within GMP regulations [30]. The use of such systems has several benefits: (1) a closed system minimizes the risk of contamination, providing options to ultimately manufacture CTMPs outside of stringently controlled and expensive cleanroom class B environments, (2) the automation reduces labor costs and process variability [25], (3) the controlled environment within the device provides more accurate information about the state of the cell cultures. This information can be used to improve and further standardize the expansion process. All these factors will lead to more robust production processes at reduced costs. We initiated cultures with pre-cultured cells to obtain sufficient cells to seed all conditions and we chose 20 million cells as a starting inoculum since this amount was previously used as an inoculum for other bioreactor studies [25, 31]. The total cell numbers achieved in our study (average 324 million cells) would require an equivalent of 130 T175 flasks, an amount that requires multiple operators to refresh and harvest in a realistic time-frame. In contrast, the total volume of the bioreactor can be refreshed and harvested by one operator within 10 min and 1 h respectively. An interesting feature of the Scinus bioreactor is the option to culture at any volume between 100 to 1400 mL or to increase the volume of the bioreactor bag during the process. This allows the seeding of very low cell numbers at manufacturing start. Currently, protocols using this feature are being optimised to generate clinically relevant MSC numbers from a single BM biopsy. In these protocols the culture volume is gradually increased and MSC colonise freshly added microcarriers by bead-to-bead transfer [32, 33]. These versatile properties make the medium-sized Scinus bioreactor particularly suited for production scale-out.

We demonstrated that a new microcarrier-based bioreactor system fulfils essential requirements to consider its integration in the further development and scale-up of flask-based MSC manufacturing methods. This system allows MSC expansion to clinically relevant cell numbers, and the resulting cell populations are similar in identity, purity and potency to MSCs expanded in traditional flasks. The system has the potential to efficiently scale-out the production of MSCs for clinical applications.