Key words

bioreactors; ceramics; polymers; perfusion; culture

7.1 Introduction

In general, bioreactors are used to perform biochemical or biological reactions under controlled conditions of temperature, pressure and fluid composition. Large-scale applications such as fermentation and waste-water treatment (Visvanathan et al., 2000) are widely appreciated but bioreactors are also being used in a variety of pharmaceuticals manufacturing processes including those that result in antibiotics, growth factors and vaccines (Shevitz et al., 1990). The more recent uses of bioreactors in tissue engineering include in vitro expansion of cell products and implantable tissue (Barron et al., 2003) and drug testing (Wua et al., 2008). The broad definition of regenerative medicine leads to strong overlaps in the application of bioreactor technologies with other sub-disciplines such as cell therapy (Mason et al., 2011) and tissue engineering (Griffith and Naughton, 2002). Perhaps regenerative medicine is somewhat different in the added emphasis on restoring function of tissue or organs with stem cells (Gurtner et al., 2007) that can grow cell populations in ways that go far beyond the expansion by simple division of differentiated cells.

The ability of stem cells to differentiate into different cell types is constrained to different degrees depending on the nature and source of the stem cells. At the very least, stem cells are able to differentiate into one specific cell type (unipotent) while also possessing the ability to renew and preserve their undifferentiated population. Obviously, stem cells that are able to differentiate into a wider range of cell types are potentially most useful. These can be multipotent stem cells such as hematopoietic stem cells (HSCs) which differentiate into a range of related blood cell types as shown in Fig. 7.1. Most multipotent stem cells in adult or fetal tissue are present in relatively small numbers, at specific locations called stem cell niches. These locations have been identified in many organs including: the epidermis, hair follicle, intestine, brain and bone marrow and are reviewed elsewhere (Fuchs and Segre, 2000; Moore and Lemischka, 2006). The stem cell niche provides a local microenvironment capable of maintaining the stem cell population and protecting cells from differentiation and apoptic stimuli that would otherwise deplete their numbers. When the natural processes of tissue repair are required the stem cells in the niche are activated to produce progenitors that are committed to a particular differentiation lineage.

Other multipotent stem cells such as human mesenchymal stem cells (MSCs) do not appear to have a specific niche, although first discovered in the bone marrow; they have since been isolated in other adult and fetal tissue and may be present in all post-natal organs and tissue (Chamberlain et al., 2007). Human MSCs cells can differentiate into a range of tissues including bone, fat and cartilage. Pluripotent stem cells might be even more attractive as a cell product because of their ability to differentiate into most cell types in the body if the differentiation process can be controlled. The embryo is a commonly considered source of pluripotent stem cells because they can be propagated indefinitely in the embryonic state (Thomson et al., 1998). The problem with embryonic stem cells has always been the ability of in vitro culture to control the differentiation to specific cell lineages amongst all the possibilities. Nevertheless, some success has been demonstrated for certain cell types for well over a decade (Brustle et al., 1999). The controversy surrounding embryonic stem cell sources could be addressed by culturing to induce pluripotency in cells from an adult source. This has now been achieved for several types of cell with reduced risk of generating cancer cells (Yamanaka, 2009).

Finally stem cells can be used to produce cell types that are not normally able to divide, such as red blood cells and plasma. All blood cells, including the 200 billion red blood cells produced daily by an adult human, are derived from HSCs predominantly located in the bone marrow. Therefore the in vitro culturing of HSCs will be a potential avenue to useful blood cell products.

Whether considering classic tissue engineering techniques in which cells are loaded into scaffolds and designed to grow by cell division or considering regenerative medicine therapies that require the presence of stem cells, it is necessary to remove cells from the source, sort them and then culture the cells in vitro to a population size suitable for the cell therapy or tissue replacement. Traditionally, the culturing of cells has been accomplished in two dimensions using dishes under static media culturing conditions. The culture media is then replaced periodically, thereby exposing the cells to large discontinuities in media concentration over time which can affect the culture results. However, culturing has been shown to be more successful when the cells are cultured in three-dimensional environments subject to flowing media (Griffith and Swartz, 2006). This is commonly done in bioreactor systems, some of which involve dilute suspensions of cells and others in which the cells are immobilized in a core cell compartment and subject to media flow. Bioreactors most suited to in vitro cell culturing will be reviewed, emphasizing the materials used and their applications before proceeding to comment on the potential for creating in vitro niche materials for culturing stem cells in scalable bioreactor technologies.

7.2 The need for large volume cell culturing

Before reviewing the different types of bioreactors that have been used to culture stem cells, two examples will be used to emphasize the need to advance bioreactor cell culturing technologies to the point where they can be scaled to the need for stem cells, cell products and tissue rather than simply being used in scientific studies of cellular processes. The first example will be the need for transplantable adult liver tissue and the second is the need to expand human HSCs and control their differentiation into different blood cell products.

7.3 Bioreactors for tissue engineering

For the purposes of this chapter bioreactors will be split into two main groups. The first is the stirred tank reactors in which cells are allowed to float in dilute suspension of circulated media. The second main category is perfusion bioreactors where the cells are immobilized and media is flowed passed them in a circumstance that more closely resembles physiologic conditions. The key difference between bioreactors on the one hand and conventional static culture on the other is perfusion and the potential for uniform media concentrations over time. Perfusion is a particularly important limiting factor in larger, 3D scaffolds because it controls the supply of soluble nutrients, insoluble gases such as oxygen and carbon dioxide and the removal waste products. In general, aggregates of cells or constructs without vascularization are limited to approximately 1 mm in diameter if necrosis of the cells is to be avoided in the middle of aggregate (Sutherland et al., 1986). Vascularization is not usually a prime concern in stirred tank reactors if the cells form small aggregates that are free to move in an excess of media and so the nutrient and gas supply might be expected to be relatively uniform.

Perfusion bioreactors involve cell masses immobilized in a cell compartment that might be expected to experience more problems with spatially and temporally non-uniform concentrations in nutrients, gas and waste products in the media. Direct perfusion reactors have been developed in 2D systems in which cells are cultured in layers for which vascularization is not necessary. However, these systems are limited in their practical ability to culture large cell masses. For such large cell masses 3D culturing will be necessary in scaffold materials that allow cell seeding and flow of media to take place directly through the pores of the scaffold and thereby mitigating, to a certain extent, the need for a vasculature in the early stages of culturing. Direct perfusion systems have been used to successfully culture several cell types such as chondrocytes, for example (Davidson et al., 2002), and allow mineralization and growth of bone cells (Bancroft et al., 2002). An alternative perfusion technology, the hollow fiber membrane reactor, can also control flow condition and perfusion distances at a fine scale (millimeters to hundreds of microns) using the arrangement and the spacing of the hollow fibers (Schmelzer et al., 2010). The fine scale architectural requirements of perfusion have also stimulated the use of planar fabrication patterning techniques and 3D prototyping methods to fabricate bioreactor cores. It is clear that the planar patterning techniques will help in experimental modeling of cell culturing and drug testing using small cell mass bioreactors (Sakai et al., 2010). However, it is not clear how they will be applied to larger cell masses.

Bioreactors have also been used to culture stem cells which, it must be remembered, have the additional requirements of controlling expansion of the stem cells and also controlling the stem cell differentiation into specific lineages. Stirred tank bioreactors have been used to successfully culture HSCs and MSCs for prolonged periods at volumes up to 2 liters volume (King and Miller, 2007). The benefits of perfusion and frequent feeding have been confirmed but optimal culture parameters are still hard to predict. Continuous perfusion of fresh or recycled media has also been shown to increase oxygen transport to cells and bioreactors have increased MSC density within the interior of scaffolds (Zhao et al., 2005) without loss of their multi-lineage differentiation potential (Zhao and Ma, 2005). Also human bone marrow stromal cells that were expanded and differentiated in vitro under perfusion conditions, functioned more effective after they were implanted in mice (Branccini et al., 2005). Importantly, HSC expansion in bioreactors has been promoted by the ability of the bioreactor perfusion to remove inhibitory cytokines produced by their more differentiated progeny (Madlambayan et al., 2005).

The second important issue after perfusion is seeding of the cells in bioreactors. Non-uniform seeding can influence cell behavior and it should be no surprise that this is an important issue in all perfusion bioreactors and also in stirred tank or rotating wall bioreactors used to perfuse scaffolds that do not contain microvasculature. Since the cells have to be immobilized in a scaffold or a cell compartment, provision should be made to uniformly seed the cells into the required location at relatively high cell density (Freed et al., 1997; Holy et al., 2000). Manual methods appeared to result in low seed densities and non-uniform seeding (Kim et al., 1998). Better seeding of polymer scaffolds has been achieved when they are placed in stirred tank bioreactors due to enhanced convection. This suggests perfusion is important even in the seeding step when the media is commonly static (Vunjak-Novakovic, 1996). There are indications that this is also important for culturing of stem cells (Zhao and Ma, 2005). The different types of bioreactors will now be reviewed in more detail.

7.3.2 Perfusion bioreactors

The way in which the perfusion is accomplished for immobilized cell masses can obviously affect the spatial uniformity of the media composition that cells experience over time so perfusion bioreactor control might be expected to present a greater challenge compared to the stirred tank reactors. However, if the perfusion is controlled at sufficiently small length scales, perfusion bioreactors may offer the potential for controlled tissue formation from the multiple cell types often found in the stem cell niches. In consequence, perfusion bioreactors may have a logical advantage in applications where cells must assemble into tissue during in vitro culture (Barron et al., 2003), provide extracorporeal organ support (Jasmund and Bader, 2002) or even drug testing (Wua et al., 2008). Perfusion reactors range in complexity from two-dimensional culture geometries that involve culturing layers of cells. This geometry works well for situations where the levels of structure in tissue are well simulated by layers of cells such as the culturing hepatocytes in the liver plate morphology. Examples include the collagen-sandwich in which hepatocytes are immobilized on a surface between two layers of collagen (Dunn et al., 1991; Tilles et al., 2001). Additionally planar fabrication techniques have contributed to our ability to pattern cells in highly controlled but small three-dimensional systems. However, scaling this approach to applications that require large cell masses may be challenging. Three-dimensional bioreactor material systems will be split into two groups including direct perfusion systems and hollow fiber membrane reactors. In terms of stem cells, three-dimensional perfused culturing systems have achieved notable success in culturing of MSCs, for example, on scaffolds which result in more uniform distribution of cells than static culturing conditions (Zhao and Ma, 2005). Additionally the MSCs have been shown to retain their multi-lineage potential (Chen et al., 2006).

Two-dimensional perfusion bioreactors

Perfusion systems have been developed in which cells are grown on a flat surface, usually a polymer material, over which the liquid media and gas are flowed as shown in Fig. 7.2. As already mentioned, collagen sandwiches in which hepatocytes are immobilized on a surface between two layers of collagen (Dunn et al., 1991) have been used to preserve liver cell function over about 4 to 6 weeks. This has been extended to single cell sheets of hepatocytes co-cultured with single sheets of endothelial cells in order to provide relevant cell associations (Kim et al., 2012). The co-culture was more effective at maintaining albumin secretion over 28 days compared to a single hepatocyte sheets. The two-dimensional culturing methods have also been applied to stem cells including HSCs (Koller and Palsson, 1999). Perfusion resulted in a 20 to 25-fold expansion of bone marrow mononucleated cells over 2 weeks when initially seeded before perfusion to facilitate stromal attachment (Palsson et al., 1993). While the two-dimensional culturing techniques side-step the need for a microvasculature, they are very limiting in terms of the cell mass because of the necessary culture area. The systems have been used to study bioreactor variables but have limited potential for scaling.

Microfabricated bioreators

The potential for fabricating biomimetic culturing structures through the application of soft lithography patterning techniques has been recognized for some time (Whitesides et al., 2001). Indeed a recent example has demonstrated the importance of micropatterning in the co-culturing of hepatocytes and rodent fibroblasts to encourage the formation of liver tissue-like structures (Khetani and Bhatia, 2008). In this case polydimethylsiloxane stencils were used to pattern 500 μm islands of collegen-1 hydrogel on a polystyrene substrate. Human hepatocytes were then allowed to attach to the collegen-1 islands and rodent fibroblasts were allowed to attach to the polystyrene substrate that remained uncovered by hepatocytes. Albumin production and CYP450 activity was used to demonstrate improved liver specific function compared to culturing of pure hepatocytes under the same conditions. While it is quite clear that this arrangement might facilitate drug testing on hepatic function, for which it intended, it is not clear that such patterning techniques will address applications that require large cell masses such as the extracorporeal liver support that could bridge liver disease patients to transplantation. The potential for 3D bioreactor architectures and the control exercised over oxygenation using 3D microfabrication techniques, microelectromechanical techniques or modular small-scale reactors to support culturing in large liver cell masses with microvascularization has been reviewed elsewhere (Sakai et al., 2010). This will be challenging and has yet to be clearly demonstrated.

Direct perfusion bioreactors

The simplest types of three-dimensional bioreactors represented in tissue engineering (Martin et al., 2004) are the direct perfusion reactors in which the cells are seeded onto a scaffold in the bioreactor core and media are forced to flow directly over the cells through the pores of the scaffold which is attached directly to an external recirculation system (Wendt et al., 2003). Understandably, most of these systems use porous polymers scaffolds in the reactor core. For example, direct perfusion was shown to be important to in vitro culturing of rat cardiomyocytes on collagen sponge, the thickness of which is limited by oxygen diffusion to 200 μm. The interior of the larger scaffolds remained largely acellular. Incorporating the collagen into a perfusion system for both seeding, 10 minutes after inoculation and also for culturing gave significantly higher number of live cells and higher cell viability throughout the 1.5 mm thick scaffold. In response to electrical stimulus the perfused culture contracted synchronously whereas the dish cultured scaffold contracted with arrhythmic contractile patterns (Radisic et al., 2004).

One possible solution to the difficulty of culturing in thicker scaffolds, and the apparent need for microvascularization, is to decellularize the connective tissue of a real organ, usually from an animal source, and then recellularize it. This concept takes advantage of the natural vasculature, which has been optimized for perfusion and the decellularization–recellularization technique has been demonstrated for a variety of tissues (Yoo et al., 1998; Badylak, 2007; Ott et al., 2008). This approach has been applied to liver tissue (Uygen et al., 2010), which has been largely limited to very thin cultures due the high oxygen consumption of hepatocytes (Ohashi et al., 2007; Wang et al., 2008). Importantly, this decellularize–recellularize approach to the liver organ preserved the extracellular matrix and the natural vasculature that could be connected directly to a bioreactor circulation system via the portal vain. The seeding in the recellularization was then accomplished via a four-step portal vein perfusion that resulted in 95% engraftment of the rat hepatocytes. The graft was perfused for 5 days and the hepatocytes were mainly distributed in the parenchyma around larger vessels suggesting they had penetrated local sinusoids. Unfortunately the interior of the scaffold away from large vessels appeared to remain acellular. Nevertheless, metabolic activity measured by urea production was greater than for a hepatocyte sandwich culture although albumin production was not statistically different. This suggests a viable way of moving to larger three-dimensional structures for liver tissue where the high rates of oxygen consumption by primary hepatocytes (Cho et al., 2007). would be expected to result in damage.

Direct write methods have recently been extended to patterning positive microvascular networks in thicker cell masses. A positive copy of a micro-vasculature is first fabricated and then it can be directly surrounded with cells and extracellular matrix before the positive copy of the vasculature is dissolved away (Miller et al., 2012). This concept had been previously used for rapid direct casting of sugar fibrils to produce a degradable sugar network (Bellan et al., 2009) which was consequently impregnated with cells in an extracellular matrix hydrogel. The sugar positive copy of the microvasculature must then be dissolved out before the cells die. The patterned vasculature is an important requirement that allows the scaffold to be attached directly to an external perfusion system. The effect of this type of structure on the culturing of stem cells has yet to be explored in detail.

Ceramic direct perfusion reactors are relatively rare in the culturing of stem cells. However, some ceramic reactor cores have been used to culture sheep MCSs in a cylindrical three-dimensional foam tube that was 14 mm in diameter and 30 mm long and directly connected to a bioreactor core inlet port (Xie et al., 2006) It was found that the daily glucose consumption in the perfused bioreactor was much higher than for a static culturing. Additionally the cells grew throughout the perfused scaffold but only penetrated into two or three layers of pores from the external surface for samples cultured under static conditions. The ceramic scaffold was processed using a templating technique in which an organic sponge was cut to the correct dimensions and then coated with β-tricalcium phosphate (β-TCP) powder before sintering. In this reactor arrangement the media were forced to enter the reactor core directly through the TCP scaffold and flow into the bioreactor tank. The scaffold had a porosity of 75% with spherical pores of 530 μm with gates roughly 150 μm in size.

Rapid prototyping techniques such as drop-on-demand have not gone unnoticed by the tissue engineering community because of their shaping flexibility and the potential for manufacturing ceramic scaffolds with tailored pore structures and external dimensions the match the tissue gap in specific patients. Ceramics ink-jet printing has been used to produce hydroxyapatite (HAp) scaffolds (Leukers et al., 2005) and β-TCP scaffolds (Khalyfa et al., 2007). Similarly extrusion-based rapid prototyping methods have been used to fabricate tissue engineering scaffolds for in vivo applications (Cesarano et al., 2005). Negative molding techniques have also been applied to calcium phosphate materials for bioreactors (Woesz et al., 2005). In terms of stem cells, one study compared two rapid prototyping techniques in terms of their effect on the culturing of a mouse bone marrow clonal stromal cell line (Woesz et al., 2005). Both techniques produced HAp scaffolds with controlled porous lattice structures with pore sizes in the range 300 μm separated by 500 μm thick struts although the pore volume fractions were low, in the range 37–44%. Cell proliferation seemed higher in the drop-on-demand scaffolds while cell differentiation seemed more prevalent on the negative mold technique. The location of the scaffolds in a perfusion bioreactor after 3 days of static culture also improved the culturing results.

Ceramic scaffolds have been used to successfully in vitro culture rat hepatocytes that then supported increased albumin production when transplanted into Nagase analbuminemic rats (Higashiyama et al., 2003). Since that time in vitro culturing of liver cells has been attempted on ceramic scaffolds made with a technique compatible with a positive–negative casting model in which a positive copy of a natural vasculature is produced by injecting an organ vasculature with a thermosetting resin as shown in Fig. 7.3. A ceramic foam scaffold was cast around the positive thermoset copy before the ceramic was sintered and burnt-out to leave a negative copy of the vasculature in the porous ceramic scaffold (Finoli et al., 2012). This ceramic foam structure has been used to culture human liver cells in a hydrogel extracellular matrix. Of the sample cast using liver cells in an extracellular matrix, albumin measurements showed that adult liver function was best supported in the ceramic scaffold over a period of 28 days with collagen 1. Figure 7.4(a) shows the structure on which liver cells could be seeded in collagen 1 hydrogel and Fig. 7.4(b) shows how the processing can be manipulated to produce a ceramic wall on the negative copy of the microvascular which could act as a substrate wall for co-seeding endothelial cells.

Hollow fiber membrane reactors

The culturing of stem cells in hollow fiber membrane reactors is still not commonplace and the expansion of HSCs in these systems has been less successful to this point compared with stirred tank reactors. In hollow fiber membrane reactors, cells are held in cell compartments between the hollow fiber membranes through which the media will flow (Jasmund and Bader, 2002). The hollow fiber membrane bioreactors are capable of controlled continuous perfusion immediately after seeding because the perfusion flow through the lumens of the fiber is separated from the cell compartment by the porous membrane fiber walls. The pore in the fiber walls (Fig. 7.5) allows the medium to diffuse into the cell compartment. The materials used to make the fibers are usually polysulfone or cellulose (Whitford and Cadwell, 2009). Figure 7.6 shows a reactor of this type. The liquid medium and the less soluble gases are usually pumped through the lumens of the fibers in a closed loop that includes a reservoir. The reservoir is then exchanged gradually over time. The advantage of this system is that the cells are exposed to controlled media concentration gradients that are relatively uniform over time when compared with static culturing which has the attendant need for periodic media replacement. Efficient mass exchange can be provided by controlling both the flow rate of media through the fiber lumens and the arrangement of the fibers in the core. An example of a small-scale hollow fiber membrane reactor has been used to culture human fetal liver cells (Schmelzer et al., 2010). This reactor had a diameter of approximately 20 mm and a 1 ml capacity specifically to study cell culturing in a small volume that can be directly observed with an optical microscope. The reactor housing was cast in polyurethane then two layers of hollow fiber membranes were placed through the chamber and connected to external ports. More polyurethane was applied to seal the outside of the chamber between applications of the fiber layers and then a transparent coverslip was fixed in place containing a silicone port for inoculation of the cells into the cell compartment. The medium supply was delivered with a cross-perfusion mode with counter flow occurring through adjacent fiber membranes. This creates a driving force for flow that is designed to maximize perfusion. The fibers used for the medium supply were made from polyethersulfone with an outer diameter of 500 μm, an inner diameter of approximately 300 μm and a wall pore size of approximately 0.5 μm.

The independent oxygenation fibers, 280 μm in outer diameter and 200 μm inner diameters, were also used for carbon dioxide perfusion used to control the pH in the chamber. There were 25 gas capillaries placed alternately between and an equal number of medium capillaries in each fiber layer in the cell compartment. The two fiber layers were arranged so that the fiber membranes containing medium in the lower layer were directly above the fiber membranes containing medium in the upper layer but the flow direction of the medium was opposite direction in the fibers in each layer. Also the flow was opposite in adjacent medium perfusion fiber membranes within each layer even though they were separated by oxygenation fibers. This illustrates the flexibility and precise spatial control of flow conditions that can be achieved by the use of independent media and gas supplies and the arrangement of the fibers in the core. In this example the liver cells were mixed with precursor of a hyaluronic acid (HA) and then placed into the bioreactor core between the hollow fiber membranes before the gel crosslinked to mimic the extracellular matrix of the fetal liver. The dynamic perfusion conditions of the hollow fiber membrane bioreactor showed a tenfold increase in albumin secretion and much higher CYP4503A4 expression compared to a static culture over a period of 10 days. This suggests that the perfusion conditions in the bioreactor had markedly promoted differentiation of fetal hepatoblasts into mature hepatocytes over static culture, although proliferation assays suggest about the same cell numbers existed under both conditions. This may be explained by the fact that the perfusion in the bioreactors was thought to provide constant mass transfer which would be expected to remove albumin from the cell compartment and thereby induce increased albumin production by the liver cells compared to a static culture where the albumin would remain in contact with the cells.

There has been some success in using hollow fiber membrane reactors in extracorporeal liver support (Nyberg et al., 1993; Rozga et al., 1993) to bridge patients with chronic liver disease or acute liver failure to liver transplant. The important role of the liver in metabolism of toxins and regulation has incentivized the direct use of human liver cells rather than porcine cells to avoid the potential for contamination with viruses. One specific example (Sauer et al., 2002) of a bioreactor used for extracorporeal liver support sustained eight patients with over different periods of time from 7 to 144 hours. Two patients with acute liver failure were successfully bridged to transplantation as were two others with acute chronic liver failure. Two other patients were successfully bridged to retransplantation after failure of their initial liver grafts and two patients who were unsuitable for transplantation were supported over a period of acute deterioration. In this bioreactor system the liver cells for the bioreactor were obtained from livers considered to be unsuitable for liver transplant and possessed lower cell viability. The cells were harvested by a five-step collagenase perfusion procedure through the liver portal vein and resulted in a cell suspension consisting mainly of hepatocytes that was injected into the bioreactor to produce a cell mass of 400 to 600 grams that could function as equivalent to a single liver lobe. The bioreactor was composed of three independent and interwoven arrays of hollow fiber membranes used for independent media inflow, media outflow and one for independent oxygenation and carbon dioxide removal. The space between the hollow fiber membranes was the cell compartment where the liver cells spontaneously formed aggregates with evidence of liver tissue-like structures. These were usually located on the outside of the hollow fiber membranes.

7.4 The future of large bioreactors through in vitro mimicry of the stem cell niche

While most differentiated cell types can increase their number by simple cell division, stem cells have an important advantage in terms of enhanced expansion rate and the ability to differentiate into a number of different mature cell types. MSCs have been identified in various tissues in the body and differentiate into cells that make up connective tissue including adipose tissue and neural cells. Organ-specific stem cells have also been identified, including: HSCs that can differentiate into a range of cells including red blood cells, plasma and cells for the immune system and liver specific precursor cells such as hepatoblasts sourced from adult and fetal livers can differentiate into mature hepatocytes.

Organ-specific stem cells have been found to be associated with specific microenvironments in the body commonly referred to as the stem cell niche (Ohlstein et al., 2004; Moore and Lemischka, 2006). The first such niche was reported for HSCs in 1978 (Schofield, 1978). Since then a great deal has been discovered about stem cell niches for several human organs including: the small intestine (Barker et al., 2007), brain (Valcanis et al., 2001), muscle (Sherwood et al., 2004) and bone marrow (Sacchetti et al., 2007). Several aspects of the microenvironment have been found to be of central importance to determining stem cell fate in vivo. These include: the presence of extracellular matrix, local cell associations, perfusion conditions controlled by the surrounding tissue and particularly the microvasculature. The niche must support functions including expansion to maintain the stem cell population as well as differentiation into progenitors that are committed to specific mature cell lines. The details of the microenvironment will obviously be different in different organs and the microenvironment for expansion of stem cells may be somewhat different from the microenvironment for differentiation, as in the case of HSCs.

In the bone marrow niche (Fig. 7.7) proximity to a vasculature controls the differentiation of HSC into blood cell progenitors which are then released into the blood stream (Kopp et al., 2005; Wilson and Trumpp, 2006). While a full description of the niches and their categorization is beyond the scope of this chapter, some of the details of the bone marrow niche can be illustrative. The bone marrow niche is the microenvironment that houses HSCs in adult humans and produces about 500 billion cells per day (Fliedner et al., 2002). In the embryo the HSCs reside in the yolk sac and then move to the fetal liver and eventually to the bone marrow cavities before birth. Migration then continues to other bone marrow cavities during maturation to the adult (Wright et al., 2001). In the bone marrow the HSCs constitute a very small fraction of the cell population. Obviously the niche will contain more than just the HSCs. The bone marrow contains a vascular network that extends from arterial vessels and divides into sinusoids separated by smaller capillaries. The HSCs appear to be preferentially located close to the surface of the trabecular bone (Calvi et al., 2003) where they remain relatively quiescent and conditions appear to promote expansion of the stem cell population. In order to participate in differentiation, the HSCs must migrate to the vascular zone before they reach an environment that promotes the differentiation process (Heissig et al., 2002). The high concentration of extracellular calcium adjacent to the endosteal surface is thought to play an important role in locating the HSCs in the bone marrow niche due to calcium sensing receptors expressed by HSCs (Adams et al., 2006). This may control the homing of the HSCs from the fetal liver to the developing bone marrow niche and preserving the location of adult hematopoiesis. Control of these receptors may be a method to control engraftment of transplanted HSC or mobilize them into the blood stream when they need to be harvested from peripheral blood. Indeed the coating of cellulose sponge with hydroxyapatite has been shown to promote the homing of circulating hematopoietic and mesenchymal progenitors in rat subcutaneous tissue (Tommila et al., 2009). Therefore it is logical to expect that the design of bioreactor core, whether they are direct perfusion or hollow fiber membrane, should attempt to recreate these basic characteristic of the niche to control HSC fate in perfusion reactors. This may start with a HAp scaffold that simulates trabecular bone onto which osteoblasts, HSC and the attendant stromal cells can be seeded. It is very important to include perfusion in the bioreactor core and in the case of HSC the local environment around the vasculature may promote differentiation into progenitors that are then released from the bioreactor core. This must inform the design of ceramic bioreactor cores in which vascularization channels are included, whether they are formed as a negative copy of a vascular system in a direct perfusion core or as an arrangement of fibers in a hollow fiber membrane reactor. Additionally, in the case of bone marrow perfusion bioreactors, the release of calcium into the bioreactor core could be controlled by using scaffolds with biphasic calcium phosphates in which one phase (TCP) dissolves and resorbs much faster than the other (HAp).

Some bioreactors will involve stem cells that require an attendant extracellular matrix to be present in the niche, especially if the cells are sorted into cell fractions that are not able of producing their own extracellular matrix. Hydrogels have been used in tissue engineering to mimic the extracellular matrix that interacts with cells by further controlling: the diffusion of nutrients, metabolites and growth factors, biomechanical cues and adhesion sites with controlled degradability. (Lutloff and Hubbell, 2005; Seliktar, 2012) These behaviors combined with the ability of injectable hydrogels to subsequently crosslink without negative effects on the cells makes them particularly attractive for use in bioreactor cores. There are several types of hydrogel that have been used in tissue engineering including natural polymers such as collagen, hyaluronate, fibrin, alginate, agarose and chitosan that are biocompatible but tend to show batch to batch variability as well as synthetic polymers such as polyacrylic acid, polyethylene oxide, polyvinylalcohol, polyphosphazene and polypeptides for which some of the precursors may have negative effects (Lee and Mooney, 2001). Hydrogels have also been demonstrated to play an important role in controlling the fate of stem cells. One of the previously mentioned liver tissue bioreactors provides a good example of the importance of this choice. When fetal liver cells were cultured in a hollow fiber membrane reactor the choice of HA over collagen 1 was made because previous studies had shown that human liver stem cells expressed CD44H which is an HA receptor (Dan et al., 2006). HA was capable of maintaining the immature phenotype in static culture but also allowed the differentiation into mature hepatocytes when perfused. Again this suggests that scaffold materials to be used in bioreactor cores must be carefully considered for a specific task. If the goal of culturing liver progenitors, from fetal liver cells, is to differentiate them and spontaneously form liver tissue capable of adult liver function then it will be necessary to properly perfuse the core and allow tissue formation. This might best be done with a scaffold material that contains a three-dimensional microvasculature which is capable of completely dissolving during tissue formation. Better differentiation and hepatic function of rat fetal liver cells has been shown in static culture on 3D poly-L-lactic acid (PLLA) scaffolds when compared with monolayer controls (Hanada et al., 2007). Similarly improved behavior was seen for rat fetal liver cells cultured on 3D polyvinyl formal resin constructs (Koyama et al., 2009). Therefore a dissolvable three-dimensional scaffold consistent with the goals of culturing should contain a microvascular to promote differentiation to hepatocytes capable of albumin production. This is only one example in which the details of the niche are used to instruct the design and material selection of perfusion bioreactors and this trend should continue.

7.5 Conclusions and future trends

Bioreactors will continue to be optimized for culturing differentiated cells, but importantly they will also be required to expand and maintain stem cells and progenitor cells in the numbers required for the range of proposed clinical applications in regenerative medicine. Many stem cells are derived from specific microenvironments or niches in the body and the available cell sources are often limited, as in the example of the liver where the cells must be harvested from donated organs and must compete with the needs of transplantation, the only clinical therapy available for chronic liver disease or acute liver failure. Obviously the widespread use of these and other stem cells will require the use of dedicated bioreactor technologies for expanding and maintaining specific cells for periods of days and weeks. At the present time the results of bioreactor culturing are mixed in the case of some important stem cells from specific organs. For example, HSCs have been successfully expanded in 2D perfusion reactors and stirred tank bioreactors. However, the differences in bioreactor design, cell sources, media composition and operating conditions make comparisons difficult outside individual studies. Even when the conditions are optimized, stirred tank reactors and flat plate perfusion reactors have difficulty operating under physiologically relevant conditions and cell densities. Nonetheless, these types of bioreactor will serve as important technologies for expanding cells for cell therapies and also performing drug testing where relatively small amounts of cells are needed.

If the range of capabilities are to be extended to larger cell masses for tissue engineering constructs, bioartificial organs or extracorporeal organ support it will be necessary to design bioreactor systems that can culture cells in three dimensions at higher cell densities and thereby approximate physiologic conditions over larger length scales. Two-dimensional bioreactors, such as those developed for culturing liver cells, will require large area stackable modules to house enough cells to support organ function. This could be technically challenging to control. Alternative three-dimensional bioreactors required to culture larger cell masses will probably immobilize cells in a core, like existing perfusion reactors, but they must also control perfusion in the core over small length scales while protecting the stem cells from the high shear conditions experienced when they are directly exposed to flowing media. In principle, hollow fiber membrane bioreactors are one category that are capable of this at longer length scales (approximately a millimeter in fiber spacing); however, greater control over the local culturing conditions experienced by cells is required. Other approaches may involve large-scale microfabrication, if indeed it is practical to construct and seed large volumes using these methods. Alternatively a vascular structure may be grown in the bioreactor core during bioreactor operation to the facilitate cell perfusion in the interior.

As mentioned earlier, bioreactors may also be used to vascularize tissue engineering constructs before implantation and thereby avoid the all too common problem of necrosis inside the implanted construct resulting from very slow intrusion of the natural vascular system into the interior of the constructs after implantation. The design of appropriate three-dimensional bioreactor cores will be a complex engineering problem, requiring seeding and immobilization of multiple cells, control of local cell associations in the bioreactor core and perfusion on the appropriate length scales by an artificial or spontaneously generated vascular structure. Obviously the architecture of the core and the culturing conditions will have to be optimized for each stem cell type. This optimization might be profitably accomplished using the conditions of the stem cell niche as a model by which control could be exercised over the stem cells or progenitors; the association with supporting cells and the architecture and the chemical signaling. The bioreactor core must also be able to simulate the events that cause cells to expand or differentiate into the desired progenitor cells within the same core. The design principles involved in this undertaking have yet to be widely applied in bioreactors that mimic specific stem cell niches. However, these in vitro systems will not only hold the promise of scientific models that enrich our understanding of stem cell niches, they will also provide a crucial range of biotechnology support tools for regenerative medicine therapies. Such therapies will ultimately rely on our ability to expand and manipulate stem cells from the available stem cell sources in sufficient quantities.

7.6 Acknowledgements

The author gratefully acknowledges the patient guidance of Professor Jorg Gerlach and Dr Eva Smelzer of the McGowan Institute of Regenerative Medicine at the University of Pittsburgh concerning the subject of bioreactors and the attendant human cell biology during our collaboration. The author would also like to thank Anthony Finoli for his comments on this manuscript. Finally the author would also like to acknowledge the financial support of the National Science Foundation grant 0900254.