The authors have declared that no competing interests exist.
Conceived and designed the experiments: MJO MRD MK. Performed the experiments: MJO TM BB. Analyzed the data: MJO TM BB HJG. Contributed reagents/materials/analysis tools: TM HJG KA. Wrote the paper: MJO TM WBL MRD. Provided knowledge and experimental protocol for MSC isolation from placenta: KA.
Large numbers of Mesenchymal stem/stromal cells (MSCs) are required for clinical relevant doses to treat a number of diseases. To economically manufacture these MSCs, an automated bioreactor system will be required. Herein we describe the development of a scalable closed-system, packed bed bioreactor suitable for large-scale MSCs expansion. The packed bed was formed from fused polystyrene pellets that were air plasma treated to endow them with a surface chemistry similar to traditional tissue culture plastic. The packed bed was encased within a gas permeable shell to decouple the medium nutrient supply and gas exchange. This enabled a significant reduction in medium flow rates, thus reducing shear and even facilitating single pass medium exchange. The system was optimised in a small-scale bioreactor format (160 cm2) with murine-derived green fluorescent protein-expressing MSCs, and then scaled-up to a 2800 cm2 format. We demonstrated that placental derived MSCs could be isolated directly within the bioreactor and subsequently expanded. Our results demonstrate that the closed system large-scale packed bed bioreactor is an effective and scalable tool for large-scale isolation and expansion of MSCs.
Mesenchymal stem/stromal cells (MSCs)-based therapies have potential utility in the treatment of inflammatory diseases, the direct regeneration of mesenchymal tissues, or the up-regulation of innate tissue repair processes [
Regardless of the tissue source, MSC populations will require
Bioreactor designs used for MSC expansion include micro-carrier suspensions in spinner flasks, stirred tank reactors, and perfusion reactors, such as fixed beds or hollow fibre bioreactors [
The bioreactor design described here overcomes these problems by incorporating a gas permeable polydimethylsiloxane (PDMS) shell, which decouples the bulk medium perfusion from the supply of oxygen. This allows a reduced perfusion flow rate or even a single pass medium supply. Fused polystyrene pellets are used to construct a scaffold, that is subsequently air plasma treated to generate charged functional groups on the surface, which promotes cell attachment similar to commercial tissue culture polystyrene (TCP) [
This system contained a 1.5 cm diameter by 7.5 cm long scaffold providing a total surface area of 160 cm2, connected to a single pass circuit (
(A) Process flow diagrams of the single pass small-scale 160cm2 bioreactor and (C) recirculating perfusion for the large-scale 2800cm2 bioreactor. (B) Picture of 160 cm2 left and 2800 cm2 right bioreactor scaffolds made from 3 mm polystyrene pellets (scale bar is 10 mm). (D) 2800 cm2 bioreactor inside the incubator. (E) Mass transport model result for radial oxygen diffusion with cell density of 100,000 cells/cm2 in small-scale (160 cm2) and (F) large-scale (2800 cm2) packed bed bioreactors.
A recirculating system was implemented as shown in the process flow diagram (
Scaffolds are constructed from 3 mm diameter by 2.5 mm length fused polystyrene cylindrical pellets (a generous gift from Paul Reynolds of Styron, PA, USA). The pellets were fused together by passing Acetone through a mould with a 1.5cm diameter for the small-scale and a 5 cm diameter for the large-scale. The surface area of the scaffold was calculated by using the mass of the column and the volume estimate was confirmed by measuring the void volume.
Details of the plasma reactor used to treat the polystyrene scaffolds have been reported previously [
To prevent bubble formation due to pressure drop and temperature changes in the system, the system was pressurised with air at 2 PSI (13.79 kPa) using a low pressure two stage gas regulator (Gascon Systems, Thornbury, Vic, Australia) and a pressure safety valve set at 2 PSI (Generant, Butler, NJ, USA).
Surface analysis was carried out using Axis Ultra DLD spectrometer (Kratos, Manchester UK.) using a monochromatic Al Kα X-ray source operating at 225 W, which corresponds to an energy of 1486.6 eV. The area of analysis was 0.3 x 0.7 mm and an internal flood gun was employed to minimize the charging of the samples. The survey spectra were collected at a dwell time of 55 ms with 160 eV pass energy at steps of 0.5 eV with three sweeps. The collected data was then analysed and processed using CasaXPS (ver.2.3.16 Casa Software Ltd.
Placenta was obtained from full term elective Caesarean sections with written patients’ consent and ethics approval from Queensland University of Technology Human Ethics Committee (1000000938). The method used to isolate pMSCs from the placenta has previously been reported in [
Primary green fluorescence protein labelled mouse mesenchymal stromal cells (GFP-mMSCs, generous gift from Mater Medical Research Institute) were isolated from the bone marrow of UBI-GFP/BL6 mice in accordance with The University of Queensland Animal Ethics Committee, previously published and described in [
Non-tissue culture treated polystyrene 60 mm Petri dishes (Sigma-Aldrich) were plasma treated in the same manner as the polystyrene scaffolds above. To assess cell attachment, the air plasma treated Petri dishes and T75 flasks (Nunc) were seeded at 2000 cells/cm2 with GFP-mMSCs in DMEM 10% FBS. The cells were permitted to attach for one and a half hours. The adhered cells were detached and then counted on an FC 500 flow cytometer (Beckman and Coulter, USA) using flow cytometry counting beads (Beckman and Coulter, USA). To compare the growth rate, plasma treated Petri dishes and T75 flasks were seeded with GFP-MSCs at a density of 1000 cells/cm2 and incubated for 3 days in DMEM 10% FBS. The cells were then harvested and counted via flow cytometry as described above.
The 160 cm2 bioreactor was seeded at 1000 cells/cm2 of GFP-mMSCs suspended in DMEM 10% FBS. To distribute the cells evenly, the bioreactor was placed on a tube rocker roller set at 5 RPM for 10 minutes, followed by 5 minutes of rest in an incubator and repeated for 3 hr. As a 2D control, GFP-mMSCs were seeded at a density of 1000 cells/cm2 in T175 flasks (Nunc). The GFP-mMSCs were grown either under static conditions or perfusion (5 mL/day) in the bioreactor. The cell number was quantified every two days by replacing the medium with a 1:50 dilution of AlamarBlue in DMEM 10% FBS medium and incubated for three hours. The reacted AlamarBlue medium was replaced with fresh DMEM 10% FBS and the culture was continued. The fluorescence signal of the AlamarBlue in medium was measured in a plate reader (FLUOstar Omega, BMG Labtech, Germany) along with a cell titration to infer cell numbers using excitation and emission filters of 544 and 590.
The bioreactor was washed with one reactor volume of PBS and replaced with one reactor volume of 0.01% Trypsin EDTA (Gibco). The Bioreactor was then incubated (5% CO2 and 37°C) for 15 minutes. The detached cells were removed by first draining the Trypsin solution and then pumping through three reactor volumes of DMEM 10% FBS. Live cells numbers were determined by an automated cell counter (Bio-Rad TC20, CA, USA) using Trypan Blue. This number was compared to the Alamar Blue result to estimate the percentage of live cells removed from the bioreactor.
For direct imaging of the GFP-mMSCs attached to the 160 cm2 column at the end of culture, the column was removed and fluorescence microscopy (Nikon Eclipse Ti-u with a Nikon Digital Sight Ds-Qimc camera, Japan) was used to image GFP expressing cells.
To image the whole bioreactor using the IVIS Imaging System 200 series (Caliper, PerkinElmer, MA, USA), the cells were fixed with 4% paraformaldehyde and were stained with 10 μg/mL propidium iodide (PI, Invitrogen) for 10 minutes. The bioreactor was then washed with PBS and stored on ice prior to imaging. As negative controls, bioreactors containing no cells underwent the same fixing and staining process. The bioreactors were imaged with an excitation filter of 520 nm and an emission filter of 620 nm. The thresholds were adjusted to remove background auto florescence.
The 160 cm2 scaffold bioreactor was seeded through the injection port with 1000 cells/cm2 of passage four pMSCs suspended in DMEM 10% FBS, using the same seeding process, culture conditions and 2D controls as the expansion of the GFP-mMSCs. However, in the perfusion experiment the cell numbers were quantified by AlamarBlue at the end of the culture. The cells were then harvested and characterised by mesodermal tri-lineage differentiation capacity.
The placental tissue was digested by the same method outlined in the pMSC isolation and cell culture section above. Instead of a density gradient the filtered cell suspension from the 10 g of starting tissue was seeded into the bioreactor or into a T175 flask, using the same seeding method outlined in the GFP-mMSC expansion. After 24 hours, the cell suspension was removed, the medium was replaced and the flow rate was adjusted in the bioreactor to 5 mL/day. Once the pMSC colonies were observed in the T175 flasks, the cells were harvested in both the bioreactors and flasks and reseeded into the same bioreactor or flask to continue the expansion process. The cell number was monitored using the AlamarBlue. The cells where harvested once reaching confluence and characterised by mesodermal tri-lineage differentiation capacity.
The 2800 cm2 bioreactor was seeded at 1000 cells/cm2 with GFP-mMSCs suspended in DMEM 10% FBS using the injection port. The T175 flasks function as a 2D control. The same seeding procedure as per the small-scale expansion experiments was followed. The reservoir was filled with 250 mL of DMEM 10% FBS to give a total medium volume of 360 mL in the circuit (bioreactor volume 110 mL). The medium was pumped in the recirculating circuit at 0.5 mL/min. A 1 mL medium sample was taken from the injection port each day. In addition, at the end of the culture period an additional 1mL sample was taken from the bulk medium in the reservoir. Glucose and lactate concentrations were quantified by the Mater Hospital Pathology laboratory. At the end of expansion, the cell numbers were quantified by the AlamarBlue method.
Osteogenic differentiation was achieved by culturing pMSCs in induction medium containing high glucose DMEM (HG DMEM, Invitrogen) supplemented with 10% FBS, 10 mM β-glycerol phosphate (Sigma-Aldrich), 100 nM Dexamethasone (Sigma-Aldrich) and 50 μM L-ascorbic acid 2-phosphate (Sigma-Aldrich). The calcium deposits were visualised by staining with AlizarinRed S (Sigma-Aldrich). Alkaline phosphatase (ALP) activity was quantified with p-nitrophenyl Phosphate (Sigma-Aldrich) following manufacturers’ instructions, measuring the absorbance at 405 nm after a 30 minute incubation.
Chondrogenesis was induced using cultures of 200,000 cell pellets grown in HG DMEM supplemented with 110 μg/mL sodium pyruvate (Invitrogen), 10 ng/mL recombinant human Transforming Growth Factor β1 (TGF-β1, Peprotech), 100 nM dexamethasone, 200 μM ascorbic acid 2-phosphate, 40 μg/mL L-proline (Sigma-Aldrich) and 1% ITS-X (Invitrogen). 75% of the chondrogenic medium was changed every two days, the collected medium was stored at -20°C for subsequent glycosaminglycan (GAG) analysis. GAG was visualized by staining frozen sections of the cell pellets with Alcian Blue (Sigma-Aldrich). GAG was then quantified by first digesting the pellets in papain and then staining with 1,9-Dimethyl-methylene blue zinc chloride double salt (Sigma-Alrich).
Adipogenic differentiation was induced over 2 weeks with HG DMEM supplemented with 10% FBS, 10 μM insulin (Invitrogen), 1 μM dexamethasone, 200 μM indomethacin (Sigma-Aldrich) and 500 μM 3-isobutyl-1-methyl-xanthine (Sigma-Aldrich). Lipid droplets were visualised by staining with Oil Red O (Sigma-Alrich) and were quantified with an adipogenensis detection kit (Abcam) following the manufacturer’s protocol.
The DNA was quantified by PicoGreen dsDNA Reagent Kit (Invitrogen) following the manufacturer’s protocol.
A mass balance calculation crudely estimates the concentration of oxygen at any point along the radius by a simple source sink method. A more sophisticated theoretical treatment was applied, but will be published elsewhere. It was assumed that there was no axial flow and that oxygen only diffused radially; no axial diffusion was considered.
However, as the bioreactor vessel space is filled with non-gas permeable particles, a correction factor (
Correction factor for oxygen diffusion coefficient in medium Dm (3.290x10-05 cm2/s [
Tortuosity is a function of the voidage and sphericity (
Sphericity is calculated based on the number of particles (n), the voidage (ε) and the surface area to volume ratio (a). Equation taken from [
To calculate the shear stress from the medium perfusion, the pressure drop (ΔP) was first calculated by the Ergun equation:
The viscosity (μ) of the medium is 0.008 poise [
Results are expressed as means and standard deviations of four biological replicates. Differences were determined by T-test using SPSS statistics (ver 17, SPSS Inc, Chicago) and values of p≤0.01 were considered significant.
The bioreactor mass transport model was used to estimate the steady-state oxygen concentration along the radius at a cell concentration five times the maximum confluence of MSCs (100,000 cells/cm2). The mass balance predicted no significant oxygen gradient in the small-scale 160 cm2 bioreactor (
The Ergun equation was used to calculate the shear due to fluid flow in the bioreactor. The hydrodynamic shear from the medium perfusion in the small bioreactor (5 mL/day) and large bioreactor (0.5 mL/min) was calculated to be 9.5x10-5 Pa and 1.25x10-3 Pa, respectively.
The rotary air plasma treatment greatly increased the charged oxygen groups on the surface on the polystyrene (
(A) Surface composition of the air plasma treated polystyrene scaffold determined by XPS. (B) GFP-mMSC attachment after one and half hours, initially seeded at 3000 cells/cm2 (n = 4). (C) The growth after 3 days of culture of GFP-mMSC seeded at 1000 cells/cm2 (n = 6). (D) Growth of GFP-mMSC in the bioreactor (BR) under static conditions (n = 4) and (E) under 5 mL/day perfusion (n = 4). (F) Cell harvest recovery and viablity from the bioreactor. (G, H, I and J) Fluorescent microscopy showed that the GFP-mMSC attached to the scaffold (scale bar is 500 μm). (K) IVIS imaging of the fluorescent intensity of PI stained GFP-mMSC in the bioreactor under static and (L) 5 mL/day perfusion conditions. IVIS images are a red (low) / yellow (high) heat map of fluorescent intensity.
The attachment and growth of GFP-mMSC were compared on our plasma treated surface in 2D and commercial TCP. Only half of the 3000 cells/cm2 seeded attached to the plasma treated surface after one and half hours (
The GFP-mMSCs growth rate in the bioreactor was significantly less than the 2D controls under both static and perfusion conditions (
The cells formed a monolayer on the fused pellets and maintained a spindle-like morphology. In addition, the cells were observed to grow across the connection between the beads (
Passage 4 human pMSCs were expanded under static conditions and perfusion conditions. The doubling time of bioreactor expanded cells took longer than the 2D controls (T175 flask), 63.5±1.5 hr and 43.7±1.5 hr, respectively for static (
(A) pMSC expansion in the small-scale 160 cm2 bioreactor (BR) in static (n = 4) and (B) 5 mL/day perfusion (n = 4), with the 2D controls (2D). (C) IVIS imaging of PI stained pMSC under perfusion conditions. IVIS images are a red (low) / yellow (high) heat map of fluorescent intensity. (D) Two week tri-lineage mesodermal differentiation induction of bioreactor expanded pMSC and 2D controls down the adipogenic (Oil Red O, 10x, scale bar is 100 μm), osteogenic (Alizarin Red, 5x, scale bar 500 μm) and chondrogenic (Alcian Blue, 10x, scale bar 500 μm) lineages. Quantification of (E) triglycerides (n = 4), (F) ALP activity (n = 4) and (G) GAG production.
The cells were characterised by mesodermal tri-lineage differentiation There was no difference between the bioreactor expanded cells in both static and perfusion conditions and the corresponding 2D controls when stained with Oil Red O, AlizarinRed S or Alcian Blue (
pMSCs were isolated from 10 g of digested placenta directly into the bioreactor was expanded under 5 mL/day perfusion. The growth rate of pMSCs was slower in the bioreactor compared to the 2D controls (
(A) cell growth of pMSCs isolated directly from a placenta in the 160 cm2 packed bed bioreactor undergoing 5 mL/day perfusion (n = 4). On day 13, the cells were redistributed back into the same flask or bioreactor. # indicating no 2D control result was measured as the flask was harvested on day 18. (B) Two week mesodermal tri-lineage differentiation induction of bioreactor expanded pMSC and 2D controls down the adipogenic (Oil Red O, 10x, scale bar is 100 μm), osteogenic (Alizarin Red, 5x, scale bar 500 μm) and chondrogenic (Alcian Blue, 10x, scale bar 500 μm) lineages. Quantification of triglycerides (n = 4) (C) and ALP activity (n = 4) (D) GAG (E) production compared to its matching 2D control (n = 4).
The qualitative differential potential of the osteogenic, chondrogenic and adipogenic lineages were similar as shown by staining between the bioreactor expanded pMSCs and the 2D controls (
GFP-mMSC was used to demonstrate the scale up potential of a larger 250 mL vessel that provided 2800 cm2 of growth surface. The cell growth was significantly slower in the bioreactor than in the 2D controls, with doubling times of 21.5 ± 0.1 hr and 20.8 ± 0.4 hr, respectively (
IVIS images are a red (low) / yellow (high) heat map of fluorescent intensity.
The IVIS imaging showed heterogeneous fluorescence intensity, implying that the seeding method is not as robust when scaled-up to the large-scale bioreactor compared to the small-scale bioreactor (
The packed or fixed bed bioreactors, like many other bioreactor designs, rely either on the perfusion or mixing of the medium to provide the two limiting metabolites, glucose and oxygen. Due to the low oxygen solubility, the perfusion or mixing rates must be high to meet the oxygen demand from the expanding cells and to prevent depletion and significant gradients [
The overall growth rate of pMSCs expanded in the bioreactor was less than that observed for traditional flask expanded cells (Figs
This impaired bioreactor growth rate relative to traditional tissue culture flasks could be attributed to surface chemistry and geometry of the scaffold or to the method used to seed the bioreactor. However, our air plasma treated polystyrene surface provided a comparable surface to commercial grade tissue culture plastic (
Despite the significant decrease in growth rate in our bioreactor, the pMSCs maintained the capacity to differentiate down the osteogenic, adipogenic and chodrogenic lineages. No differences in mesodermal tri-lineage differentiation were observed under the perfusion expansion protocol (
To further characterise the bioreactor’s effect on MSCs stemness, and to demonstrate the flexibility of the design, pMSCs were isolated from a digested placenta directly into the bioreactor. The pMSCs produced from this protocol maintained their potential to differentiate down the adipogenic, osteogenic and chondrogenic lineages (
Although the growth rate of pMSCs in the bioreactor was less than the cells grown on tissue culture plastic, the few extra days required for MSCs to reach confluence in the packed bed bioreactor represents an acceptable trade-off to have the potential for an automatable method to digest, isolate and expand pMSCs in a single closed bioreactor system. An isolation/expansion process based on our design, although performed manually here, could be easily automated [
The packed bed bioreactor designed to decouple the oxygen transport from the bulk nutrient supply from the medium flow operates efficiently at a lower perfusion rate, resulting in lower shear stress acting on the cells. We observed a 10 fold expansion of pMSCs in the bioreactor after a one week culture, while still maintaining their differential potential. In addition, pMSC isolated and expanded directly onto the bioreactor still maintained their mesodermal tri-lineage differentiation potential. This design is scalable and potentially automatable. Although the growth rate of pMSCs in the bioreactor was less than the cells grown on tissue culture plastic, the few extra days required for MSCs to reach confluence in the packed bed bioreactor represents an acceptable trade-off to have the potential for an automatable method to digest, isolate and expand pMSCs in a single closed bioreactor system.
We wish to thank Josie Tarren and the Mater Hospital Pathology Unit for their assistance in performing the glucose and lactic acid quantification. We like to also thank Paul Reynolds of Styron, PA, USA for donating the polystyrene pellets. This research was supported by Wound Management Innovation CRC and Inner Wheel Australia.