• Open Access

Recent advances towards development and commercialization of plant cell culture processes for the synthesis of biomolecules

Authors


Correspondence (fax 413 545 1647;
email sroberts@ecs.umass.edu)

Summary

Plant cell culture systems were initially explored for use in commercial synthesis of several high-value secondary metabolites, allowing for sustainable production that was not limited by the low yields associated with natural harvest or the high cost associated with complex chemical synthesis. Although there have been some commercial successes, most notably paclitaxel production from Taxus sp., process limitations exist with regards to low product yields and inherent production variability. A variety of strategies are being developed to overcome these limitations including elicitation, in situ product removal and metabolic engineering with single genes and transcription factors. Recently, the plant cell culture production platform has been extended to pharmaceutically active heterologous proteins. Plant systems are beneficial because they are able to produce complex proteins that are properly glycosylated, folded and assembled without the risk of contamination by toxins that are associated with mammalian or microbial production systems. Additionally, plant cell culture isolates transgenic material from the environment, allows for more controllable conditions over field-grown crops and promotes secretion of proteins to the medium, reducing downstream purification costs. Despite these benefits, the increase in cost of heterologous protein synthesis in plant cell culture as opposed to field-grown crops is significant and therefore processes must be optimized with regard to maximizing secretion and enhancing protein stability in the cell culture media. This review discusses recent advancements in plant cell culture processing technology, focusing on progress towards overcoming the problems associated with commercialization of these production systems and highlighting recent commercial successes.

Introduction

In 1902, Gottlieb Haberlandt first attempted to initiate plant cell cultures (Haberlandt, 1902). Although he was unable to induce cell division, he is recognized as the founder of plant cell culture science (Georgiev et al., 2009). Plant cell cultures can now be created from virtually any plant through the isolation of plant tissue (Figure 1). Explants (i.e. isolated plant tissue) are sterilized using an exogenous chemical treatment and plated on solid growth media that is formulated to contain the growth hormones and nutrients necessary for each species (Hall, 2000). With correct media composition, explants proliferate into a callus of dedifferentiated cells, which can be screened for product(s) of interest and then isolated and transferred to liquid medium to create suspension cultures. Differentiated organ cultures of roots, shoots and embryos can also be created and maintained in vitro. Organ cultures often produce similar concentrations of secondary metabolites to the full plant but are limited by the prohibitive expense of large-scale culture (Verpoorte et al., 2002). Hairy root cultures are generated through infection of plant material with Agrobacterium rhizogenes. Whereas regular root cultures grow slowly and require exogenous phytohormone supply, hairy root cultures can stably produce high levels of secondary metabolites, do not require growth hormones and can be adapted to reactor systems (Giri and Narasu, 2000; Srivastava and Srivastava, 2007).

Figure 1.

 Development of a plant cell suspension culture. Sterile explants from the whole plant (a) are plated on solid culture medium. With the correct nutrients and hormones combination, explants grow into a callus of dedifferentiated cells (b). Callus cells are transplanted into liquid media, creating a suspension culture (c), which can be scaled up for growth and production in a controlled bioreactor (d). To avoid contamination, aseptic technique must be used to maintain the cell cultures, resulting in increased manufacturing costs.

Plant cell culture was initially considered a potential option for the production of high-value secondary metabolites, which is limited commercially owing to environmentally unsustainable natural harvest and complex chemical synthesis (Roberts, 2007). Recently, commercial plant cell culture has been extended to the production of heterologous proteins. Plant systems are beneficial for protein production because they allow for proper protein assembly, folding and glycosylation, without the threat of contaminants such as endotoxins, pathogens and oncogenic DNA associated with microbial and mammalian cell culture (Ma et al., 1995; Commandeur et al., 2003; Twyman et al., 2003). Plant cell culture also allows for isolation of genetically modified plants from the environment, eliminating risk of transgene migration (Fox, 2006). Conditions are more easily controlled compared to whole plant systems, allowing for more consistent yields (James and Lee, 2001), and secretion of proteins to the culture media can significantly decrease downstream processing costs (Shih and Doran, 2009). The production of both secondary metabolites and proteins via plant cell culture is limited by low product yields and unpredictable scale-up. Additionally, secondary metabolite accumulation over time within a single cell line and among cultures of the same cell line is often variable, which is likely related to cellular heterogeneity (Kolewe et al., 2008). Despite these limitations, several secondary metabolites and heterologous proteins have been successfully produced on a commercial scale using plant cell culture (Table 1).

Table 1.   Products produced commercially via plant cell culture (Kolewe et al., 2008; Obembe et al., 2010)
ProductSpeciesManufacturerUse/notes
  1. *http://www.protalix.com.

Secondary metabolites
 AnthocyaninsEuphorbia miliiNippon Paint Co., Ltd., Osaka, JapanTextile dyes
Colouring agents for fruit juices, wine and other beverages
Aralia cordata
 ArbtinCatharanthus roseusMitsui Chemicals, Inc., Tokyo, JapanPigment
Antiseptic
 BerberinesCoptis japonicaMitsui Chemicals, Inc.Anticancer
Antibiotic
Anti-inflammatory
Most commonly used for bacterial diarrhoea and intestinal parasite and eye infections
Thalictrum minus
 BetacyaninsBeta vulgarisNippon Shinyaku Co., Ltd., Kyoto, JapanRed to red-violet pigment
Food colourant and dye
 CarthaminCarthamus tinctoriusKibun Foods, Inc., Tokyo, JapanRed pigment
Food colourant and dye
 Echinacea polysaccharidesEchinacea purpureaDiversa, Ahrensburg, GermanyImmunostimulant
Anti-inflammatory
Echinacea angustifolia
 GeraniolGeraminea spp.Mitsui Chemicals, Inc.Essential oil
Primary component of rose, palmarosa and citronella oils
 GinsengPanax ginsengNitto Denko Corporation, Osaka, JapanDietary supplement
Wild ginseng stem cellsUnhwa Biotech Corp., Jeonbuk, KoreaDietary supplement
Cosmetics
 PaclitaxelTaxus spp.Phyton Biotech, Inc.Anticancer
FDA approved for the treatment of ovarian, breast and lung cancers
Largest application of commercial plant cell culture
Genexol®– Samyang Genex, Seoul, South Korea
 PodophyllotoxinPodophyllum spp.Nippon Oil, Tokyo, JapanAnticancer
Starting compound for the anticancer agents etoposide and teniposide (Chattopadhyay et al., 2002)
 Rosmarinic acidColeus blumeiA. Nattermann & Cie. GmbH, Cologne, GermanyAnti-inflammatory
 ScopolamineDuboisia spp.Sumitomo Chemical Co., Ltd., Tokyo, JapanAnticholinergicum
Antimuscarinic
Used in the treatment of motion sickness, nausea and intestinal cramping
 ShikoninLithospermum erythrorhizonMitsui Chemicals, Inc.Red pigment
Antibiotic
Heterologous proteins
 Vaccines
  HN protein of Newcastle  disease virusTobacco suspension culturesDow AgroSciences LLC, Indianapolis, INVaccine for Newcastle disease in poultry
First FDA approved plant-derived vaccine
Currently not on market
Therapeutic proteins
 Human glucocerebrosidase (ULYPSO)Carrot suspension culturesProtalix Biotherapeutics*Enzyme replacement therapy for Gaucher’s disease
Licensed by Pfizer, Inc.
Completed Phase III clinical trial in September 2009
Awaiting FDA approval
 Recombinant α-galactosidase-A (PRX-102)Carrot suspension culturesProtalix Biotherapeutics*Potential Fabry disease treatment
In preclinical development
 PEGylated recombinant human acetylcholinesterase (PRX-105)Carrot suspension culturesProtalix Biotherapeutics*Potential use in biodefence against nerve agent attacks
Completed Phase I clinical trials
 Anti-tumour necrosis factor (PRX-106)Carrot suspension culturesProtalix Biotherapeutics*Potential therapeutic for autoimmune diseases, such as rheumatoid arthritis
In preclinical development

Plant secondary metabolites and production routes

Plant secondary metabolites are low-molecular weight compounds that aid in the adaptation of plants to their environment but do not directly affect growth and development. Because of their significant biological activity, plant secondary metabolites have been used in traditional medicine for centuries. Currently, over 60% of anticancer drugs and 75% of drugs for infectious disease are either natural products or analogues of natural products (Newman et al., 2003; Cragg and Newman, 2009). Secondary metabolites are also commonly used as insecticides, dyes, flavours and fragrances (Srivastava and Srivastava, 2007).

Natural harvest

One of the main problems associated with the commercial supply of secondary metabolites is limited compound availability. Secondary metabolites typically represent <1% dry weight of the plant, so natural harvest is often impractical (Georgiev et al., 2009). For instance, 340 000 kg of Taxus bark or 38 000 trees were required to meet the 25 kg per year demand for the anticancer drug paclitaxel (Taxol®; Bristol-Myers Squibb, New York, NY) (Cragg et al., 1993). Harvesting is also limited by seasonal availability, species abundance and plant growth rate (Roberts, 2007). Despite these challenges, several compounds continue to be harvested from their native plant owing to lack of better commercialization options (Table 2). As an alternative to natural harvest, secondary metabolites can also be synthesized and supplied through three general approaches: total or partial chemical synthesis, heterologous expression of the biosynthetic pathway in other organisms and in situ production via plant cell culture.

Table 2.   Examples of secondary metabolites commercialized through natural harvest
UseCompound/productPlant sourceYield (% dry weight)References
Anticancer agentCamptothecinCamptotheca acuminata0.4–0.5Lopezmeyer et al. (1994)
VinblastineCatharanthus roseus0.01Ishikawa et al. (2009)
VincristineC. roseus0.0003Ishikawa et al. (2009)
AnalgesicMorphinePapaver somniferum0.05–0.1Odell et al. (2008)
CodeineP. somniferum0.01–0.1Odell et al. (2008)
Malaria treatmentArtemisininArtemisia annua1.5Covello (2008)
InsecticidePyrethrinsTanacetum cinerariifolium1.8–2.5Morris et al. (2006)
Neem (Azadirachtin)Azadirachta indica0.1–0.9Morgan (2009)

Chemical synthesis

Many plant secondary metabolites are produced economically through total chemical synthesis. Vanillin is the most popular flavour compound, but <1% of the annual demand is met through extraction from Vanilla planifolia. Because of the limited availability and cost of the natural compound, it is most commonly chemically synthesized from the substrates guaiacol, eugenol or lignin (Schwab et al., 2008). One limiting factor for chemical synthesis of many secondary metabolites is their large size and presence of multiple chiral centres (Figure 2). For instance, morphinan alkaloids (e.g. morphine and codeine) have five centres of chirality, so the complex chemical synthesis route is not cost efficient (Gerardy and Zenk, 1993). Semisynthesis of the antimalarial compound artemisinin from both artemisinic acid and arteannuin B has been achieved, but even though artemisinic acid is present at concentrations 8–10 times higher than artemisinin in Artemisia annua, these routes are still not commercially feasible (Roth and Acton, 1989; Nowak and Lansbury, 1998). Two different routes for the total synthesis of paclitaxel were developed in 1994, but the processes involved over 40 reactions, utilized harsh solvents, and had low product yields, making them economically and environmentally unfavourable (Holton et al., 1994a,b; Nicolaou et al., 1994). Commercial paclitaxel was achieved through semisynthesis from two paclitaxel precursors that could be extracted from the needles of Taxus, which is discussed below in Success Stories: Phyton Biotech, Inc. (Ahrensburg, Germany).

Figure 2.

 Examples of the complex structures of commercially available secondary metabolites. (a) Vanillin, a food flavouring derived from Vanilla planifolia; (b) artemisinin, an antimalarial compound from Artemisia annua; (c) morphine, an analgesic from Papaver somniferum; (d) camptothecin, an anticancer compound from Camptotheca acuminata; and (e) paclitaxel, an anticancer compound from Taxus spp.

Heterologous production

Heterologous synthesis of plant secondary metabolites has been investigated in bacteria, yeast and alternative plant species. Engineered microbial species such as Escherichia coli and Saccharomyces cerevisiae are ideal because of their fast doubling times compared to plant species (minutes vs. days), inexpensive carbon sources, ease of genetic modification and well-established scale-up technologies (Chang and Keasling, 2006; Chang et al., 2007; Roberts, 2007). Microbial production routes could overcome the inherent production variability associated with plant suspension cultures, but synthesis of terpenes (e.g. artemisinin and paclitaxel) and other complex compounds through transfer of complete pathways in E. coli is limited by the difficulty of cytochrome P450 (CYP450) expression. CYP450s often lose functionality in E. coli owing to improper folding, translation and insertion into the cell membrane, inefficient cofactor pools and a lack of CYP450 reductases (Chang et al., 2007; Ajikumar et al., 2010). For functional CYP450 expression, the membrane anchor can be engineered to achieve proper membrane translation or CYP450 chimeras can be created which mimic proteins found in the native plant (Chemler and Koffas, 2008). This approach has been successfully employed for the production of hydroxylated flavonoids (Leonard et al., 2006) and isoflavones (Leonard and Koffas, 2007). The secondary metabolic pathways for compounds such as paclitaxel have also been introduced into other plant systems, such as Arabidopsis (Besumbes et al., 2004), tomato (Kovacs et al., 2007) and Physcomitrella (Anterola et al., 2009). The heterologous production of plant secondary metabolites has recently been reviewed in (Chemler and Koffas, 2008; Zhang et al., 2011). The main limitation of heterologous production is the lack of fully characterized secondary metabolic pathways. In addition to using microbes for the production of natural products, analogues of natural products have been created by introducing genes with modified substrate specificity to plant-derived pathways that can be expressed in microbes, such as the flavonoid (Katsuyama et al., 2007; Werner et al., 2010) and carotenoid pathways (Schmidt-Dannert et al., 2000). Similar studies have also been conducted on the more complex alkaloid biosynthetic pathway in Catharanthus roseus cell cultures (Runguphan and O’Connor, 2009; Runguphan et al., 2009).

Since 2000, biosynthetic vanillin has been produced and commercialized through the microbial fermentation of ferulic acid, with molar yields as high as 54.5% in Streptomyces sp. (Hua et al., 2007) and 86.6% in E. coli (Lee et al., 2009). This production route is more expensive than the chemical synthesis of vanillin, but consumer value is increased because it can be labelled as a natural flavouring (Schwab et al., 2008). To decrease biosynthesis costs, vanillin production using glucose ($0.30/kg) as a substrate, rather than ferulic acid ($5/kg), is being investigated. Vanillin was first produced from glucose through the conversion of glucose to vanillic acid by heterologous E. coli, which was then converted to vanillin by aryl aldehyde dehydrogenase extracted from Neurospora crassa (Li and Frost, 1998). Complete biosynthesis from glucose was achieved in S. cerevisiae and Schizosaccharomyces pombe, reaching vanillin concentrations of 45 and 65 mg/L, respectively (Hansen et al., 2009). In silico design was used to engineer S. cerevisiae for higher production levels, and experimental implementation of the optimized strains resulted in a maximum production level of 500 mg/L (Brochado et al., 2010).

Early genes in the paclitaxel biosynthetic pathway, which contains 19 putative pathway steps from geranylgeranyl diphosphate (GGPP) (Croteau et al., 2006), have been introduced into S. cerevisiae and E. coli. The introduction of GGPP synthase and taxadiene synthase (TASY) into S. cerevisiae resulted in the production of 1 mg/L of taxadiene (DeJong et al., 2006). Additional pathway genes were introduced in this study, but expression levels of the first CYP450 enzyme, taxadiene 5α-hydroxylase (T5αH), were low, resulting in only 25 μg/L of the next intermediate, taxadiene-5α-ol. By eliminating the effect of native metabolism on the heterologous biosynthetic pathway and expressing a codon-optimized TASY, taxadiene levels of 8.7 mg/L were ultimately achieved (Engels et al., 2008). In E. coli, the two precursors to GGPP, isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP) are synthesized through the 1-deoxyxylulose-5-phosphate (DXP) pathway. To increase the availability of GGPP, enzymes from the DXP pathway were coexpressed with GGPP synthase and TASY in E. coli, leading to production of 1.3 mg/L of taxadiene (Huang et al., 2001). By balancing precursor supply with GGPP conversion to maximize taxadiene production while minimizing inhibiting effects of indole accumulation, a final level of 1.02 g/L of taxadiene was achieved (Ajikumar et al., 2010). Similar taxadiene yields of 0.87 g/L were achieved using in silico analysis of the DXP pathway in E. coli (Meng et al., 2011). Through codon optimization, modification of the N-terminal transmembrane region and the generation of chimera enzymes fused with a Taxus CYP450 reductase, the expression of a Taxus CYP450 in E. coli was recently achieved to allow for the 5α-oxidation of taxadiene to taxadiene-5α-ol (Ajikumar et al., 2010). Expression of the optimized CYP450 resulted in the production of nearly 60 mg/L taxadiene-5α-ol.

Plant cell culture

Although plant cell culture technology allows for sustainable production of secondary metabolites, research is being directed at overcoming problems associated with low product yields and cell culture variability. Dedifferentiated cell cultures (i.e. callus or suspension) often produce low levels of secondary metabolites compared to differentiated cells (i.e. shoots or roots) (Lorence and Nessler, 2004; Liu et al., 2006; Pasqua et al., 2006). Recent work in our laboratory and others have shown that secondary metabolite production can be unstable over long periods of time in suspension cultures (Ketchum and Gibson, 1996; Kim et al., 2004; Naill and Roberts, 2005b; Sanchez-Sampedro et al., 2009), which may be correlated with variations in the cellular ploidy level (Baebler et al., 2005). This variation in ploidy has been linked to differences in gene expression, with higher ploidy resulting in an increase in the number of silenced genes (Mittelsten Scheid et al., 1996). Gene silencing within a secondary metabolic pathway could significantly affect metabolite production.

Cell culture conditions can influence secondary metabolite production in plant cell cultures. For example, large variations in paclitaxel accumulation have been seen across cell cultures that were derived from the same parent culture (Ketchum and Gibson, 1996). High paclitaxel-producing cultures can exhibit growth inhibition when subcultured, which was linked to accumulation of paclitaxel and related taxanes in the media (Kim et al., 2004; Expósito et al., 2009). The tendency of plant cells to grow in aggregates can lead to the development of heterogeneous subpopulations in culture where cells differ metabolically and morphologically (Naill and Roberts, 2005a,b; Kolewe et al., 2010). Individual cells within a culture accumulate variable levels of a product. For instance, only 10% of cells in a culture of C. roseus were found to produce the secondary metabolite anthocyanin (Hall and Yeoman, 1987), and a broad distribution of paclitaxel accumulation was observed in cultured Taxus cells (Naill and Roberts, 2005b). The existence of cell subpopulations necessitates the study of suspension cultures at the single-cell level to fully understand and ultimately engineer cultures for optimal performance (Kolewe et al., 2008).

Whereas single cells are typically around 20–40 μm in diameter, larger aggregates can be several millimetres in diameter. Oxygen and nutrient diffusion limitations lead to the development of microenvironments within a larger aggregate (Hulst et al., 1989; Naill and Roberts, 2004). The average aggregate size of a culture often varies over the culture period and can affect secondary metabolite accumulation. The majority of research aimed to link aggregate size to secondary metabolite accumulation involves filtration and collection of different aggregate size classes and subsequent product extraction from biomass. No conclusive trends have been found which span all plant species (Kolewe et al., 2010). A new strategy was used to study the accumulation of secondary metabolites in cultures with different sized aggregates that enabled consideration of both cell- and media-associated accumulation (Kolewe et al., 2011). Cultures composed of small aggregates of Taxus cuspidata cells were found to produce up to 20-fold more paclitaxel after methyl jasmonate elicitation than cultures composed of larger aggregates. The mechanism for the increase in paclitaxel accumulation is currently under investigation.

Increasing yields of secondary metabolites in plant cell culture

Elicitation

Elicitation is a traditional method used to increase secondary metabolite accumulation in plant cell cultures (Bourgaud et al., 2001). Many plant secondary metabolites are linked to plant defence against pathogens and herbivores, so cellular metabolism can be shifted to favour secondary metabolism through the introduction of chemical or physical stresses (Pauwels et al., 2009). Various environmental conditions (e.g. plant disease, UV light, air quality) and chemical treatments (e.g. CuSO4, NaCl, H2SO4) have been found to affect furanocoumarin content in a variety of plant species (Bourgaud et al., 2001). Jasmonic acid and its conjugates and precursors, collectively known as jasmonates, are lipid-derived compounds involved in plant growth and development (Wasternack, 2007). They are able to induce a metabolic shift from growth to defence in many species, allowing a plant to defend itself from a hostile environment (Pauwels et al., 2009). Jasmonic acid and its methyl ester, methyl jasmonate, were first shown to increase the accumulation of secondary metabolites in suspension cultures of 36 distinct plant species (Gundlach et al., 1992), and exogenous methyl jasmonate addition increased paclitaxel production up to 46-fold over unelicited cells (Mirjalili and Linden, 1996; Yukimune et al., 1996; Ketchum et al., 1999). In Arabidopsis thaliana suspension cultures, methyl jasmonate was found to first induce genes involved in transcriptional regulation and jasmonic acid biosynthesis, followed by repression of cell cycle genes and induction of genes leading to phenylpropanoid production (Pauwels et al., 2008). Elicitation can also be used to understand the regulation of secondary metabolic pathways through differential gene expression studies of cultures that accumulate low/no versus high levels of secondary metabolites.

In situ product removal

The accumulation of secondary metabolites can be limited by both feedback inhibition of product synthesis and product degradation (Wang et al., 2001). In situ product removal from the culture media using two-phase solid–liquid or immiscible liquid–liquid systems can overcome these limitations and increase culture productivity, while simplifying product recovery and allowing for semicontinuous operation (Yan et al., 2005). Product adsorption using a nonaromatic resin led to a significant increase in the production of trans-resveratrol and isotopically labelled trans-resveratrol in Vitis vinifera after elicitation with either salicylic or jasmonic acid (Yue et al., 2011). Product accumulation increased mainly as a result of introduction of the adsorbent, rather than the elicitor, which suggested that the rate-limiting step occurred after biosynthesis of trans-resveratrol. The use of a hydrophobic macroporous resin in the culture of Salvia miltiorrhiza hairy root cultures along with semicontinuous operation resulted in tanshinone production 7.4 times higher than batch culture productivity (Yan et al., 2005). Production levels were only increased by adsorbent addition when the concentration of product in the media was high, suggesting that production was limited by feedback inhibition at high concentrations. Two-phase systems can also help to simplify purification of products from the media. For example, a sixfold increase in paclitaxel production was achieved in Taxus chinensis suspension cultures in aqueous-organic two-phase systems with sucrose feeding, with 63% of the product released to the media (Wang et al., 2001).

Immobilization

Immobilization of plant cells can be used to overcome some of the problems associated with suspension cultures, such as low and variable product yields, high susceptibility to shear and slow growth rates, because of the potential for higher cell densities and continuous product removal (Brodelius, 1985; Dornenburg and Knorr, 1995). Cells have been successfully encapsulated in calcium alginate, agar, agarose, gelatin, carrageenan, polyacrylamide, polyurethane foam and hollow fibre membranes, with the most popular technique being encapsulation in an alginate gel (Smetanska, 2008). Immobilization of Taxus baccata cells in calcium alginate beads in a bioreactor resulted in total (i.e. cell-associated and extracellular) paclitaxel yields of 43.43 mg/L after 16 days, representing a production rate of 2.71 mg/L/day (Bentebibel et al., 2005). These were the highest paclitaxel levels reported to date achieved in a laboratory-scaled bioreactor. Elicitation of these cells led to an increase in the production of paclitaxel and its precursor, baccatin III, but the compounds were maintained within the alginate beads (Bonfill et al., 2007), which complicates purification. On the other hand, immobilization of Solanum chrysotrichum cells in a similar matrix resulted in an overall decrease in spirostanol saponin; however, there was an increase in metabolite secretion to the extracellular medium (Charlet et al., 2000), simplifying purification. In systems that secrete the secondary metabolite of interest, encapsulation or immobilization can simplify downstream processing and allow for continuous product removal.

Metabolic engineering

Both transient and stable transformation procedures can be used to study the effect of the up-regulation or down-regulation of biosynthetic pathway genes or transcription factors on metabolite profiles or gene expression. Transient transformation techniques such as particle bombardment, tissue electroporation, DNA injection, agroinfiltration and viral expression systems lead to higher gene expression levels and transformation efficiencies in nearly all plant species over stable transformations but only allow for short-term expression (Lessard et al., 2002; Komarova et al., 2010). Stable transformation techniques, such as Agrobacterium-mediated transformation, are more time-consuming than transient techniques, but because genes are incorporated into the plant genome, expression can be maintained over time (Newell, 2000). Although this transformation technique is ideal for metabolic engineering of plant cell cultures, there are many industrially important species, such as soy beans, cereal grains and tree species, that are recalcitrant to Agrobacterium-mediated transformation (Gelvin, 2003b). As an alternative, stable gene expression can be achieved through direct gene transfer techniques such as particle bombardment, silicon carbide fibres and electroporation. While transient gene expression is achieved when DNA coated particles are transferred into the cell, stable gene expression requires integration of the DNA into the genome, which is a much less frequent event (Altpeter et al., 2005). As a result, a selection marker, such as an antibiotic resistance, must be transformed into the cells to allow for the isolation and scale-up of transformed cells for multiple months post-transformation. Once the transgenic line has been established, transgene expression, which can be affected by the site of integration into the genome, gene silencing and the promoter utilized, must be studied to identify highly expressing cell lines that maintain expression over time (Page and Minocha, 2004). Efficient transformation protocols must first be established for a plant species to enable metabolic engineering studies. Both transient and stable transformation techniques, including particle bombardment and Agrobacterium-mediated transformation, have been developed for the Taxus (Ketchum et al., 2007; Vongpaseuth et al., 2007) and Catharanthus species (Goddijn et al., 1995; van der Fits and Memelink, 1997; Guirimand et al., 2009).

One major limitation to the metabolic engineering of plant cells for increased secondary metabolite accumulation is the lack of fully defined metabolic pathways. The rate-influencing genes within a metabolic pathway must be identified to enable targeted metabolic engineering approaches. Several techniques are commonly used for the identification and characterization of unknown metabolic pathway genes, including precursor feeding, gene overexpression and inhibition, mutant selection or differential gene expression studies using elicitation. Many of the steps of the paclitaxel biosynthetic pathway have been elucidated using these techniques. Precursor feeding led to identification of the first committed step of the paclitaxel biosynthetic pathway, the cyclization of GGPP to taxadiene by TASY (Koepp et al., 1995). Using degenerate primers based on the sequences of similar enzymes in other species, a hybridization probe was identified and used to screen a cDNA library to obtain the full sequence of TASY (Wildung and Croteau, 1996). The identified gene sequence was expressed in E. coli and found to be catalytically active on GGPP. The use of differential gene expression studies on elicited and unelicited cells allowed for discovery and characterization of many of the genes involved in the 19 putative steps of paclitaxel biosynthesis, as reviewed in Croteau et al. (2006); Vongpaseuth and Roberts (2007). In addition, these studies identified putative genes that could be involved in the undefined regions of the pathway. Transcript profiling of Taxus suspension cultures using RNA gel blot analysis and RT-PCR was used to identify potential rate-influencing steps in paclitaxel biosynthesis (Nims et al., 2006). Recently, 454 pyrosequencing was used to create an expressed sequence tag library from T. cuspidata needles (Wu et al., 2011). This represents the largest EST collection for Taxus to date and could be used to further elucidate the metabolic pathway and identify potential transcription factors involved in the regulation of taxane production.

Transcription factors are regulatory proteins that have the ability to control multiple genes within a metabolic pathway through protein–protein interactions or the binding of specific DNA sequences (Broun, 2004). They have been considered a viable alternative to single-gene metabolic engineering because they can be used to redirect metabolism by regulating the expression of multiple genes within a biosynthetic pathway (Gantet and Memelink, 2002; Grotewold, 2008). In addition, transcription factor engineering does not require complete knowledge of the biosynthetic pathway of interest, rendering this approach particularly useful for complex secondary metabolites with undefined biosynthetic pathways. The use of transcription factors for plant metabolic engineering has been thoroughly reviewed (Gantet and Memelink, 2002; Broun, 2004; Grotewold, 2008; Iwase et al., 2009). Up-regulation of both G10H, a gene responsible for the production of the terpenoid moiety, and Orca3, a transcription factor involved in terpene indole alkaloid (TIA) biosynthesis (van der Fits and Memelink, 2001), in C. roseus hairy root cultures resulted in a 6.5-fold increase in production of the TIA catharanthine (Wang et al., 2010). The engineering of both transcription factors and rate-influencing genes within the TIA biosynthetic pathway could lead to a significant increase in TIA accumulation in C. roseus (Liu et al., 2007). By affecting the expression of multiple genes within a metabolic pathway, transcription factor expression can often overcome overall pathway flux limitations associated with single-gene metabolic engineering (Martin, 1996).

Plants as a production system for heterologous proteins

Biopharmaceuticals are proteins, antibodies or nucleic acid-based products used in the treatment of disease (Walsh, 2005). There are over 200 biopharmaceuticals on the market, 58 of which have been approved in the last 4 years. Of the 58 products recently approved, 55% are produced using mammalian cell culture, 29% using E. coli and the remaining 16% using yeast (S. cerevisiae and Pichia pastoris) and nontraditional systems, such as transgenic animals or baculovirus-insect cells (Walsh, 2010). Bacterial systems have very short production times, and yields are typically 0.5–0.8 g/L; higher yields on the order of multiple grams per litre have been reported with additional engineering and process control (Joly et al., 1998; Choi et al., 2000; Chen et al., 2004). This system is limited by contamination by bacterial toxins and inability to support post-translational modifications, such as protein folding and glycosylation, which can lead to protein aggregation or loss of function (Gomord and Faye, 2004). Yeast systems are able to overcome some problems associated with post-translational modifications and offer relatively high yields between 0.1 and 1.0 g/L (James and Lee, 2001), but proteins can be hyperglycosylated, which can influence protein folding and function (Hamilton and Gerngross, 2007). Mammalian cells are capable of proper post-translational modifications and glycosylation and offer the highest production levels, between 1 and 3 g/L (Boehm, 2007). On the other hand, commercialization of mammalian cell culture is costly and complicated, owing to expensive media components, difficult handling and the potential for viral and oncogenic contamination (Boehm, 2007; Yin et al., 2007).

Heterologous proteins were first introduced into plants just over two decades ago (Hiatt et al., 1989; During et al., 1990; Sijmons et al., 1990). Since this time, the use of whole plants and plant cell culture as production systems has been widely explored. Plants are able to avoid some of the problems associated with mammalian and microbe-based production systems. For instance, plant systems do not produce endotoxins, pathogens or oncogenic DNA and are able to fold and assemble complex proteins (Commandeur et al., 2003; Twyman et al., 2003). This was demonstrated in tobacco with the expression of the four protein monomers of secretory immunoglobulin A which were assembled into the functional, high-molecular weight complex (Ma et al., 1995). While heterologous plant-produced proteins and native human proteins have similar post-translational modifications, some differences exist. Engineering strategies have been applied to further humanize the glycosylation patterns, enabling the production of fully functional proteins in plants (Gomord and Faye, 2004; Schähs et al., 2007; Gomord et al., 2010). Although whole plant farming of pharmaceutical proteins has been utilized, there are benefits to the use of plant cell culture as a production platform. The US Department of Agriculture’s concern about transgene migration and contamination of the human food chain has been a major limitation to the use of field-grown crops, especially when food crops are the system of choice (Fox, 2006; Murphy, 2007). Plant cell culture allows for the isolation of genetically modified plant cells, reducing the risk of contaminating food sources. They can also have more consistent yields under controlled conditions, eliminating the threat of crop destruction and inconsistency owing to unpredictable environmental conditions or pathogen contamination (James and Lee, 2001). For the recovery of foreign protein from field-grown corn, downstream processing costs represent 94% of the operating costs annually, with only 83% product purity (Evangelista et al., 1998). Although the material and operating costs for aseptic plant cell culture that meets cGMP demands are significantly higher than those for agricultural production, the secretion of proteins in vitro has the potential to significantly decrease downstream processing costs (Shih and Doran, 2009). Additionally, the use of bioreactor systems allows for the use of expression systems that utilize chemically inducible promoters or viral vectors that are limited in field-grown crop applications. The use of plant cell culture for the production of heterologous proteins is on the rise, with multiple biopharmaceuticals in preclinical and clinical development (Table 1). To increase the commercial feasibility of plant cell culture production, research is being conducted to increase protein yields, which typically vary from 0.005 to 200 mg/L, through the development of optimized expression systems and enhancement of protein stability after production (Hellwig et al., 2004).

Increasing yields of heterologous proteins in plant cell culture

Gene transformation strategies

Heterologous genes can be introduced into plants both transiently and stably. For transient transformations, gene expression is typically maintained for 6–14 days, allowing for screening of expression systems, small-scale production for protein characterization and large-scale manufacturing (Vaquero et al., 1999; Huang and McDonald, 2009; Komarova et al., 2010). Transient gene expression can be accomplished in plant cell culture through the use of Agrobacterium infiltration or viral vectors (Fischer et al., 1999; Gleba et al., 2007). This procedure produces protein yields that are comparable to those found in the Agrobacterium infiltration of intact plant tissue and is a better platform for scale-up under aseptic conditions, where the culturing conditions can be more easily controlled (Andrews and Curtis, 2005). Additionally, the Agrobacterium strain containing the viral expression system can be more easily contained in vitro compared to whole plants grown in an open field (Shih and Doran, 2009). It is hypothesized that transient transformations allow for high levels of gene expression prior to the onset of post-translational gene suppression (PTGS). As a result, some transient transformations have resulted in higher protein yields than stable transformations (Wroblewski et al., 2005). Agrobacterium infiltration was used to demonstrate transient expression of an antibody directed against the carcinoembryonic antigen before stably introducing the production system into Nicotiana tabacum. Expression levels in the transiently transformed cells were approximately five times higher than those found in the stably transformed plant (Vaquero et al., 1999, 2002). Stable transformations can be accomplished through the use of Agrobacterium-mediated transformation (Gelvin, 2003a), particle bombardment (Christou, 1992) or electroporation (Joersbo and Brunstedt, 1991). Because these stably transformed cells take months to years to establish, protein expression studies often begin with transient expression. Once an expression system is established, the genes of interest can be integrated into the cells using a stable transformation technique to allow for expression over longer culture periods (Vaquero et al., 1999). Stably transformed cell lines are beneficial for larger-scale production of proteins but can be prone to genetic instability, resulting in the loss of transgene expression (Gao et al., 1991). As a result, seed stocks can be established using cryopreservation to limit production variability among batches (Van Eck and Keen, 2009).

Expression systems

Both constitutive and inducible promoter systems can be utilized to regulate gene expression for heterologous protein synthesis in plant cell culture systems. Constitutive promoters allow for gene expression throughout the entire growth phase of the cell culture. The cauliflower mosaic virus 35S (CaMV 35S) promoter is one of the most commonly used because it allows for high expression levels; however, viral promoters can result in transgene-induced gene silencing (Odell et al., 1985; Vaucheret et al., 1998). To minimize this potential problem, plant-derived promoters, such as ubiquitin and actin, have been developed (Lessard et al., 2002). If the protein product affects cell growth or viability, an inducible promoter can be used to create an expression platform where genes are expressed in response to specific and timely chemical, metabolic or physical stimuli (Huang and McDonald, 2009). For plant cell culture, chemically induced promoters are commonly used because inducers such as tetracycline, alcohol or sugars can be added directly to the bioreactor (Fleißner and Dersch, 2010).

Viral vectors offer a platform for the high expression of heterologous proteins using transient agroinfiltration. In addition to expressing wild-type genes, viral vectors express the gene of interest under the control of a strong viral promoter (Gleba et al., 2007). A summary of the viral vector expression systems utilized in plants can be found in Shih and Doran (2009). Through the cocultivation of Nicotiana glutinosa suspension cultures with Agrobacterium, a transient viral vector expression system was successfully scaled up to a 51-L stirred tank bioreactor with no reduction in protein production levels (M. O’Neill et al., 2008). A cucumber mosaic virus inducible viral amplicon (CMViva) was developed that, upon induction by estradiol, amplifies production of the CMV coat protein, allowing for high levels of gene expression at specific cell growth conditions (Sudarshana et al., 2006). Heterologous protein production using this system in the plant tissue of Nicotiana benthamiana was 30 times higher than production using the CaMV 35S promoter, with a 170-fold increase in functional protein levels. This system also led to a fourfold increase in protein production levels in N. benthamiana suspension cultures, with a twofold increase in functional protein (Huang et al., 2009). Scale-up of the transgenic N. benthamiana cultures to a 2-L semicontinuous bioreactor resulted in a 25-fold increase in protein production over batch systems (Huang et al., 2010). The protease activity and phenolic concentration in the extracellular medium were lower than in batch cultures and the system allowed for sustained production levels and steady-state bioreactor operation. Co-expression of viral gene silencing suppressors in the bioreactor system could decrease PTGS and allow for even higher productivity levels (Mallory et al., 2002; Voinnet et al., 2003).

Subcellular targeting and secretion

Proteolytic degradation within plant cells is believed to be an important factor leading to low heterologous protein yields (Benchabane et al., 2008). To minimize degradation, signalling tags are used to target protein production to specific regions within the cell (Fischer et al., 2004). For more information on localization and targeting of proteins in plants, see Mackenzie (2005). Subcellular targeting can enhance protein stability owing to differences in the redox potential, pH, and the presence or absence of chaperones, glycosyltransferases and proteolytic enzymes in different organelles (Doran, 2006a; Benchabane et al., 2008). The endoplasmic reticulum (ER) is often targeted because it is an oxidizing environment that promotes the formation of disulphide bridges. In addition, there is a high concentration of chaperones in the ER, which aid in complex protein folding, and a low concentration of proteases (Fischer et al., 2004), which result in decreased protein degradation. Storage vacuoles also have lower levels of proteases (Neuhaus and Rogers, 1998). The presence of enzyme activity in the storage vacuole that leads to exposure of terminal mannose residues on complex glycans (Lerouge et al., 1998) was exploited for the production of a functional recombinant human glucocerebrosidase in carrot suspension cultures, as discussed in Success Stories: Protalix Biotherapeutics (Shaaltiel et al., 2007).

One benefit of plant cell culture over whole plant systems is that heterologous proteins can be secreted into the cell culture medium to simplify purification. For secretion, a signal peptide is used to target unfolded peptides to the ER for folding and assembly. After folding, the protein is transported from the ER through the secretory pathway, enters the Golgi apparatus and is secreted from the cell (Kermode, 1996). Some of the benefits to protein secretion are: (i) no need to destroy the cell, which results in complicated purification strategies owing to release of intracellular compounds to the media; (ii) cells can be easily filtered from the liquid media; (iii) plant cells do not secrete many proteins, so purification is simplified because there are few contaminating proteins; and (iv) plant cells can be reused or maintained within the bioreactor, allowing for additional production cycles (Fischer et al., 2004; Huang and McDonald, 2009). A major drawback to secretion is that proteins can be degraded by proteolytic enzymes in the culture media (Doran, 2006a). In N. tabacum suspension cultures, functional heterologous protein was identified in samples containing both cells and media, but not in media-only samples (Schiermeyer et al., 2005). Endogenous proteases secreted to the medium degraded the protein, and addition of a protease inhibitor or EDTA to the media was found to reduce degradation. The conversion of callus cultures to suspension cultures resulted in the rapid loss of hepatitis B surface antigen expression in Glycine max (Ganapathi et al., 2007). By adding protease inhibitors to the culture media, product recovery was restored by 50% to 12.1 ng/g of fresh weight.

Increasing secreted protein stability

Secreted proteins are also not always stable in the media environment (Tsoi and Doran, 2002). To prevent degradation or protein unfolding, stabilizing agents such as dimethyl sulfoxide, polyethylene glycol, polyvinylpyrrolidone (PVP), gelatin, bovine serum albumin or protease inhibitors can be added to the culture media (Franconi et al., 2010). In transgenic N. tabacum suspension cultures, functional heterologous protein production was found to correlate with cell growth until stationary phase, where functional protein levels decreased significantly (Becerra-Arteaga et al., 2006). Total extracellular protein in these samples was found to either increase or remain constant over time, suggesting that loss of function was owing to denaturation rather than proteolytic degradation. The addition of two known protein stabilizers, PVP and bacitracin, led to a reduction in the loss of activity in stationary phase (Becerra-Arteaga et al., 2006). Surface adsorption of secreted proteins can decrease protein levels in the media and be mitigated by coating glass vessels with a protein-resistant polymer (Doran, 2006b). Effective monitoring and adjusting of culture conditions can lead to increased yields of secreted proteins. An increase in the media pH of N. benthamiana suspension cultures resulted in a fourfold increase in recombinant protein production, with a twofold increase in the ratio of functional to total protein (Huang et al., 2009). At higher pH, protein stability was increased through a decrease in proteolytic activity, resulting in maximal protein recovery.

In situ protein removal has also been effective at increasing functional protein recovery. Aqueous two-phase systems (ATPS) are produced by mixing two water-soluble polymers in water. Nutrients, metabolites, proteins, cell particles and whole cells can be effectively partitioned between the two phases based on size, protein confirmation (folded vs. denatured) and the charge and ionic composition of the molecule (Albertsson, 1970; Hooker and Lee, 1990). Extractive fermentation using ATPS has been used successfully for a diverse range of compounds, from low-molecular weight products to high-molecular weight enzymes (Banik et al., 2003). The use of ATPS with plant cell culture was first demonstrated using suspension cultures of N. tabacum in a 3% PEG and 5% crude dextran system (Hooker and Lee, 1990). Culture growth rates and cell densities were comparable to those measured in standard suspension cultures. Enzyme production of N. tabacum in ATPS cultures was explored using a 4% polyethylene glycol and 7.5% dextran system (Ilieva et al., 1996). Despite the reduced growth rate observed, the extracellular acid and alkaline phosphomonoesterase yields were increased by 18- and 10-fold, respectively, although the overall enzyme yield was comparable to the control system.

Protein-binding resins can also be used to remove proteins in situ. A hydroxyapatite resin leads to a 20%–21% increase in the recovery of functional monoclonal antibody from hairy root and suspension cultures of N. tabacum (Sharp and Doran, 2001). Protein production was increased through more frequent harvestation of the resin from the media. An affinity chromatography bioreactor circulates media from a bioreactor through a column to remove secreted proteins during cell growth. This system was used with protein G and iminodiacetic acid metal affinity resin for suspension cultures of N. tabacum (James et al., 2002). Production of the heavy chain of a mouse monoclonal antibody was increased by eightfold, while a twofold increase was observed in the production of a HIS-tagged human granulocyte macrophage colony-stimulating factor. These increases were attributed to the removal of proteins from the degrading media and prevention of product inhibition pathways.

Industrial production using plant cell culture

For plant cell culture, the transition from shake flask to bioreactor is complicated. As a result, a wide variety of bioprocessing strategies have been specifically designed for large-scale cultivation of plant cells. Bioreactors for plant suspension cultures (Eibl and Eibl, 2008; Georgiev et al., 2009; Huang and McDonald, 2009) and hairy root cultures (Kim et al., 2002b; Georgiev et al., 2007; Srivastava and Srivastava, 2007) have been recently reviewed. This section will focus on technologies used for commercial bioprocessing of plant cell culture.

Bioreactor design and operation

Plant cells (20–50 μm in diameter and 100–500 μm in length) are significantly larger than bacterial (<1 μm in diameter), fungal (5–10 μm in diameter and <100 μm in length) and mammalian cells (10–100 μm in diameter), with intracellular vacuoles that account for up to 90% of cell volume (Huang and McDonald, 2009). As a result, plant cell suspensions are subject to shear sensitivity, which limits the mechanical agitation techniques available to meet oxygen demands for cell growth. Oxygen transport in plant cell bioreactors has been modelled, accounting for the transfer of oxygen through the gas, liquid and solid phases (Curtis and Tuerk, 2006). Although cell culture medium typically behaves as a Newtonian fluid, excretion of extracellular polysaccharides can significantly increase medium viscosity (Georgiev et al., 2009). Culture broth rheology can also be affected by aggregate size distribution, cell morphology, biomass concentration and culture conditions (Huang and McDonald, 2009). Changes in these culture properties can lead to diffusion and mixing problems, so reactor conditions (e.g. agitation, aeration, temperature, pH, etc.) must be optimized to maximize cell growth. Azadirachta indica was grown in a 3-L stirred tank bioreactor with both a setric and centrifugal impeller (Prakash and Srivastava, 2007). Owing to a decrease in shear and increase in oxygen transfer, biomass accumulation (15.5 vs. 18.7 g dry weight per L) and azadirachtin production (0.05 vs. 0.071 g/L) were increased when using the centrifugal impeller. Response surface methodology was used to optimize elicitor treatment of Azadirachta indica cells, and azadirachtin yields were increased to 0.16 g/L after treatment with a combination of salicylic acid, jasmonic acid and chitosan (Prakash and Srivastava, 2008). A variety of bioreactors have been designed specifically for the culture of plant cell suspension cultures and are summarized in Table 3.

Table 3.   Bioreactors designed for plant cell suspension cultures (Huang and McDonald, 2009)
Bioreactor typesFeatures/advantagesDisadvantages
Stirred tank bioreactorCommonly used for all cell types
Easy to scale up
Useful for high-viscosity cell culture
Able to achieve high oxygen transfer
Good fluid mixing
Alternative impellers available
Ease of compliance with cGMP requirements
High shear stress around the impeller
High capital and operational costs
Heat generated from mechanical mixing
High energy cost owing to mechanical agitation
Contamination risk with mechanical seal
Pneumatic bioreactor: bubble columnSuitable for plant and animal cells
Easy to construct and scale up
Low operational cost
Low contamination risk
Low shear stress
No heat generation owing to lack of mechanical agitation
Poor oxygen transfer capabilities
Poor fluid mixing in highly viscous cultures
High levels of foaming under high-aeration conditions
Pneumatic bioreactor: air-lift and modified air-liftSuitable for plant and animal cells
Easy to construct and scale up
Low operational cost
Low contamination risk
Low shear stress
No heat generation due to lack of mechanical agitation
Multiple choices of internal draft tubes
Better oxygen transfer than bubble column
Circulating flow patterns to aid in gas and nutrient transfer
Relatively poor oxygen transfer capabilities
Poor fluid mixing for highly viscous cultures
High levels of foaming under high-aeration conditions
Membrane bioreactorDisposable equipment
Ability to concentrate biomass and protein product in membrane compartment
Easy to withdraw extracellular product
Low shear stress
Low operational cost
Simplified media exchange
Difficult to scale up
Oxygenation required
Low heat transfer rate
Difficult for online monitoring of culture conditions
Increased cost owing to disposability
Wave bioreactorDisposable equipment
Low shear stress
High oxygen mass transfer
Useful for high cell density culture
Low operational cost
Reduce cleaning, in-house sterilization requirements
Increased operational flexibility—batch, semi-batch, perfusion configurations
Light weight
Difficult to scale up
Difficult to apply advanced cell culture operational strategies
Low heat transfer rate

The use of hairy root cultures in bioreactors adds additional complexities for bioprocess design. The branching of hairy roots creates a tight matrix within a culture, leading to nutrient transport limitations that result in areas of senescent cells (Srivastava and Srivastava, 2007). These nutrient gradients also attribute to variability in root growth and productivity (Eibl and Eibl, 2008). Hairy roots are sensitive to shear stress, and mechanical agitation can result in root wounding (Srivastava and Srivastava, 2007). Three types of bioreactors are typically used for the cultivation of hairy root cultures: (i) liquid-phase reactors, where roots are submerged in liquid media; (ii) gas-phase reactors, where media are either sprayed onto the roots or delivered as a mist; and (iii) hybrid reactors (Kim et al., 2002a). Hybrid reactors combine the properties of both liquid- and gas-phase reactors, allowing an initial growth phase to occur in a liquid reactor before transitioning to a gas-phase system (Srivastava and Srivastava, 2007). The reactor type can significantly affect culture growth and product accumulation. Growth and productivity of a mammalian immunomodulator, interleukin 12, in N. tabacum hairy roots were compared in shake flask, air-lift and mist bioreactors (Liu et al., 2009). The highest root quality was found in the shake flask and mist bioreactors. Sections of dark roots were found in the air-lift reactor, which suggests that roots were nutrient starved. The highest protein levels were found in the mist bioreactor at 5.3 μg/g DW, which was 49.5% higher than that produced in the air-lift bioreactor. Arachis hypogaea hairy root biomass production was examined in different scales of mist bioreactors utilizing disposable culture bags (Sivakumar et al., 2010). Production levels of 12.75 g dry weight per litre were comparable to those in shake flask cultures (11.10 g dry weight per litre), but scale-up to a 20-L reactor led to a decrease in biomass accumulation to 7.77 g dry weight per litre. The main factors affecting biomass accumulation upon scale-up were flow rate to the nozzle, timing of flow rate changes as biomass increased and an increase in the frequency of misting.

The use of disposable bioreactors for plant cell culture is on the rise and has been successfully implemented by Protalix Biotherapeutics with their ProCellEx™ production platform (see Success Stories: Protalix Biotherapeutics). Benefits include high flexibility, ease of handling, reduction in cross-contamination and savings in both time and cost (Lim and Sinclair, 2007; Mauter, 2009), which can be attributed to the presterility of the container in which the cells are cultured (Eibl et al., 2010). Many different classes of disposable bioreactors have been designed for plant cell culture, including wave-mixed, stirred and bubble columns, as reviewed in Eibl et al. (2009). Suspension cultures of transgenic N. tabacum produced three times more human anti-rabies monoclonal antibody than the whole plant system (Girard et al., 2006). Biomass and protein production were comparable to shake flask production levels when scaled up to 10 L in a plastic bag disposable bioreactor, demonstrating that these inexpensive, disposable reactors are suitable for promoting both biomass and heterologous protein synthesis. The presterility of disposable bioreactors enables compliance with strict good manufacturing practice standards associated with bioreactor systems, and therefore these systems have the potential to reduce manufacturing costs.

Bioreactor design depends on whether product formation is growth- or non-growth-associated and where the product is stored, either within the cell or secreted extracellularly. If a metabolite is produced during exponential growth phase, one reactor is usually suitable for growth and product recovery. When the product is synthesized after cell growth, one reactor is often used during exponential growth to increase cell number, while another reactor is used for metabolite production. For products maintained within the cell, the reactor is usually run in batch mode so that the cells can be permeabilized to release the product after the run is completed. If the product is secreted to the media, a continuous reactor can be used for longer processing times and product can be removed as it is synthesized (Bourgaud et al., 2001). A perfusion reactor was used to culture transgenic Anchusa officinales suspension cultures producing secreted acid phosphatase over a 4-week period, with yields reaching 490 units/L/day, as opposed to 100–150 units/L/day for batch cultures (Su and Arias, 2003). Batch or semi-batch production systems are used for the commercialized products highlighted in this review, but coupling protein or secondary metabolite secretion with semicontinuous or continuous reactors could decrease the cost of both reactor operations and downstream processing.

Product recovery

Secondary metabolite recovery and purification

Many secondary metabolites are hydrophilic, so they are primarily stored in the cell vacuole; however, hydrophobic molecules are often stored in the cell membrane, vesicles, dead cells and extracellular sites, such as the cell wall (Guern et al., 1987). Paclitaxel accumulates within the cell wall (Choi et al., 2001), with as little as 7%–10% of the product released to the extracellular medium (Wickremesinhe and Arteca, 1994; Pestchanker et al., 1996). The use of cell wall-digesting enzymes resulted in the recovery of up to 90% of total paclitaxel in the extracellular medium (Roberts et al., 2003). Cell walls can be ruptured using homogenization, sonication, cell wall-digesting enzymes or steam explosion, but all biomass is destroyed. To maintain cell viability and allow for further use of biomass, cells can be permeabilized using pH shock (Thimmaraju et al., 2003) or chemical treatments (Brodelius, 1988; Park and Martinez, 1992). If the product is secreted to the medium, cell rupture can be avoided and biomass can be separated from the medium prior to downstream processing. Promoting secretion by utilizing ATPS, media exchange and resin adsorbents is beneficial for product purification because unwanted by-products are minimized.

Adsorption, precipitation, distillation, membrane separation and extraction are examples of different techniques used for initial recovery of secondary metabolites from plant cell culture. There are complications associated with each separation technique, such as protein fouling on membranes and low selectivity of extraction procedures owing to complex media compositions (Schügerl, 2005). Extraction is the most common technique used for this first stage of product purification. The initial extraction of the product from biomass is typically achieved through liquid–liquid extraction or ATPS (Romanik et al., 2007). The extraction solvent must be chosen carefully based on the physicochemical properties of the product of interest (Georgiev et al., 2009). Solvent properties such as pH and polarity affect product stability and separation efficiency. In addition, interactions between the solvent and molecule of interest sometimes lead to undesired structural changes in the product (Maltese et al., 2009).

Following initial recovery, chromatography is typically performed, and although this technique can be expensive and complicated, it is rarely avoided (Georgiev et al., 2009). Countercurrent chromatography is commonly used for preparative chromatography for all scales of product purification and has been recently reviewed in Pauli et al. (2008). Because the liquid and stationary phases compose a biphasic solvent mixture, there is no irreversible adsorption onto either of the phases. The composition of the phases is nearly unlimited, allowing for high product selectivities. The stationary phase has a loading capacity that is only limited by the solubility of the product, and the overall solvent consumption is lower than solid-phase chromatography (Georgiev et al., 2009). High-speed countercurrent chromatography is used for the preparative separation of both natural and synthetic products and can purify multiple grams of product in several hours (Ito, 2005). For the purification of some compounds, such as anthocyanins, isoflavones and flavanols, countercurrent chromatography can be followed by high-performance liquid chromatography to increase product purity (Valls et al., 2009).

Protein recovery and purification

Downstream processing can account for 80% of total production costs for heterologous proteins expressed in plant cell systems (Abranches et al., 2005). As a result, the protein extraction and purification techniques must be optimized to maximize protein recovery while minimizing processing expenses, as recently reviewed (Wilken and Nikolov, 2011). During purification, there are benefits and disadvantages to secreted verses intracellular proteins. Secreted proteins require no cell disruption, which minimizes the presence of contaminating proteins, secondary metabolites and cell debris. On the other hand, the product is dilute and prone to degradation and instability in the bulk cell culture medium. The first step in downstream processing is removal of bulk cell mass from the medium, typically using a decanter, disc-stack separator or semicontinuous or continuous centrifuge (Hellwig et al., 2004). Intracellular proteins reduce the volume of material being processed downstream, resulting in higher product concentrations; however, the feedstock is more complex, and there is a higher concentration of proteases and oxidizing substances. For these proteins, the first processing step is to isolate the cell biomass and disrupt cells using wet milling, sonication, pressure homogenization or enzymatic treatment (Xu et al., 2011). This process is nearly identical to the extraction required for the whole plant (Hellwig et al., 2004). The optimal purification process for intracellular proteins is dependent upon the localization of the protein, as demonstrated for protein targeted to the apoplast, plasma membrane and ER in tobacco leaf tissue (Hassan et al., 2008).

After the initial step, downstream processing is the same for both intracellular and extracellular proteins, beginning with clarification, which is typically accomplished using dead-end or cross-flow filtration (Hellwig et al., 2004). As an alternative, expanded bed adsorption (EBA) can be used to concentrate the protein of interest without having to perform upstream clarification (Bai and Glatz, 2003). EBA was used to achieve 60% recovery, with 90% purity of a recombinant protein from N. tabacum, but clarification could not be avoided owing to an interaction between the particulates and adsorbent (Valdes et al., 2003). In contrast, EBA was successfully able to recover 72% of a 34-fold purified recombinant protein from canola with no need for clarification. The expanded bed diethylaminoethyl (DEAE) resin was found to have similar binding and elution properties to the packed bed DEAE resin (Bai and Glatz, 2003). ATPS can be used during this primary protein recovery step to enrich the protein content while removing cell debris and other contaminates (Aguilar and Rito-Palomares, 2010). An ATPS was used to purify a recombinant protein from alfalfa, with 88% of the protein recovered in the top phase and 94% of the contaminant proteins at the interface or in the bottom phase (Ibarra-Herrera et al., 2011).

ATPS, along with chromatography and membrane filtration, are often used for protein purification. When compared to ion-exchange chromatography, ATPS led to a 37% decrease in purification costs, with 97% of a penicillin acylase recovered in a PEG-rich phase (Aguilar et al., 2006). Other chromatographic methods include affinity chromatography, hydrophobic interaction chromatography and size-exclusion chromatography (Xu et al., 2011). Utilizing four different immobilized affinity chromatography resins (histamine, tryptamine, tyramine and phenylamine), two therapeutic antibodies and a protease were successfully purified from both tobacco and maize extracts, showing that one binding resin is capable of isolating multiple proteins (Platis and Labrou, 2008). For all resins, the maize extract was purified more than the tobacco, possibly owing to the presence of phenolic substances binding to the column matrix. Affinity tags, such as glutathione S-transferases, maltose-binding proteins and polyhistidine, can be used to simplify chromatographic methods and increase product purity (Fischer et al., 2004). Chloroplast production of a HIS-tagged protein in tobacco allowed for one-step chromatography with >85% recovery (Leelavathi et al., 2003). By co-expressing an antibody with a fusion protein containing both an antibody-binding site and a cellulose-binding domain, the product antibody was able to be captured on cellulose beads (Hussack et al., 2010). Protein A, Protein G or Protein L chromatography can also be used for recovery of plant-derived antibodies, but these resins can be expensive, and there is a risk of leaking of the proteinaceous ligands, which can elicit an immune response in humans (Terman and Bertram, 1985; Platis and Labrou, 2006, 2008). Finally, membrane systems such as ultrafiltration, microfiltration, pervaporation and pertraction have also been considered for purification of plant-derived protein products (Xu et al., 2011).

Success stories

Phyton Biotech, Inc.

To explore the medicinal properties of natural products, the U.S. Department of Agriculture led the collection and screening of plants for the discovery and development of plant-derived anticancer agents between 1960 and 1982 (Cragg et al., 1996). Approximately 35 000 plants were tested throughout the United States and Mexico, and a number of anticancer compounds, such as paclitaxel and camptothecin, were discovered (Cragg et al., 1993). The extract of the bark of the Pacific yew tree, Taxus brevifolia, was first isolated in 1963 and showed anticancer activity (Mountford, 2010). In 1964, the active ingredient was isolated, and the structure of Taxol® (Bristol-Myers Squibb; generic name paclitaxel) was first published in 1971 (Wani et al., 1971). Further investigation into the efficacy of paclitaxel was not resumed until 1978 owing to limited solubility in water and low yield (0.014%) from the bark of the yew tree (Kingston, 2000). In 1979, Horwitz determined that paclitaxel binds to cell microtubules, promoting microtubule assembly into bundles and preventing mitosis (Manfredi et al., 1982). This discovery increased interest in the compound, and Phase I clinical trials began in 1983 (Cragg et al., 1993). Paclitaxel was approved for marketing as an anticancer agent in 1992 (Bristol-Myers Squibb), and with $1 billion dollars in commercial sales in 1998, it became the best-selling anticancer drug in history (Kingston, 2000; McChesney et al., 2007). Used in the treatment of ovarian, breast and lung cancers as well as AIDS-related Kaposi’s sarcoma, paclitaxel usage continues to increase dramatically from the initial 25 kg per year required to treat ovarian cancer in the United States (Cragg et al., 1993; Vongpaseuth et al., 2007). Demand could potentially exceed 200–300 kg per year as applications are being developed for paclitaxel in the treatment of Alzheimer’s and post-heart surgery patients (Cragg et al., 1993; Nims et al., 2006).

Initially, paclitaxel demand was met by harvesting bark of T. brevifolia. Because of the low paclitaxel content in the bark, 340 000 kg of Taxus bark or 38 000 trees were required to extract the desired 25 kg of drug per year (Cragg et al., 1993). Taxus trees are found in North America, Europe and Asia, but never in very high abundance, so availability is limited (Patel, 1998). Harvesting is also dependent upon seasonal variability and the slow growth rate of the tree, making it nearly impossible to keep up with demands (Roberts, 2007). To prevent the extinction of Taxus species and decrease processing costs, alternative paclitaxel production methods were investigated. Paclitaxel total synthesis has been accomplished, but reaction schemes are complex with low yields (Holton et al., 1994a,b; Nicolaou et al., 1994; Wender et al., 1997a,b). As a result, total synthesis is economically and environmentally unfavourable. The European yew tree, T. baccata, contains about 0.1% of two precursors (baccatin III and 10-deacetylbaccatin III), which can be synthetically converted into paclitaxel (Denis et al., 1988; Patel, 1998; Wuts, 1998). Initially, semisynthesis yields were approximately 50%, but a commercially viable semisynthesis route was patented in 1992 (Holton, 1992). This process continued to depend upon seasonal availability but did not require destruction of trees for harvesting (Heinig and Jennewein, 2009). With this process, paclitaxel synthesis became commercially viable and Taxol® went on the market in 1993 (Mountford, 2010).

Semisynthesis required 11 chemical transformations using 13 solvents and 13 organic reagents, making it costly and harmful to the environment (Vongpaseuth and Roberts, 2007). In 2002, Bristol-Myers Squibb discontinued the use of the semisynthetic production route, switching entirely to a plant cell culture fermentation process for both environmental health and safety reasons and a financial incentive for a green pharmaceutical process (Mountford, 2010). This production system, developed by Phyton Biotech, Inc., is the largest commercial application of plant cell culture, utilizing the Chinese yew (T. chinensis) cultivated in 75 000-L bioreactors (Huang and McDonald, 2009) and has recently been described in detail by Mountford (2010). Taxus cells that have been cryogenically frozen in a production cell bank are used to start each fermentation batch. These cells are placed on solid culture media to form a callus, before being transplanted into liquid medium for scale-up. Once the biomass has been accumulated, cells are fed a production medium to induce paclitaxel synthesis. Phyton Biotech, Inc. has quantified paclitaxel production after elicitation by many compounds, including jasmonic acid, methyl jasmonate, silver thiosulfate and 3,4-methylenedioxy-6-nitrocinnamic acid, an inhibitor of the IPP pathway (Bringi et al., 1995). Paclitaxel is purified from the medium using liquid–liquid extraction and chromatography, followed by crystallization of the pure product (Mountford, 2010). Whereas the semisynthetic production route decreased costs to 25% of that for natural harvest, the plant cell fermentation process reduced costs to just 20% of that for natural harvest (Mountford, 2010). Bristol-Myers Squibb was awarded a Presidential Green Chemistry Challenge Award by the U.S. Environmental Protection Agency, acknowledging the development and use of an environmentally friendly, sustainable manufacturing technique for paclitaxel (Tabata, 2006). Production of paclitaxel through this plant cell culture platform is currently sustainable, but demand for paclitaxel usage in combination with chemotherapy is continuing to rise, while additional applications are being developed in the treatment of Alzheimer’s and post-heart surgery patients. For these reasons, research efforts need to be directed towards the development of superior plant cell culture processes (e.g. using metabolic engineering or establishment of optimal cell lines) to increase culture yields and decrease production costs.

Protalix Biotherapeutics

Gaucher’s disease is the most common lysosomal disease, caused by decreased activity of the lysosomal enzyme acid β-glucosidase (glucocerebrosidase; GCD), resulting in lysosomal accumulation of glucosylceramide, the enzyme’s main substrate (Grabowski, 2008). The disease can lead to enlargement of the liver and spleen, anaemia, a decrease in blood platelets and skeletal deterioration (Grabowski and Hopkin, 2003; Jmoudiak and Futerman, 2005). Since 2005, the main treatment option for patients with severe Gaucher’s disease is enzyme therapy, but treatment can cost from $100 000 to >$200 000 per year (Barton et al., 1991; Weinreb et al., 2002, 2004; Grabowski, 2008). In May of 1994, Cerezyme® (Genzyme Corporation, Cambridge, MA), a recombinant GCD expressed in mammalian Chinese hamster ovary (CHO) cells, was approved for use in patients with Gaucher’s disease (Jmoudiak and Futerman, 2005). Proper glycosylation of GCD is required for optimal enzyme activity and targeting to macrophages. As a result, functional GCD cannot be produced by E. coli, and deglycosylated GCD from the human placenta is not active (Furbish et al., 1981; Grace and Grabowski, 1990). The GCD produced from CHO cells must be enzymatically remodelled to expose the mannose residues required for macrophage uptake, which significantly increases production costs (Friedman et al., 1999).

Protalix Biotherapeutics (Carmiel, Israel) developed a recombinant human GCD, taliglucerase alfa, produced by suspension cultures of transgenic carrot cells (Shaaltiel et al., 2007). GCD expression is controlled by the CaMV 35S promoter followed by a tomato mosaic virus omega translation enhancer element. Protein expression is targeted to the storage vacuole through the ER using a signal peptide from A. thaliana and a storage vacuole target from tobacco. Expression in carrot cells results in a functional protein that does not require enzymatic modification after production, and the glycosylation patterns are highly reproducible from batch to batch (Shaaltiel et al., 2007; Aviezer et al., 2009). The recombinant plant-derived GCD (prGCD) was found to be structurally homologous to Cerezyme® with comparable enzymatic activity and uptake in macrophages (Shaaltiel et al., 2007). In December 2009, Pfizer, Inc. (New York, NY) purchased the worldwide rights to the drug, excluding Israel, for $115 million. With this deal, Protalix Biotherapeutics maintained control of manufacturing and 40% of expenses, receiving 40% of revenues in return (Ratner, 2010). The drug completed phase III clinical trials in September 2009 and Protalix Biotherapeutics submitted a new drug application. In the meantime, both the FDA and European Medicines Agency requested supply of prGCD for select patients under an expanded access programme (Ratner, 2010). At the time of writing this review, the protein is still awaiting approval after the FDA issued a Complete Response Letter requesting additional information regarding testing specifications and assay development in February 2011. The FDA did not request data from additional clinical studies and found the Protalix Biotherapeutics manufacturing facilities acceptable (http://www.protalix.com). Two therapies are currently on the market for the treatment of Gaucher’s disease: Genzyme’s Cerezyme® and Shire’s velaglucerase alfa (Vpriv®), which was approved in 2010 (Opar, 2011).

According to the Protalix Biotherapeutics’ website, the prGCD is produced using a patented ProCellEx™ production platform (http://www.protalix.com). This platform utilizes presterilized, flexible polyethylene containers that are >400 L in volume for culturing and harvesting in consecutive cycles. Cultures are provided with oxygen, nutrients, inoculants and culture media, and excess air and waste gases are easily removed. The temperature, lighting, air and nutrients used for operation have been optimized to maximize cell productivity. This reactor system was designed specifically for plant cell culture, is rapidly scalable at a low cost, requires minimal initial capital investment and is easy to use. As a result, Protalix Biotherapeutics is currently working on the development of several other plant-produced protein therapeutics using the ProCellEx production platform, as seen in Table 1. The company is also developing lyophilized plant cells as a drug delivery vehicle for the prGCD protein. The cellulose of the cell wall protects the enzyme from degradation in the digestive tract, aiding in the delivery of the active enzyme into the blood stream.

Future outlook for plant cell culture commercialization

Several plant cell culture systems have been used for the industrial production of secondary metabolites and heterologous proteins, but a further understanding of plant cellular metabolism and the development of new optimization strategies could dramatically affect the field and enable more widespread use of the technology. A better knowledge of the inherent variability and heterogeneity associated with plant cell cultures could allow product yields to be maximized by adjusting process conditions. A new technique has been developed to determine the aggregate size distribution of plant cell suspension cultures using a simple Coulter counter (Kolewe et al., 2010). This technique can be used to rapidly analyse the aggregate size distribution of a culture over time, which has been widely shown to affect metabolite production. To fully understand cell culture variability and heterogeneity, cultures can be studied on the level of the individual cell. Typical cell culture studies focus on culture-averaged parameters (e.g. fresh weight, dry weight, culture productivity, etc.), where samples containing millions of cells are analysed together, and no information is collected regarding cell–cell differences. Flow cytometric techniques using isolated single particles (e.g. cells, protoplasts, nuclei, mitochondria and chromosomes) can be used to analyse cell cycle participation, genome size, ploidy level and metabolite accumulation of individual cells (Galbraith, 2004; Gaurav et al., 2010). For example, to study paclitaxel accumulation in single Taxus cells, an immunostaining protocol was developed using an anti-paclitaxel monoclonal antibody (Naill and Roberts, 2005b). Paclitaxel accumulation was highly variable among cells in culture, and approximately 20% of cells did not accumulate any cell-associated paclitaxel (Naill and Roberts, 2005b), even with elicitation with methyl jasmonate.

One reason for plant cell culture heterogeneity is that traditional suspension cultures are created from dedifferentiated callus cultures that originated as a mixture of specialized cell types isolated from plant tissue. These cells could remain heterogeneous over time and contribute to the variability associated with suspension cultures (Roberts and Kolewe, 2010). To avoid this dedifferentiation process, cambial meristematic cells (CMC) have been isolated from T. cuspidata, Panax ginseng, Ginkgo baloba and Solanum lycopersicum (Lee et al., 2010; Roberts and Kolewe, 2010). Taxus cuspidata CMC were found to have lower growth variability, smaller aggregate size and less sensitivity to shear in stirred tank and air-lift bioreactors than traditional Taxus suspension cultures. Additionally, paclitaxel production levels were higher than those found in traditional cultures, and process optimization could potentially create a more homogeneous culture and further increase product yields (Lee et al., 2010). Development and optimization of CMC cultures from a variety of species could expand the potential for commercial plant cell culture.

Functional genomics approaches have been used to identify key genes and transcription factors involved in the secondary metabolism of genome-sequenced and unsequenced plant species, as reviewed in Yonekura-Sakakibara and Saito (2009). Metabolite and transcript profiling of methyl jasmonate-elicited tobacco cells yielded both known and novel genes potentially involved in Nicotiana secondary metabolism, resulting in the discovery of a new nicotine transporter (Goossens et al., 2003; Morita et al., 2009). A similar study conducted on elicited C. roseus cultures identified several genes putatively involved in TIA biosynthesis (Rischer et al., 2006). These functional genomics techniques can be used to identify genes involved in the biosynthesis of secondary metabolites, allowing for single-gene metabolic engineering of plant cell cultures. They can also identify putative transcription factors involved in the regulation of secondary metabolic pathways, allowing for the metabolic engineering of cultures without fully elucidated biosynthetic pathways. Metabolic engineering for increased productivity could decrease production costs associated with existing commercial plant cell culture systems as well as make other plant cell culture systems commercially feasible.

The sales of biologics are projected to continue rising dramatically, and it is predicted that four of the five top-selling drugs will be protein therapeutics by the year 2013 (Goodman, 2009). As patents begin to expire on existing biopharmaceuticals, the production of biosimilars, or replicate versions of the original therapeutic proteins or antibodies, is on the rise. Owing to their complex structures, the development of generic biologics is far more difficult than the development of generic small-molecule pharmaceuticals. Identical replication of a therapeutic protein cannot reasonably be achieved, so biosimilars will likely have to undergo clinical trials to prove their safety and efficacy (Ledford, 2010). Despite these complications, the approval of biosimilars is on the rise, with 14 drugs being approved in Europe in 2010, and a regulatory pathway for biosimilar approval has recently been created in the United States (Walsh, 2010). Biosimilars, such as the prGCD produced by Protalix Biotherapeutics, offer an opening for plant cell culture production technologies, which could allow for decreased production costs over traditional biopharmaceutical production systems. Also, the FDA’s request for supply of Protalix Biotherapeutics’ prGCD to patients with Gaucher’s disease before approval signals the acceptance of plant cell-based production systems (Ratner, 2010). To compete with existing biologic production systems, the benefits of plant cell culture, such as safety, controlled culture conditions and protein secretion, must be promoted and exploited. In addition, functional product yields can be improved by manipulating the plant cell host to produce proteins with human-like glycosylation patterns, creating low protease cell culture lines and/or overcoming gene silencing. Optimization of viral and inducible gene expression systems and improving product secretion and stability in the culture media will help to decrease production costs associated with plant cell culture systems, making them more commercially attractive for the production of biosimilars and other biologics.

Acknowledgements

The authors would like to acknowledge the support of grants from the National Science Foundation (CBET 0730779) and the National Institutes of Health (GM070852). In addition, S.A.W. would like to acknowledge the support of the National Science Foundation-sponsored Institute for Cellular Engineering IGERT program DGE-0654128.

Ancillary