Design of bioreactors for mesenchymal stem cell tissue engineering



There is a growing need for efficient mesenchymal stem cell (MSC) production processes for a wide spectrum of clinical conditions ranging from the treatment of life-threatening graft-versus-host disease to cartilage repair. Development of cell and tissue engineered products derived from MSCs will be enabled by scalable production processes with standards of safety and efficacy similar to those established for the pharmaceutical industry. Many of the bioreactor design principles established for production of biopharmaceuticals can be applied to production of MSC products; however, specific control of the microenvironment is required for MSC expansion and differentiation. We review how porous scaffolds and bioreactor technologies are applied to the study of the MSC microenvironment. Current methods for MSC production are based on growth in tissue culture flasks, which is labour-intensive and expensive. We suggest that stirred, perfusion or microfluidic bioreactor technologies can more efficiently address the clinical need for large-scale MSC production. Soft lithography and microfluidic design offer extreme geometric precision to study MSC function as well as defining the microenvironment for tissue engineering at the micron scale. Copyright © 2008 Society of Chemical Industry


This mini-review complements the article on ‘Mesenchymal stem cells in tissue engineering’ in this issue and as such focuses on the bioreactor design requirements for expansion of mesenchymal stem cells and growth of differentiated tissues. The role of bioreactors is to transform laboratory-based experimental approaches into scalable cell and tissue production processes with standards of safety and efficacy similar to those established for pharmaceutical therapeutics. Mesenchymal stem cell (MSC) therapies can be divided into two broad areas: (i) repair or regeneration of connective tissues, namely tendon, bone and cartilage, using tissue-engineered constructs; and (ii) MSC transfusion therapy, such as the treatment of steroid-refractory graft-versus-host disease.

Orthopaedic disorders affect a large proportion of the population, with a correspondingly high economic impact: the cost of orthopaedic repair in the USA alone exceeds US $ 28 billion per year.1 In 2005, over 21% of US adults (46.4 million people) were reported as having arthritis2 and nearly 27 million had clinical osteoarthritis.3 This is set to rise as our population ages. MSCs are being used to treat orthopaedic conditions in two ways: implantation of MSCs in a scaffold or carrier system for direct repair of site-specific defects, or by systemic infusion of isolated MSCs to correct degenerative bone diseases such as osteogenesis imperfecta (OI) or osteoporosis. Osteoporosis is a common bone disorder, affecting up to 18% of women and 6% of men in the USA and its intractable nature means that health care costs are high, estimated at US $ 10–15 billion per year.4 Another area of growing interest is the use of MSCs to treat inflammatory conditions. Phase II–III clinical trials using MSC to treat Crohn's disease and steroid-refractory graft-versus-host disease (see have had promising results, which have been attributed to their immunosuppressive properties.5

Bioreactors and fermentation processes have provided the production needs of the food and biopharmaceutical industry over many years. Many of the principles established by these applications can be translated into MSC production. However, there are several important differences between biopharmaceutical and stem cell production processes that are relevant for the design of bioreactors. The most obvious is that the product is a cell, rather than a protein. Accurate characterization of the cell product and development of regulatory guidelines for cell and tissue production are formidable challenges. MSC growth and differentiation are complex and tightly regulated, requiring addition of scarce or expensive components to media (growth factors and serum components). Pharmaceutical-like economy of scale will be difficult to realize, particularly if MSC products need to be individualized and tissue matched to avoid graft rejection.

An innovative and multidisciplinary approach has led to the emergence of MSC tissue engineering. The topics covered by this mini-review are:

  • (i)MSC metabolism and physiological requirements;
  • (ii)mass transport;
  • (iii)bioreactor designs;
  • (iv)clinical process development.


Fetal calf serum (FCS) provides many of the key elements of culture media, including the proteins responsible for transporting essential metabolic substrates such as lipids, iron and fat-soluble vitamins. Growth and differentiation of MSC are regulated by growth factors, extracellular matrix proteins, proteoglycans and hormones. The key to development of fully defined culture media for clinical application will be to define the components within serum that are responsible for MSC expansion, and to understand the microenvironmental factors that regulate differentiation. The influences of soluble and mechanical factors on MSC growth and differentiation to cartilage and bone are summarized in Fig. 1.

Figure 1.

Influence of soluble and mechanical factors on MSC renewal and differentiation into chondrogenic and osteogenic lineages. The components of serum that promote MSC renewal, and those that comprise chondrogenic or osteogenic media, are shown over the arrows. Oxygen concentration, cyclic strain and shear stress can have positive (+) or negative (−) effects on growth and differentiation.

Growth factors and hormones

Extracellular matrix proteins such as fibronectin, laminin and vitronectin are adsorbed from serum onto tissue culture grade polystyrene plasticware and may provide the necessary adhesive requirements for MSC proliferation.6 The lysophospholipid sphingosylphosphorylcholine increases the proliferation of MSC in a dose-dependent fashion, via G-protein signalling and the c-Jun N-terminal kinase pathway.7 In another study FGF-4 was shown to increase the in vitro expansion rate of human MSC.8

Johnstone et al. described the in vitro chondrogenesis of MSCs by pellet culture.9 Inclusion of 10−7 mol L−1 dexamethasone to the culture medium was sufficient to induce chondrogenic differentiation, though efficiency could be greatly enhanced with the addition of transforming growth factor-β1 (TGF-β1) with or without dexamethasone. Shirasawa et al. showed that addition of BMP2 to TGF-β1 and dexamethasone dramatically increased cartilage pellet size and the synthesis of cartilage matrix by MSC isolated from synovium.10 Fibroblast growth factor-2 (FGF-2) and dexamethasone have also been shown to increase the production of hyaluronan in two-dimensional culture of elastic cartilage-derived cells.11 Osteogenic differentiation from MSC was optimized by Jaiswal et al. by adding dexamethasone (100 nmol L−1), L-ascorbic acid-2-phosphate (0.05 mmol L−1) and β-glycerophosphate (10 mmol L−1) to culture media.12

Oxidative microenvironment

Oxygen tension plays an important role in regulation of MSC renewal and differentiation. Fehrer et al. studied the replicative senescence of MSC during long-term culture and showed that low oxygen tension (3% atm) prolonged proliferative lifespan by about 10 population doublings. The in vitro differentiation capacity of MSC was restricted by low oxygen during long-term culture, resulting in reduced adipogenic and osteogenic potential. These effects could be reversed by cultivating cells at 20% pO2.13 Oxygen concentration changes have been shown to induce phenotypic changes in chondrocytes, such as a switch in collagen production from type II to type I, which can then lead to the formation of cartilage tissue with inferior biomechanical properties.14

Oxidative species generated by culture media components15 could also influence intracellular redox potential and differentiation. Changes in intracellular redox potential can modify cellular signalling pathways, transcription factors and gene expression.16 Reyes et al. used autofluorescent spectroscopy to study the change in intracellular redox potential of MSC differentiated using osteogenic medium.17 The ratio of reduced pyridine nucleotides to oxidized flavoproteins was used to estimate intracellular redox potential. Differentiation down the osteogenic path was associated with a reduced ratio (oxidation). These authors also showed that cell density and contact could also reduce the ratio and intracellular redox potential.

Strain and shear stress

Mechanical factors have also been reported to influence chondrocyte metabolism. Cyclical compression or strain has been shown to alter the biosynthetic activity of chondrocytes in tissue-engineered cartilage constructs. Depending on the magnitude of strain, high frequencies are associated with greater biosynthetic activity.18–20 Chondrocytes embedded in agarose have a frequency-dependent response in both proliferation and proteoglycan (GAG) synthesis, with optimal GAG synthesis at 1 Hz for a strain amplitude of 15%.20

MSC expansion and osteogenic differentiation are sensitive to fluid shear stress and strain. Using elastomeric membranes with parallel grooves, Kurpinksy et al. showed that MSCs align parallel to the strain axis (5% strain at 1 Hz over 2–4 days). Mechanotransduction was associated with increased MSC proliferation.21Zhao et al.22 showed that MSC proliferation as determined by colony-forming units–fibroblast (CFU-F) was enhanced in highly porous poly(ethylene terephthalate) matrices at low flow rates (0.1 versus 1.5 mL min−1) over a 20-day period of culture. The higher flow rate (1.5 mL min−1) upregulated osteogenic differentiation potential. The authors concluded that metabolic factors alone could not account for enhanced proliferation, and that shears in the range 10−5–10−4 Pa were responsible for the shift towards osteogenesis. Differentiation of osteocytes has been shown to be enhanced in the range 0.5–1.5 Pa.23

Oxygen and glucose uptake

The specific uptakes of oxygen and glucose for MSCs and chondrocytes reported in the literature are shown in Table 1. Chondrocytes may have lower oxygen requirements and increased glucose consumption, though direct comparison is often not possible due to differences in experimental design, and no studies have directly compared MSC and chondrocytes.

Table 1. Specific uptake of oxygen and glucose (mol s−1 per cell)
MetaboliteCell type
Oxygen2.5 ± 1.3 × 10−171472–4 × 10−17148
 3.3 × 10−17870–1.5 × 10−18149
Glucose1.4 × 10−161503.2 × 10−15151
 9.3 × 10−16152


The role of mass transport is to ensure that cell metabolism is kept within a physiological range by provision of metabolic substrates and removal of toxic degradation products. Understanding how device geometry is related to convective or diffusive transport limitations is therefore a key element of bioreactor design.

The role of diffusion and convection

There is a range of substrates that need to be supplied to the biomass. The rate of diffusion is proportional to their concentration gradient, the constant of proportionality being the diffusion coefficient. The Stokes–Einstein equation relates the radius of the diffusing particle and temperature to the diffusion coefficient:

equation image

where k is Boltzmann's constant, T is temperature and R is the particle radius.24 The volume of a sphere is proportional to the cube of its radius, equation image, so the diffusion coefficient is approximately inversely related to the cube root of molecular weight. Larger substrates including low-density lipoproteins, free fatty acids, transferrin and growth factors will have a lower diffusion coefficient. Their specific cellular uptake (mol s−1 per cell) is many orders of magnitude lower than oxygen, as is the molar concentration required for binding to cell surface receptors (nano- to picomolar range). The diffusion of growth factors is also limiting at low concentration, particularly if high density culture is not supplemented with additional growth factors.

The mass flux is also related to the gradient that can be generated at the cell/medium interface (Fick's law of diffusion). For a stagnant boundary around a cell the mass transfer is by diffusion alone, and is limited by the thickness of the stagnant layer and the concentration at the boundary of the stagnant layer. Mass transfer is increased by reducing the thickness of the stagnant boundary layer surrounding cells. The metabolite with the lowest solubility relative to specific uptake (Table 1) is oxygen: approximately 0.2 mmol L−1 in room air (partial pressure of oxygen is 0.2 atm) at 37 °C. Flask culture systems rely primarily on diffusion, and to a lesser extent natural convection, for transport of oxygen to cells. The depth of media in the flask limits the supply of oxygen from the gas phase.25

Within a polymer scaffold (e.g., polygycolic acid), such as those used to generate cartilage,26 the material properties of the system are spatially and temporally heterogeneous. Galban et al. considered a seeded polymer scaffold as consisting of two phases:27, 28 a void phase (β), which contains the nutrient fluid and some polymer matrix, and the cell phase (γ), which includes the cells, nutrient fluid, extracellular matrix and some polymer matrix. Diffusive transport within the two phase system is modelled by coupling boundary conditions and diffusion equations for both phases (β, γ). The volume-averaging method was utilized to derive a single averaged nutrient continuity equation that allows calculation of effective diffusion coefficients as a function of cell volume fraction and time. Leddy et al. measured the diffusivity of tissue-engineered cartilage as a function of scaffold material, culture conditions and time in culture.29 Diffusivity in these constructs was much greater than in native cartilage. A decrease in diffusivity over time was most likely related to new matrix synthesis and matrix contraction by cells in the fibrin and gelatin scaffolds.

The role of convection is to reduce diffusion limitations imposed by stagnant boundary layers surrounding cells. Hydrodynamic modelling can be used to predict the flow field for various geometries by solution of the Navier–Stokes equations. This is only trivial for simple geometries that generate well-defined flow fields such as parallel plates,30, 31 membranes32 or hollow-fibre bioreactors.33

Improving local perfusion of thick tissue constructs remains a significant challenge for scaffold-based devices. Static culture of cell-seeded 3D scaffolds typically produces thin tissue growth localized to the construct periphery.34 Scaffolds can be perfused by housing them within a flow-through column,35–37 or by suspending them within rotary culture devices or spinner flasks.38 While flow rate can be used to modulate media exchange, pore sizes, connectivity and anisotropy can impart vastly different rates of media exchange and shear stress on cells within the construct.

Calculating flow-mediated shear stress within a porous scaffold is not a trivial exercise. Porter et al. utilized computational fluid dynamics to model the flow of media through scaffolds.34 Micro-computed tomography was used to define scaffold geometry, with generation of a detailed simulation of the flow field within pores. The authors were able to estimate average shear stress within the scaffold and to estimate the fluid shear stresses equation image that correlate with increased osteogenic proliferation. Zhao et al. extended this numerical approach to include terms for the diffusion and consumption of oxygen within the scaffold.22

Selective exchange using membrane systems

Gas-permeable or dialysis membranes offer independent control of respiratory gases or small molecules. The quantity of oxygen that can be delivered to the biomass is limited by its solubility in water. Media perfusion rates can be drastically reduced if oxygen is supplied from a silicone membrane in close proximity to cells.30 A dialysis membrane partitions proteins and growth factors from metabolites of less than 8000 Da. Cells and proteins are dialysed against media without growth factors. Cells are grown in direct contact with the membrane at zero shear, while culture media are exchanged by a perfusion system on the other side of the membrane. Membrane perfusion systems can support cell densities that approach those found in tissues.39


A bioreactor is defined as any device that provides the physiological requirements of the cell (e.g., nutrients, growth factors and mechanical environment) for study of cellular function or scaling production of cells and their products. Figure 2 shows a classification of mammalian cell culture methods that is based on culture device geometry. Figure 3 indicates how convection is used to facilitate mass transfer for various bioreactor geometries. The simplest bioreactors are static culture devices (flasks, bags or dishes) consisting of a single unstirred compartment where nutrients diffuse to cells. Gas exchange (oxygen and carbon dioxide) occurs at the media/gas interface. Mixing will reduce the stagnant boundary layer surrounding cells, scaffolds or micro-carriers, as well as creating a more homogeneous cellular microenvironment. Mixed vessels can be ‘fed’ batchwise, or if cells are immobilized onto scaffolds or micro-carriers, using a continuous perfusion system. A perfusion system attempts to replace the function of the microcirculation so that cells can be grown near tissue density. In the following sections we will review the application of these devices to MSC growth and differentiation.

Figure 2.

Classification of bioreactor devices.

Figure 3.

Bioreactor geometry and mass transfer. The diagrams indicate how convection facilitates mass transfer for various bioreactor geometries: (A) media are mixed by a magnetic impeller; (B) media inside a rotating-wall bioreactor have the same angular velocity, and particles are suspended in ‘free-fall’; (C) bag culture on a rocker generates a wave motion which mixes cells and media; (D) ‘plug’ flow through a column generates a uniform perfusion, though media may not penetrate the interstices of the porous matrix; (E) the parallel plate design generates a parabolic velocity profile with shear stress directly proportional to flow rate. A silicone membrane (upper plate) is used for exchange of oxygen and carbon dioxide; (F) adherent cells growing on the inside of a dialysis fibre with media flowing over the outside of the fibre.

Micro-carriers and scaffolds

Porous micro-carriers were developed for mammalian cell recombinant protein production in stirred vessels. The beads generally range between 100 and 400 µm in diameter, providing a large surface area for attachment of cells. Beads increase attachment surface area and are less likely to foul filtration devices used to separate cells from media and secreted products. Granet et al. investigated the use of Cytodex and Biosion for use in osteoblastic culture.40 Cytodex beads are made of a thin layer of denatured type I collagen, chemically coupled to a matrix of cross-linked dextran. Biosion beads are small, globular, negatively charged plastic particles. Micro-carrier beads have also been used for MSC expansion.41

The rationale for the use of scaffolds in manufacture of bone is to closely mimic the anatomical organization of bone and its tissue matrix. Hydroxyapatite (HA), which in different forms has been approved by the US Food and Drug Administration, is currently used clinically as a bone void filler, and is a logical choice for development of tissue-engineered bone grafts because of its osteoinductive properties. Porous HA is manufactured using a variety of processes including hydrothermal conversion of coral, deproteinization of bovine bone, foaming of HA slurries and porogen leaching. It is also possible to manufacture HA structures with well-defined porous geometry with bubble jet printing or robotic deposition.42, 43 There are numerous studies examining the in vitro growth of MSCs on HA scaffolds.44–55 Pre-clinical proof of principle is provided by the sheep metatarsal model of bone healing, where composite HA-MSC transplants have been shown to have almost the same osteogenic potential as autologous bone grafts in terms of the amount of newly formed bone present at 4 months.56 This approach is gaining wider clinical acceptance as an alternative to autologous bone grafting.

Resorbable polymer scaffolds for bone regeneration include those manufactured from silk,57–61 chitin,46, 55, 62 collagen62, 63 and polyglycolic acids.64–66 Porous three-dimensional silk scaffolds have been made using three fabrication techniques: freeze-drying, salt leaching and gas foaming. Silk fibroin (SF) is a biocompatible, enzymatically degradable material, which can be processed into a bone-like structure, for generation of bone.57, 58 The resorbable polymer scaffolds that have been investigated for the generation of cartilage include polylactic and polylactic-glycolic acid,67–69 alginate,68, 70 silk,71, 72 polycaprolactone73, 74 and devitalized menisci.75

Mixed vessels

Spinner flasks

Spinner flasks are glass or plastic vessels with a central magnetic stirrer shaft and side arms for the addition and removal of cells and medium, and gassing with CO2-enriched air (Fig. 3(A)). Spinners increase the efficiency of scaffold cell seeding and survival in comparison to static culture. Scaffolds are suspended within the spinner flask by threading onto long needles embedded in the side arm stoppers or other support structures. Scaffold and spinner flasks have been used for cultivation of MSCs with osteogenic differentiation.44, 57, 58, 76

Rotating-wall bioreactors

The rotating-wall reactor (Fig. 3(B)) was originally designed to simulate microgravity effects similar to those achieved for cell growth experiments performed by NASA on the space shuttle.77 A horizontally rotating cylinder which is completely filled with culture medium (no gas–liquid interface) rotates liquid inside at the same angular rate as the wall. If the velocity of the rotating fluid is equal and opposite to the sedimentation rate of cells or micro-carriers, the cell suspension will be maintained in a state of ‘free-fall’. Viscous flow around the falling particles reduces the thickness of the boundary layer surrounding cells, resulting in a shorter diffusion distance from the bulk media to the cell. Oxygenation of the media is provided by a silicone membrane wrapped around a central cylinder. Cell-seeded scaffolds can be grown in an environment with low shear stresses.

The rotating-wall reactor has been used to suspend micro-carriers,40 porous scaffolds45, 78 or hollow micro-spheres79 for osteogenic differentiation with superior results when compared to static culture. Cartilage engineering is also particularly feasible using the rotating-wall vessel. Chondrocytes have been generated on beads,80 meshes81 and novel porous biopolymers.82 Most notably, Ohyabu et al. were able to generate large cartilage cylinders (1.25 cm × 0.6 cm, height × diameter) from rabbit marrow cells which formed spontaneously without a scaffold.83

Wave bioreactors

The wave bioreactor system (Fig. 3(C)) provides a gentle wave motion for mixing, provides higher oxygen transfer than in spinner flasks, and has been shown to perform comparably with stirred-tank bioreactors for working volumes between 1 and 100 L.84 Gas-permeable bags are simply placed on a rocker, which induces the wave motion. Cholesterol adsorption onto low-density polyethylene bags sold commercially for this application can be a problem, but can be overcome by pre-treating the bags or replacing them with inert fluorinated ethylene propylene.85 There have been no reported applications to MSC growth and differentiation, but the disposable contained bag system has obvious advantages for clinical applications.

Perfusion bioreactors

A perfusion system distributes media via a network of channels to the biomass and increases mass transfer by continual exchange of media. Various bioreactor configurations for this purpose include scaffolds packed within a column, hollow fibre arrays or ‘printed’ networks of micro-channels (microfluidics). The fluid path must be confined so that perfusion is equally distributed to cells. Mass transfer is further enhanced if the distances between immobilized cells and the perfusion channel are short. Regular geometries such as hollow fibre arrays and microfluidic designs have a more predictable flow distribution compared to porous networks.

Column bioreactors

Hydrodynamic principles that govern the flow pattern inside columns are applied to achieve uniform (plug) flow (Fig. 3(D)) through the column.33 The flow distribution will depend on the hydraulic resistance of the column bed and the design of headers that distribute flow to the column. If the hydraulic permeability of the matrix is non-uniform, flow will be unequally distributed. If the pressure drop along the column is much greater than the dynamic pressure at the inlet header, the flow should mainly depend on the hydraulic permeability of the scaffold, and will be uniform if the scaffold porosity is uniform. Bancroft et al. developed perfusion bioreactors for bone tissue engineering applications.35, 86 The scaffold is held in a cylindrical cassette sandwiched between two ‘O’ rings which seal the chamber. The fluid path is confined so that it passes through the scaffold rather than around it.

Fluid flow increases oxygen transport,87 osteogenesis and mineralization in a dose-dependent manner.35, 88–91 There have been some studies showing that osteogenic development is enhanced in perfusion bioreactors in comparison to spinner flask and rotating-wall bioreactors.38, 92 Non-uniform seeding of the scaffold has been a problem in static and perfusion systems. Investigators have partially overcome this by oscillatory flow92 or low pressure.93

The beneficial effects of perfusion have also been verified for cartilage production,67, 94–98 though it is probably fair to comment that the material produced is ‘cartilage-like’ because it does not have the mechanical strength of explanted hyaline cartilage, limiting application to cosmetic rather than load-bearing applications. One of the problems may be related to poor retention of secreted matrix molecules such as collagen II and various glycosaminoglycans, by the perfusion system. The strength of cartilage is also related to the orientation of collagen fibrils.99 The anisotropic properties of cartilage may be realized by use of oriented nano-fibrous scaffolds.100

Parallel plates

A radial perfusion bioreactor design was developed for the support of human haematopoiesis on a stromal feeder layer.101, 102 The bioreactor consists of two primary compartments: a gas compartment that is separated from the bottom compartment by a gas-permeable, liquid-impermeable membrane, and the liquid-filled bottom compartment with a tissue culture plastic surface for support of anchorage-dependent cells (Fig. 3(E)). Fresh medium enters the liquid compartment at the centre and flows radially outward over cells before exiting into the waste container. The device has been integrated into a Good Manufacturing Practice (GMP) cell production system, which consists of an incubator unit and a processor unit to inoculate and harvest cells. Each bioreactor is a disposable cell cassette which is loaded into processor or incubator platforms.103 The parallel plate perfusion device was used to significantly expand CFU-F and progenitors with osteogenic potential from bone marrow mononuclear cells.31

Hollow-fibre bioreactor

The clinical use of hollow-fibre modules for cell expansion was described almost 20 years ago by Knazek et al.104 Human tumour-infiltrating lymphocytes isolated from patients with metastatic melanoma were expanded by 2–3 logs to produce around 1010 lymphocytes. Since that time hollow-fibre systems have been applied to mammalian cell culture processes including lymphocyte expansion,50, 105 gene transfer to hematopoietic cells,106 production of recombinant proteins and viruses,70, 107–111 hepatocyte culture and extracorporeal hepatic assist devices.74, 112–116 Hollow-fibre bioreactors have been used to study cartilage development. The distribution of collagen, proteoglycan and glycosaminoglycan content within the bioreactor was determined by Fourier-transform infrared imaging spectroscopy.117, 118

A hollow-fibre bioreactor is a two-compartment system consisting of intracapillary and extracapillary spaces. Intracapillary flow is distributed by headers to a hollow-fibre bundle that is potted in resin. Modules can be designed so that flow distributes equally to each hollow fibre in the bundle.33 Flow within the lumen of each fibre has a parabolic velocity profile, so that cells attached to the inside surface are subject to a uniform shear stress which is directly proportional to the intracapillary flow rate. The hollow-fibre bundle is encased in a cylindrical shell with ports for flow of media around hollow fibres. The hollow-fibre membrane is semi-permeable, the pore size determining which molecular species are rejected.

The most widespread biomedical application is renal dialysis, where blood is passed through the intracapillary side of a hollow-fibre membrane dialyser with a low molecular weight cut-off (<10 000 Da). The dialysate is perfused through the extracapillary side of the module. Analogously, cells have been grown on the inside of the fibre (Fig. 3(F)), with perfusion of media on the outside of fibres.39, 112, 115 Alternatively, cells are inoculated into the extracellular space104, 105 with intracapillary perfusion. A macroporous membrane will allow proteins to cross compartments, whereas a dialysis membrane will retain proteins within the cell growth compartment. The hollow-fibre membrane may be modified with ligands for cell separation or attachment of anchorage-dependent cell types.119, 120

Elevated upstream pressure in hollow fibres can drive intracapillary medium out of the upstream section of the fibre, which is then driven back into the fibre downstream. This secondary flow is termed Starling flow121 and tends to drive cells towards the downstream end of the module. Starling flow is not a significant problem for hollow-fibre dialysis modules, which have a much lower hydraulic permeability compared to macroporous membranes.33

Microfluidic bioreactors

Microfluidic devices are fabricated using soft lithographic techniques originally developed by Whitesides and colleagues.122–125 Poly(dimethylsiloxane) (PDMS), which is biocompatible,126 optically transparent, permeable to respiratory gases and elastomeric, is cast onto silicon wafers that have been patterned and profiled by photolithography.127 The silicon mould is manufactured by borrowing well-established lithographic methods developed for the semiconductor industry. Two or more PDMS layers can be sandwiched together to form individually addressable reaction chambers with controlling micro-valves, multiplexers and micro-peristaltic pumps.128–132

Soft lithography is a relatively simple and cheap approach providing great flexibility for manufacturing a wide range of bioreactor device geometries with extreme precision. Diffusion of nutrients is across micron distances, so the mass transfer requirements of cells are easily met. Well-defined geometry means that viscous flow and mass transfer can be estimated by relatively simple calculations. Soft lithography therefore provides a versatile approach in studying the metabolic and physiological requirements of cells.

Pioneering studies by Leclerc et al.126 established the early cell culture applications of soft lithography. A reactor consisting of 10 PDMS layers stacked together with four cell culture chambers, and a chamber dedicated for oxygen supply, established proof of concept for the scalability of this bioreactor manufacturing approach.133 The device was used to support the growth of the hepatic cell line Hep G2. Photolithographic methods were then tested with photosensitive biodegradable polymers which could be used for tissue-engineering applications. Initial biocompatibility data were provided by the attachment and growth of a range of cell lines on microstructures created with a new photosensitive biodegradable polymer.134 PDMS micro-devices with a 3D microstructured channel network were also used to study the influence of shear stress on osteoblast growth and differentiation.135 PDMS devices for culturing cells under conditions of cyclic strain have also been developed. Micro-grooved PDMS sheets have provided a relatively simple method for studying the mechanosensing properties of MSCs.21

Toh et al. micro-fabricated a PDMS device for support of MSCs and hepatocytes. Cell–cell interactions were facilitated by perfusion-seeding cells through a pillar array which concentrated cells within the bioreactor. Cells were then stabilized by a laminar flow complex coacervation reaction of polyelectrolytes (cationic methylated collagen and anionic terpolymer of hydroxylethylmethacrylate–methylmethacrylate–methylacrylic acid).136 Histological staining of bioreactors seeded with MSCs showed that there were calcium deposits after one week of osteogenic induction.

The concept of ‘lab-on-a-chip’ analysis has been applied to cellular analysis by Gomez-Sjoberg et al.129 They have developed a microfluidic chip that creates arbitrary culture media formulations in 96 independent culture chambers and maintains cell viability for weeks. This system was used to automate analysis of MSC proliferation, motility and osteogenic differentiation in response to a range of cell culture regimes. Time-lapse imaging revealed that overall cell motility decreased in chambers where cells were stimulated with osteogenic medium. The results also showed that human MSCs need a minimum of 4 days' stimulation with osteogenic medium to fully commit to differentiation into the osteogenic lineage. The study provides an example of how microfluidic systems can rapidly optimize culture conditions for tissue-engineering applications.


A detailed analysis of the clinical and regulatory requirements for tissue engineering specifying a framework for clinical delivery of tissue-engineered products is beyond the scope of this mini-review, and we will therefore focus on some of the more important aspects of bioreactor design that have a direct impact on clinical delivery. These are the economic scale-up of MSC production, development of serum replacement media for MSC expansion and surgical integration of tissue-engineered grafts.

MSC expansion

The low frequency of MSCs in bone marrow (1:104) makes expansion a prerequisite for MSC therapies.137 For clinical applications where MSCs will be used as a transfusion product (graft-versus-host disease, renal failure, Crohn's disease, myocardial ischaemia, etc.), the optimal dose and frequency of administration can only be determined by clinical trials. The time-consuming and labour-intensive nature of conventional flask culture has restricted upper target doses in clinical trials to about 108 cells per patient.137, 138 Therapeutic efficacy is likely to lie beyond this range.

For flask culture systems, MSCs reach confluence at around 5 × 103 cells cm−2 and maximal expansion relies on seeding at relatively low density (1–50 cells cm−2). Bartmann et al. developed a two-step culture protocol in which 1.5 × 108 cells were expanded from 10 mL bone marrow over a 4-week period.139 The first and second passages required 0.2 m2 and 2.5 m2 of culture surface area, respectively. Plastic adherent cells from bone marrow mononuclear cells were cultured over the first 10–13 days using 8–10 225 cm2 tissue culture flasks. The secondary large-scale culture was achieved using 10 four-layered tissue culture flasks.

While this approach has the advantage of easy translation from laboratory-scale studies that use tissue culture ware, it is not easily scaled. Robotics could be used to reduce the labour component involved in manual inoculation and passaging provided the capital cost associated with these systems is offset by the value of the product. This makes sense for the pharmaceutical industry, which needs to perform large-scale combinatorial drug screening.140

An alternative approach is to select one of the bioreactor technologies discussed in this review. Desirable criteria would be: (i) large surface area to volume ratio to reduce the quantity of incubator space required for expansion; (ii) closed system; (iii) automated inoculation and harvesting; (iv) cheap disposable bioreactor unit; and (v) economical running costs. The quantity (estimated to be around 1 L per 108 cells) and cost of serum consumed by this process are major considerations. It is likely that regulatory authorities will require that FCS is replaced with a well-characterized human product.

Serum replacement

Current techniques for MSC expansion depend on addition of FCS to culture media. FCS is an undesirable source of xenogeneic antigens; it has the potential to transmit animal viruses and prions and has problems with batch-to-batch variability. The most stringent regulators may require that media follow pharmaceutical-like specification requirements, and as such should be composed of highly purified and well-characterized (e.g. recombinant) proteins. Another advantage of a fully defined media specification would be to improve the reproducibility of the expansion process, without the need to test serum batches individually. There is some confusion with regard to the constituents of ‘serum-free’ or ‘serum replacement’ products. These proprietary formulations may be better characterized but still contain serum components, perhaps from animal sources, which may not be highly purified or suitable for human use.141, 142

Autologous serum is a viable alternative to FCS-supplemented media and minimizes the risk of disease transmission. Mizuno et al. developed a closed-bag system to separate serum, and obtained superior human MSC expansion in comparison to FCS-supplemented media.143 A platelet lysate obtained from human thrombocyte concentrates enhanced human MSC output compared to FCS, and was proposed as a clinically applicable animal serum replacement.144 The limited supply and cost of blood products are important scale-up considerations.

Surgical integration of tissue-engineered grafts

Bone grafting is a well-developed surgical technique; osteoinductive biomaterials such as hydroxyapatite have played an important role in bone repair. Development of tissue-engineered bone grafts is likely to play an important role in reconstructive surgery where there is not enough autologous material for a graft. It appears that MSC seeding of a osteoinductive biomaterial may be sufficient for a satisfactory clinical outcome.145

Underlying subchondral bone is required to graft cartilage onto an articular surface. The mosaicplasty technique, which grafts autologous osteochondral material, has limited success but requires a donor site, and may not match the topology of the articular surface. Tissue engineering of osteochondral composites has the potential to overcome these limitations.146


The clinical potential of MSC culture systems will be realized by development of cost-effective and safe processes for large-scale expansion of MSC and derivative tissue types. The development of these therapeutic products has stalled, mainly because flask or pellet cultures employed to study biological processes are not appropriate methods for clinical delivery. The manual passaging of large-volume MSC cultures is labour-intensive and relies on the skill of the lab worker to prevent contamination of cultures. The expansion of MSCs strongly depends on medium supplemented with FCS, a problem which is coming under increasing scrutiny by regulatory authorities, with the risk of transmission of infectious agents via serum. The generation of solid tissues such as cartilage and bone will require technologies for biomimicry of essential tissue elements, microvascular architecture for exchange of metabolites, and regulation of differentiation by cell–cell interactions, extracellular matrix molecules and growth factors.

It will be necessary to move away from the tissue culture flask and pipette for clinical process development. Bioreactor technologies offer the promise of transforming laboratory-based techniques into scalable cell and tissue production processes with standards of safety and efficacy similar to those established for pharmaceutical therapeutics. Mixed-vessel reactors improve oxygen transport and create a more homogeneous environment for cell culture. Perfusion systems replace the vascular system and can support high-density cell growth. Selective exchange of gases and metabolites is possible when semi-permeable membranes are combined with perfusion systems. Soft lithography and microfluidic design offer extreme precision to define tissue geometry, leading to the next generation of bioreactor devices for tissue engineering.