Application of process quality engineering techniques to improve the understanding of the in vitro processing of stem cells for therapeutic use

Abstract

The translation of experimental cell-based therapies to volume produced commercially successful clinical products requires the development of capable, economic, scaleable (and therefore frequently necessarily automated) manufacturing processes. Application of proven quality engineering techniques will be required to interrogate, optimise, and control in vitro cell culture processes to regulatory and clinically acceptable specifications. We have used a Six Sigma inspired quality engineering approach to design and conduct a factorial screening experiment to investigate the expansion process of a population of primary bone marrow-derived human mesenchymal stem cells on a scaleable automated cell culture platform. Key cell culture process inputs (seeding density, serum concentration, media quantity and incubation time) and important cell culture process responses (cell number and the expression of alkaline phosphatase, STRO-1, CD105 and CD71) were identified as experimental variables. The results rank the culture factors and significant culture factor interactions by the magnitude of their effect on each of the process responses. This level of information is not available from conventional single factor cell culture studies but is essential to efficiently identify sources of variation and foci for further process optimisation. Systematic quality engineering approaches such as those described here will be essential for the design of regulated cell therapy manufacturing processes because of their focus on identifying the sources of and the control of variation, an issue that is at the core of current Good Manufacturing Practice.

Introduction

Cell-based therapies are currently being developed for a wide range of disease states. However, the capability to manufacture such therapies to satisfy cGMP standards and emerging regulatory requirements, and therefore to transfer clinically promising therapies to commercially successful products, is threatened by a poor understanding of the in vitro cell culture process variation. This challenge necessitates the bridging of disciplines to apply established manufacturing and quality engineering techniques to in vitro cell culture processes to expedite development of well characterised, capable and economic manufacturing processes for therapeutic cell populations (Archer and Williams, 2005).

Bone marrow-derived human mesenchymal stem cells (hMSCs) are an example of a cell type that will require a cGMP compliant in vitro expansion process to provide sufficient functional biological material for a number of putative clinical applications. This cell type is a promising candidate for some key cell-based therapies: they are simply isolated from a number of sources such as bone marrow and umbilical cord blood and retain significant ability to differentiate into therapeutically useful cell types including cartilage, bone, or muscle (Lee et al., 2004, Pittenger et al., 1999). Human MSCs therefore provide a relevant case study for investigating the in vitro cell culture process.

Conventional approaches to process experimentation using the ‘One Factor At a Time’ (OFAT) approach assume a lack of statistical interaction of variables on the process response (Anselme et al., 2002, Both et al., 2007, Sotiropoulou et al., 2006). However, for many bioprocess systems it is common for variables to be interdependent and produce statistically and practically significant interactions. Cell culture factors are considered highly likely to be interdependent, i.e. if cell seeding density mediates an effect through cell secreted factors, the effective concentration would probably also be dependent on media volume. These effects would confound the analysis of OFAT studies. In contrast, statistical design of experiments (DOE) uses multivariate design to investigate multiple parameters in parallel experimental runs. The approach therefore permits efficient use of resources to provide more information on key process interactions and reproducibility. Combined with automation to remove the variation due to operators from the experimental process DOE should be a powerful tool for in vitro cell culture process investigation. There is a commercial and regulatory driven need for the use of such quality engineering techniques to develop characterised, standardised and validated stable processes for therapeutic cell culture investigation and commercial production.

Statistically designed factorial experiments planned using a rigorous quality engineering approach have been routinely used to improve understanding, and therefore control, of similarly complex processes across a variety of industries including automotive, electronics and pharmaceutical sectors (Montgomery, 1992). A quality engineering approach requires systematic deconstruction and mapping of the process to ensure consideration of all known controllable and uncontrollable influential factors and their likely effects on process responses (output variables). This encourages a logical thought process when selecting process factors for investigation. Screening experiments to identify key process input factors contributing to process variation are an important early component of a quality engineering approach. There are a large number of potentially influential cell culture factors in vitro. However, our initial informing studies and the published literature indicate seeding density, serum, media availability and incubation time are particularly important in the culture of hMSCs (Anselme et al., 2002, Both et al., 2007, Sotiropoulou et al., 2006).

Quality engineering relies on informative responses to assess the effect of process input variables. The ideal response for hMSCs is differentiation of the cells to the lineage(s) of clinical or application interest, thereby proving the utility or multi-potent capacity. However, variation in the relatively lengthy and poorly controlled differentiation process is likely to mask effects from the expansion process and this is therefore an inappropriate response for a screening experiment. Surface markers are an alternative form of response that can act as an analogue. These are quantifiable immediately from the expansion process and are therefore more likely to be sensitive to small variations in the expansion process (although it is recognised that their interpretation is less definitive than that of the differentiation response). STRO-1 and alkaline phosphatase (ALP) are informative surface marker responses that have previously been shown to indicate the proportion of progenitors relative to the proportion of cells committed to the osteogenic lineage respectively in a bone marrow-derived population (Gronthos et al., 1999, Dennis et al., 2002, Stewart et al., 2003, Song et al., 2005). There are also many other characteristic hMSC surface markers such as CD105, CD166, CD90 and CD71 that could be used as indicators of cell character (Fibbe, 2002).

We have previously demonstrated the expansion of an hMSC population using an automated cell culture platform with appropriate manufacturing unit process, scale, and process control capability (Thomas et al., 2007). Here we describe the use of a quality engineering approach to investigate sources of variation in the automated in vitro expansion of an hMSC population. In brief, a factorial screening experiment was used to investigate the effects of different levels of selected key hMSC culture factors and their interactions on the variation of important selected process responses. This is an important early step in process optimisation and a necessary approach to achieve capable manufacturing of cell-based products.