Introduction

Regenerative medicine, of which cell-based therapies will play a significant role, has the potential to revolutionise the healthcare industry and replicate the success of the human therapeutic protein industry (Want et al. 2012). For this to be achieved however, manufacturing platforms must be developed for the reproducible production of clinically-relevant numbers of anchorage-dependent cells. Until now, anchorage-dependent stem cells, including human mesenchymal stem cells (hMSCs), have mainly been cultured as a monolayer on two-dimensional platforms such as tissue culture-treated plastic, including cell factories and roller bottles. Although such methods are commonly used and are considered reliable, they pose significant limitations with respect to scale-up. For clinical indications, which will require billions if not trillions of cells (Rowley et al. 2012), it is clear that these traditional two-dimensional expansion methods are unsuitable. As such, other methods are being investigated (Want et al. 2012; Rowley et al. 2012) and a promising alternative is the usage of microcarrier-based systems in bioreactors.

Microcarrier technology not only provides a significantly larger surface area per unit volume of bioreactor (Nienow 2006) compared to T-flask culture, but also combines the potential ease of scalability, flexible modes of operation, process monitoring and control capability associated with bioreactor cultures that makes bioreactor culture so widespread in the biopharmaceutical industry. Microcarrier culture is commonly used as a large-scale expansion technique for the culture of adherent cells in vaccine production and was first described in 1967 where positively charged DEAE-Sephadex beads were used to culture rabbit embryonic skin cells and human embryonic lung cells (van Wezel 1967). However, in vaccine development, the product of interest is usually an attenuated/killed virus, with little or no attention paid to the cell during downstream processing. For cell-based therapies, the basis of the product of interest is the cell itself, and therefore the focus is not only on cell expansion but also cell quality after harvesting from the bioreactor. A number of challenges in adapting this technology for use in the large-scale expansion of anchorage-dependent stem cells such as hMSCs therefore remain.

Human MSCs are a key human stem cell candidate for large-scale culture due to their multipotency, relative ease of isolation, self-renewal and proliferative capacity, therapeutic efficacy and immunosuppressive properties (Pittenger et al. 1999; Jackson et al. 2007). Human MSCs have been defined by the International Society for Cellular Therapy as being plastic adherent, positive for the expression of CD73, CD90 and CD105 specific surface antigens, negative for the expression of CD14, CD19, CD34, CD45 and HLA-DR specific surface antigens, and possessing the capacity to differentiate towards the chondrogenic, adipogenic and osteogenic lineages in vitro (Dominici et al. 2006). With over 120 clinical trials taking place involving the use of hMSCs (Trounson et al. 2011), there is now a clear need for the development of platform technologies for their large-scale culture which would yield them in sufficient quantity and quality. Studies have taken place investigating the culture of hMSCs on microcarriers with a range of commercially available microcarriers being used including Cytodex-1 (Frauenschuh et al. 2007; Schop et al. 2008), Cytodex-3 (Hewitt et al. 2011), Cultispher-S (Eibes et al. 2010), Cultispher-G (Sun et al. 2010) and SoloHill Plastic P102-L (Dos Santos et al. 2011). However in all the aforementioned studies, the largest volume used was no more than 200 ml and it is unclear as to whether samples of microcarrier suspension greater than 2 ml were harvested. Detachment of cells from the microcarrier surface and subsequent retention of cell quality is equally as important as cell attachment and proliferation given that the product of interest for cell-based therapies is the cells. Therefore it is of paramount importance to consider cell harvesting strategies from the outset so as to ensure a viable, holistic bioprocess.

In this work, hMSCs isolated from bone marrow were cultured in fully controllable 5 l bioreactors with a working volume of 2.5 l for 12 days on plastic microcarriers. These microcarriers were selected based on their suitability for the culture of hMSCs relative to a number of other microcarriers as identified in previous unpublished studies conducted by the authors (data not shown) as well as other groups (Dos Santos et al. 2011), in addition to their amenability for good manufacturing practise (GMP) bioprocesses, given their xeno-free composition. Whilst it would be possible to control the pH and dO2 levels in the bioreactors, it was decided that this investigation would be used to monitor and identify the change in pH and dO2 over the course of the culture as there is little published work that identifies the optimal pH and dO2 levels for microcarrier hMSC culture. Online monitoring of pH and dO2 is a rarity in hMSC culture (monolayer or microcarrier). Thus being able to shift from an unmonitored to an online monitored process with the possibility of control is a significant step. Cells were also cultured in duplicate 100 ml spinner-flasks in parallel to the bioreactors for comparison.

Materials and methods

hMSC monolayer expansion

Human MSCs were isolated from bone-marrow aspirate donated by Lonza (Lonza, Cologne AG), which had been obtained from a healthy donor after the patient provided informed consent. The local Ethical Committee approved the use of the sample for research. Cells from passage 1 were cryopreserved at 2 × 106 cells ml−1 in 2 ml 0.5 % (v/v) human serum albumin, 5 ml dimethylsulphoxide and 93 ml plasmalyte-A (Baxter, Germany). Cells were cultured with Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10 % (v/v) foetal bovine serum (FBS) and 2 mm ultraglutamine (Lonza, UK). Cells were grown in monolayer culture at 37 °C in T-flasks seeded at 5,000 cells cm−2 and were placed in a humidified, CO2-controlled incubator which had air supplemented with 5 % CO2. A complete medium exchange was performed after 72 h culture and cells were passaged at day 6 of culture (the usual time found by us for early-stage passage hMSCs to reach confluency). On passage, the hMSCs were washed with Ca2+- and Mg2+-free phosphate buffered saline (PBS) and then incubated for 4 min with trypsin (0.25 %)/EDTA solution to aid cell detachment from the culture plastic. Trypsin was then inactivated by the addition of fresh growth medium equivalent to 3× the volume of the trypsin solution used for cell detachment. The cell suspension was then centrifuged at 220g for 3 min at room temperature, the supernatant was discarded and the remaining pellet was re-suspended in an appropriate volume of culture medium. Viable cells were counted and an appropriate number of cells were then re-seeded to a fresh T-flask. Two passages were required in order to obtain sufficient cell numbers for seeding of the bioreactors and spinner-flasks.

Bioreactor and spinner-flask culture

Five litre Biostat B Plus bioreactor (diam. T = 160 mm) (Sartorius Stedim, UK), equipped with a 3-blade 45°-pitch wide blade impeller (diam. D = 70 mm) and four vertical baffles and a dished base were used, each fitted with dO2, temperature and pH probes (Hamilton, Germany) at a working volume of 2.5 l. The bioreactor was operated throughout without sparging or headspace flushing with air, though during the exchange of medium, additional dissolved O2 entered the bioreactor. pH and dO2 were measured continuously and the temperature was controlled via a water jacket at 37 °C. 100 ml spinner-flasks (diam. T = 60 mm) (BellCo, USA) with a magnetic, horizontal stirrer bar and a vertical paddle (diam. D = 50 mm) were used in parallel to the bioreactors. One cap on each spinner-flask was left loose to allow for gas exchange. Spinner-flasks were set up on a Bell-Ennium Compact 5 position magnetic stirrer platform (BellCo, USA), maintained in a 37 °C, humidified, CO2-controlled incubator with 5 % CO2. All glass vessels were siliconized using Sigmacote (Sigma, UK) and left overnight prior to autoclaving and rinsing with deionised water. 10,000 cm2 surface area of solid, non-porous Plastic P-102L microcarriers (Solohill Engineering Inc., USA) were used for each bioreactor and prepared in accordance with the manufacturer’s specification. Prior to cell inoculation, microcarriers were conditioned with 1 l growth medium for at least 1 h. Cells were seeded at 6,000 cells cm−2 = approx. 5 cells per microcarrier bead. A 50 % medium exchange was performed at day 3 and thereafter every 2 days.

Following inoculation, the culture was static for 18 h, after which the culture was agitated constantly at NJS (the impeller speed at which the microcarriers were just suspended). This speed was selected as it would ensure that all of the cells attached to the microcarrier surface would be available for mass transfer of nutrients to them and metabolites from them. These transfer rates are insensitive to agitation speed above NJS (Nienow 1997b) but the potential for damage increases rapidly as speed increases (Croughan et al. 1987; Nienow 2006). Thus, NJS as the operating speed is an appropriate choice. NJS was determined visually for the 3-blade 45°-pitch segment impeller and a Rushton turbine impeller with six vertical blades. The impellers had the same NJS but the latter has a higher specific energy dissipation rate due to its much higher power number (Ibrahim and Nienow 1995). Since higher specific energy dissipation rates correlate with a greater potential for cell damage (Croughan et al. 1987; Nienow 2006), the down-pumping impeller was chosen for the bioreactor culture. The choice of this impeller type and its relative size (D/T) fitted in well with an earlier study on microcarrier suspension (Ibrahim and Nienow 2004). The actual NJS was 75 rpm for the bioreactor, whilst in the spinner-flasks it was 30 rpm.

Analytical techniques

A 1 ml sample was aseptically obtained daily from both the bioreactors and spinner-flasks for analysis of pH, glucose lactate and ammonium using a Nova Bioprofile FLEX bioanalyser (USA). Cell counting and viability (via propidium iodide exclusion) was done using a NucleoCounter NC-100 automatic mammalian cell counter (Chemometec, Denmark). Samples for cell counting were determined whilst the cells were still attached to the microcarrier. Based on the viable cell number and metabolite data, the following parameters were determined:

  1. 1.

    Specific growth rate:

    $$ \mu = \frac{{\left( {Ln(cx(t)/cx(0))} \right)}}{\Updelta t} $$

    where μ = specific growth rate (h−1), Cx(t) and Cx(0) = cell numbers at the end and start of exponential growth phase respectively and t = time (h).

  2. 2.

    Doubling time:

    $$ t_{d} = \frac{{Ln_{2} }}{\mu } $$

    where t d  = doubling time (h) and μ = specific growth rate (h−1)

  3. 3.

    Fold increase:

    $$ = \frac{cx(f)}{cx(0)} $$

    where Cx(f) = maximum cell number and Cx(0) = initial cell number

  4. 4.

    Specific metabolite consumption rate:

    $$ q_{met} = \frac{\mu }{cx(0)} \times \frac{{c_{met(t)} - c_{met(0)} }}{{e^{\mu t} - 1}} $$

    where q met  = specific metabolite consumption rate, μ = specific growth rate (h−1), C met(t) and C met(0) = concentration of metabolite at the start and end of exponential growth phase respectively, Cx(0) = cell number at the start of exponential growth phase and t = time (h)

  5. 5.

    Lactate yield from glucose:

    $$ \frac{{Y_{Lac} }}{Glc} = \frac{\Updelta [Lac]}{\Updelta [Glc]} $$

    where Y Lac/Glc  = lactate yield from glucose, ∆[Lac] = lactate production over specific time period and ∆[Glc] = glucose consumption over same time period.

Harvesting

At day 8 of the culture, 60 ml was obtained from each bioreactor and placed into fresh 100 ml spinner-flasks. The microcarriers were allowed to settle after which the medium was aspirated. The microcarriers were then washed with Ca2+- and Mg2+-free PBS and 60 ml trypsin (0.25 %)/EDTA was added and spinner-flasks were placed in a humidified CO2 controlled incubator. Whilst incubating and based on studies of particle abrasion (Nienow and Conti 1978), the vessels were agitated briefly for 7 min at 150 rpm to aid detachment. After incubation, the cells were quenched with 70 ml growth medium and vacuum filtered using a Steriflip 60 μm filtration unit (Millipore, UK). The cell suspension was then centrifuged at 220g for 5 min and resuspended in growth medium. Cell viability was determined using the NucleoCounter NC-100.

Cell characterisation

Immunophenotypic analysis of the hMSCs was determined by flow cytometry before expansion on the microcarriers and on the cells after harvesting. This was performed using a Beckman Coulter Quanta SC flow cytometer (Beckman Coulter, UK) with excitation at 488 nm. Cells were prepared for analysis by centrifuging at 300g. The supernatant was discarded and the cells were resuspended in flow cytometry stain buffer (R&D Systems, UK). A panel of mouse anti-human monoclonal antibodies was used to target cell-surface receptors and was prepared in accordance with the manufacturer’s instructions. The antibodies were based on the panel recommended by the ISCT (Dominici et al. 2006) and included CD73-PE, CD90-PE CD105-PerCP, CD14 FITC, CD19 FITC CD34-PE-Cy7 CD45-PE-Cy5, and HLA-DR-FITC (BD Biosciences, UK; R&D Systems, UK). Cells were incubated with the antibody in the dark at room temperature for 30 min. Associated isotype controls were also prepared for all experimental conditions. A minimum of 10,000 events were recorded for each sample and the data was analysed using FlowJo computer software (Treestar Inc, USA).

The multi-lineage potential of the cells was ascertained by inducing the samples post-expansion with the StemPro Adipogenesis kit, StemPro Chondrogenesis kit and StemPro Osteogenesis kit (Life Technologies, UK). The differentiation media were prepared as per the manufacturer’s instruction. Briefly, this involved thawing the supplement overnight and adding to the basal differentiation medium.

Chondrogenic differentiation involved the formation of a micromass of hMSCs, generated by seeding 5 μl cell suspension at 1.6 × 107 cells ml−1 into a microwell plate. The microwell plate was placed into the incubator for 2 h, after which 1 ml chondrogenic differentiation medium was added; the cells were then returned to the incubator. Cells undergoing adipogenic and osteogenic differentiation were seeded onto separate microwell plates at 5,000 cells cm2 in growth medium and left in a humidified incubator at 37 °C, 5 % CO2 for 3 days. The growth medium was then replaced with either adipogenic or osteogenic differentiation medium and returned to the incubator. The differentiation medium for each differentiation assay was replaced every 72 h until day 21, at which point the hMSCs undergoing osteogenesis were stained with alkaline phosphatase (ALP) and Von Kossa stains and those undergoing chondrogenic differentiation were stained with Alcian Blue.

For chondrogenic differentiation, after fixing the cells with 2 % (v/v) paraformaldehyde for 30 min at room temperature, cells were washed three times with Ca2+- and Mg2+-free PBS and 1 % (w/v) Alcian Blue (chondrogenic) solution was added and incubated at room temperature for 60 min. Prior to observing under the microscope, the samples were washed with PBS three times and 1 ml distilled water was added. Upon final washing, the samples were observed under a light microscope (Nikon, UK).

For osteogenic differentiation, the cells were fixed with 10 % (v/v) cold neutral-buffered formalin for 20 min at room temperature. Samples were then washed with Ca2+- and Mg2+-free PBS and kept in distilled water for 15 min at room temperature before incubating with 4 % (v/v) Naphthol AS-MX phosphate alkaline solution in a darkened room at room temperature for 45 min. Prior to observing under a microscope, samples were washed three times with distilled water. Distilled water was then removed and samples incubated with 2.5 % (v/v) AgNO3 for 30 min at room temperature. The samples were then washed with distilled water three times before observing under the a light microscope.

For plastic-adherence and morphology cells obtained after harvesting were seeded into T-25 flasks to determine their ability to attach to tissue culture plastic as well as to qualitatively identify any morphological changes to the cells post-harvest. The cells were observed using a light microscope.

Results and discussion

hMSC growth

The expansion of hMSCs on microcarriers in a stirred-tank bioreactor at the litre-scale was achieved over 12 days. In parallel to the bioreactor study, hMSCs were also cultured in 100 ml spinner-flasks at the same cell seeding density to surface area ratio. The microcarrier seeding density chosen, 6,000 cells cm−2, equates to ~5 cells/bead. This was chosen based on preliminary work (Hewitt et al. 2011) and for comparability with standard T-flask culture where the seeding density used is typically 5,000 cells cm−2. In all cases, reproducible results were obtained (Figs. 1, 2). In the two bioreactors, the maximum cell densities achieved were 1.68 × 105 and 1.44 × 105 cells/ml by day 9. During this time, the dO2, though not controlled, never fell below 45 % of saturation, well above the level of 20 % where previous work with these cells had shown a deterioration in performance (Rafiq et al. 2013). Operating in this way and knowing the cell density as a function of time enabled an estimate of the specific O2 uptake rate of the cells to be calculated, the average value being ~8.1 fmol/cell h for bioreactor 1 and ~6.3 fmol/cell h for bioreactor 2.

Fig. 1
figure 1

The growth of hMSCs in 5 l bioreactor vessels over 12 days. a Viable cells/ml readings for bioreactor 1 (left) and bioreactor 2 (right), b nutrient and metabolite concentrations for bioreactor 1 (left) and bioreactor 2 (right), c pH and dO2 readings for bioreactor 1 (left) and pH and dO2 readings for bioreactor 2 (right)

Full size image
Fig. 2
figure 2

The growth of hMSCs in 100 ml spinner-flasks over 12 days. a Viable cells/ml readings for spinner-flask 1 (left) and spinner-flask 2 (right), b nutrient and metabolite concentrations for spinner-flask 1 (left) and spinner-flask 2 (right)

Full size image

The cell density achieved was similar to the maximum of 1.1 × 105 and 1.5 × 105 cells/ml achieved in the two spinner-flasks. However, it is notable that growth in the spinner-flasks was slower than that in the bioreactors as the maximum cell densities in the latter were not reached until day 10. This is also reflected in the maximum specific growth rates and doubling times, 0.014 h−1 and ~80 h for the bioreactors and 0.006–0.013 h−1 and ~100–130 h for the spinner-flasks, respectively (Table 1). The doubling times are noticeably longer than those calculated for the same hMSCs in T-flask culture which are typically 45–60 h depending on the medium exchange strategy used, if cultured at 37 °C under conditions of 100 % dO2 (Rafiq et al. 2013; Schop et al. 2009).

Table 1 Fold increase, doubling time and maximum specific growth rate in each of the vessels
Full size table

Based on the maximum cell densities achieved, a greater than sixfold expansion of the hMSCs in the bioreactors was realised, which is comparable to that from standard T-flask culture [ranging from five to tenfold depending on the medium exchange strategy used, if cultured at 37 °C under conditions of 100 % dO2 (Rafiq et al. 2013)]. The number of cells obtained from the 5 l bioreactors is equivalent to 55–65 confluent T-175 tissue culture flasks (where confluence for a T-175 for this hMSC cell line is approx. 6.5 × 106 cells. A slightly lower fold expansion was achieved in both spinner-flasks (3.66 and 5 for spinner-flasks 1 and 2, respectively).

The metabolite profile for both the bioreactor and spinner-flask cultures correlates well with the respective growth curves (Figs. 1b, 2b): with increasing cell numbers result in greater concentrations of lactate and ammonium in the medium with a concomitant decrease in the concentration of glucose. The concentration of glucose decreases steadily during the initial stages in both culture systems and subsequently begins to decrease rapidly until cells reach their maximum cell densities (e.g. days 8 and 9 for the bioreactors) and reaches <1 mM. However, immediately after a medium exchange, there is a measurable increase in glucose concentration and a decrease in lactate and ammonium as identified by the spikes at days 3, 5, 7, 9 and 11 (Figs. 1b, 2b). In both culture systems, the concentration of lactate does not begin to increase until after day 6, where for example with the bioreactors, it rises from ~5.5 to ~10 mM by day 11. Although similar trends are seen for both bioreactors and spinner-flasks, it is notable that in both spinner-flasks, a higher lactate and ammonium concentration was measured at the end of the culture (~12 mM lactate, 1.1 mM ammonium) than for the bioreactors (~10 mM lactate, 0.5 mM ammonium). It is thought that this is not simply to do with differences in cell yields but is probably due to a difference in the culture environments and this will be further explored below. However, in neither culture system do they reach growth inhibitory levels for hMSCs, which have been determined experimentally to be approx. >20 and >2 mM for lactate and ammonium, respectively (Schop et al. 2009). Therefore, whilst the accumulation of these waste products can lead to detrimental effects on hMSC growth, it is more likely that glucose limitation was a major factor here. This is supported by glucose being the primary source for ATP production through either oxidative phosphorylation (resulting in 30–38 mol ATP for each mol glucose) or anaerobic glycolysis (resulting in 2 mol ATP and 2 mol of lactate for each mol glucose). Thus, its presence is fundamental to cell proliferative capacity (Glacken 1988).

Analysis of the specific production and utilisation of lactate compared to glucose is interesting (Fig. 3). In all cases, between days 9 and 12, the amount of lactate produced when compared to glucose consumed rises above 2 mol mol−1, which is the maximum theoretical yield of lactate from glucose (Glacken 1988). This observation suggests that another carbon source, most likely glutamine, is also being consumed which, when metabolised, degrades into lactate and ammonium (amongst other metabolites), hence resulting in higher lactate yields with a concomitant increase in ammonium concentration. The fact that this occurs earlier (day 9) for the bioreactors than for the spinner-flasks (day 10–11) reflects the shorter doubling time, maximum specific growth rates and higher fold increase in the former (Table 1).

Fig. 3
figure 3

Yields of lactate from glucose for each vessel with readings from days 9 to 12. Reference point at 2 mol mol−1 as this is the maximum theoretical yield of lactate from glucose

Full size image

Whilst the lack of glucose is one factor that is likely to impact on cell expansion, it is also important to consider other factors. A significant advantage of culturing cells in a bioreactor system, as opposed to traditional monolayer culture in T-flasks or roller bottles, is the ability of online monitoring and control of various critical culture parameters, such as pH and dO2. Whilst the dO2 and pH were not directly controlled here, these parameters were monitored throughout the course of the bioreactor cultures (Fig. 1c). The measurement of these parameters was not possible in the spinner-flasks used in this work. As expected, there is a gradual decrease in both the pH and dO2 as the culture conditions become more acidic as a result of lactate accumulation, and decrease in dO2 as the cells consume O2 . Previous work has demonstrated that with this cell line, there is a significant decrease in cell yield in monolayer culture conditions when grown under 20 % dO2 in comparison to 100 % dO2 (Rafiq et al. 2013) with an associated increase in both the glucose consumption and lactate production rates. After 12 days in culture, the dO2 for both bioreactors reaches approx. 50 %. The decreasing O2 concentrations in the medium may be a contributory factor in the inhibition of hMSC expansion as was found in the monolayer culture of this hMSC line under lower O2 conditions (Rafiq et al. 2013). Likewise, the specific glucose consumption and lactate production rates (Fig. 4) increase in both bioreactors towards the end of the culture (after day 9) when the O2 level is lower, similar to the trend found previously in monolayer culture (Rafiq et al. 2013). It was thought that in the monoculture case that despite the lower cell number, the lower dO2 levels caused the cells to become quiescent (with respect to expansion), whilst still being metabolically active so as to adapt to the culture conditions. It is possible that this could also be occurring in the bioreactor when the O2 level in the culture medium starts to decrease, resulting in an increase in the glucose consumption and lactate production rate towards the end of the culture, despite there being fewer cells.

Fig. 4
figure 4

Specific growth, glucose consumption and lactate production rates for bioreactor 1 (a) and bioreactor 2 (b) over 12 days

Full size image

As well as the O2 concentration, pH is also a critical culture parameter. The pH in both bioreactors was initially ~7.2 but it steadily fell throughout the course of the culture to become increasingly acidic, reaching pH values of approx. 6.9 and 6.7 for bioreactors 1 and 2, respectively. This drop demonstrates that despite regularly refreshing the medium, which utilises a NaHCO3 buffer system, the culture conditions still become acidic as a result of lactic acid production and accumulation. This change results in a deviation from the physiological pH and therefore could also have an impact on cell growth.

In addition to the possibility that the medium culture conditions resulted in hMSC growth inhibition, another potential limiting factor is that of surface area. It was calculated that by day 9, the cells had achieved theoretical confluency of ~27 cells/microcarrier (based on equivalent monolayer culture confluence) and additional microcarriers were not added at any point during the culture. As a result, assuming confluence was achieved towards the end of the rapid growth phase, it is highly likely that this condition, in combination with the previously aforementioned factors (glucose limitations, decrease in pH and dO2) resulted in the culture entering the stationary/death phase after day 9.

The fact that the growth at the 5 l scale, a larger scale than previously reported in the available literature, is somewhat better than in the spinner-flask is clearly encouraging. It is interesting to consider whether this improvement can be explained by the different physical environment in the two cases. This difference is not only due to the potential for more precise measurement in the bioreactor, which showed that the dO2 never fell close to the level (20 %) at which a poorer performance had been observed in T-flasks (Rafiq et al. 2013); and the fall in pH was not to an unacceptable level. It is also inherent in the different geometries and scales of the two. Firstly, in the spinner-flask, the spinner is essentially a rather crude radial flow impeller whilst in the bioreactor, a down-pumping, axial flow impeller is used and this requires much less energy to suspend particles such as microcarriers (Ibrahim and Nienow 1996, 1999). In addition, the lack of baffles in the spinner-flask compared to the bioreactor, even when the microcarriers are in motion, generally leads to a much less evenly distributed suspension with the particles concentrating in the lowest part of the flask (Nienow 1997a). Thus, it is possible that the poorer particle suspension characteristics of the spinner-flask are adversely affecting cell growth. Another important parameter that changes across the scales, as well as with geometry, is the quality of homogenisation. This parameter generally deteriorates on scale-up but it is also much poorer in the absence of baffles (Nienow 2006). This improved homogeneity due to baffling will become more important when pH is controlled by the addition of base and with increasing scale.

Thus, though at this stage it is not possible to give a precise explanation for the improvement from spinner-flasks to stirred bioreactor, the changes in these different parameters with geometrical configuration and scale are encouraging in relation to the greater scale up that will be required for commercial production of cells for allogeneic therapies (Want et al. 2012).

Cell harvest

Whilst there is a clear need for the optimal attachment and proliferation of hMSCs, equally as important is the ability to successfully harvest hMSCs of sufficient quantity and quality. Using a novel and potentially scalable harvesting strategy based on earlier literature on particle abrasion (Nienow and Conti 1978), cells were harvested from the microcarriers in 60 ml of the bioreactor culture [indeed we have now successfully scaled this up to 500 ml (data not shown)]. By trypsinizing and agitating at a higher agitator speed in a separate vessel, the cells successfully detached from the microcarriers (Fig. 5). Then, using the Steriflip 60 μm filtration unit, the cells were filtered from the microcarriers and subsequently centrifuged. Samples, 60 ml, were taken from each bioreactor after day 8 and, after harvest, the total cell number from the 60 ml matched with the viable cells/ml value obtained via the day 8 sample based on the NucleoCount measurement, thereby validating the daily sampling and counting method.

Fig. 5
figure 5

Images of hMSCs which have been successfully detached from the microcarrier surface after harvesting. (A) Plastic P102-L microcarrier bead, (B) harvested hMSC

Full size image

Cell quality

In order to determine the quality of the hMSCs and to identify whether there had been any change following microcarrier expansion in a bioreactor and the subsequent harvest, the cells were analysed according to the Dominici panel of markers (Dominici et al. 2006) to ascertain their immunophenotypic expression as well as tissue-culture adherence and morphology (both pre-inoculation and post-harvest) and for their multilineage differentiation potential.

For both bioreactors (Table 2), both pre-inoculation and post-harvest, >99 % of the cell population were positive for CD73, CD90 and CD105, whilst for the negative markers CD14, CD19, CD34 and HLA-DR, <2 % of the cell population was positive for these markers pre-inoculation, and this reduced even further to <1 % post-harvest. Crucially, this set of data demonstrates that the immunophenotype of hMSCs are not affected by either the upstream or downstream processing, often a concern that cells will be sensitive to bioprocessing.

Table 2 Immunophenotypic expression of hMSCs before and after bioreactor culture
Full size table

In all cases, hMSCs adhered and grew on tissue culture plastic post-harvest whilst retaining their key morphological features, demonstrating similar morphology both before and after bioreactor culture (Fig. 6). The multilineage differentiation potential of the cells was investigated post-harvest by inducing the hMSCs to differentiate towards the adipogenic, chondrogenic and osteogenic lineages. After 21 days in lineage-specific differentiation media, samples from all experimental conditions were positive for alkaline phosphatase; and von Kossa staining demonstrated their capability for osteogenic differentiation; cells were positive for adipogenic differentiation as shown by the clear formation of lipid vacuoles, and were positive for Alcian Blue staining demonstrating chondrogenic differentiation capability (Fig. 7). Therefore, the microcarrier bioreactor culture and subsequent harvest of hMSCs had no impact on the ability of the cells to retain their differentiation capacity. As such, the characterisation analyses have demonstrated that hMSCs have retained key quality attributes (identity and function—where function is linked to differentiation capacity) after large-scale bioprocessing.

Fig. 6
figure 6

hMSC morphology when cultured on tissue culture plastic before inoculation (left) and after harvesting (right)

Full size image
Fig. 7
figure 7

Differentiation potential of hMSCs after harvesting down the adipogenic (left), osteogenic (centre) and chondrogenic (right) lineages

Full size image

Conclusions

The successful bioprocessing of human mesenchymal stem cells is not trivial. Indeed, expanding human stem cells in culture has been more of an art than a science, so much so that some laboratories do not grow human stem cells at all. Regenerative medicine promises to revolutionise clinical practice and the realisation of this promise requires the development of scalable growth and harvest strategies for the larger scale production of fully functional human stem cells. Here we report for the first time the successful growth of bone-marrow derived hMSCs in fully controllable 5 l bioreactors at a working volume of 2.5 l for 12 days on plastic microcarriers. Moreover, a scalable harvesting strategy was developed to detach a large volume of cells from the cell-microcarrier suspension, thus outlining the potential for the successful harvesting of large volumes of cells from microcarriers. Importantly, the bioprocessing unit operations had little effect, if any, on the quality of the cells for specific regenerative medicine purposes.

It is believed that this is the first instance in which both the successful hMSC cell growth on microcarriers coupled with a detachment protocol from them for subsequent downstream processing at the litre scale has been reported in a peer-reviewed journal. However, it is recognised that the productivity of even the best bioreactor run, equivalent to that produced manually from 65 confluent T-175 flasks was insufficient to meet the demand for the high numbers of cells required for a single cell-based therapy. Nevertheless the results are promising given the opportunity for further optimisation of the culture conditions including feed strategy (e.g. perfusion) and the harvesting step as well as the advantage of scalability and online control with a fully instrumented bioreactor system. Importantly, the improved performance in the 5 l baffled bioreactors compared to the spinner-flasks is particularly encouraging because such configurations are the easiest to scale-up (Nienow 2006), and there are physical studies of reactors and bioreactors in the literature to suggest that further increases in scale with such geometries are likely to be successful.