Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Human erythropoietin (EPO) is a pleiotropic cytokine with remarkable tissue-protective activities in addition to its well-established role in red blood cell production. Unfortunately, conventional mammalian cell cultures are unlikely to meet the anticipated market demands for recombinant EPO because of limited capacity and high production costs. Plant expression systems may address these limitations to enable practical, cost-effective delivery of EPO in tissue injury prevention therapeutics. In this study, we produced human EPO in tobacco and demonstrated that plant-derived EPO had tissue-protective activity. Our results indicated that targeting to the endoplasmic reticulum (ER) provided the highest accumulation levels of EPO, with a yield approaching 0.05% of total soluble protein in tobacco leaves. The codon optimization of the human EPO gene for plant expression had no clear advantage; furthermore, the human EPO signal peptide performed better than a tobacco signal peptide. In addition, we found that glycosylation was essential for the stability of plant recombinant EPO, whereas the presence of an elastin-like polypeptide fusion had a limited positive impact on the level of EPO accumulation. Confocal microscopy showed that apoplast and ER-targeted EPO were correctly localized, and N-glycan analysis demonstrated that complex plant glycans existed on apoplast-targeted EPO, but not on ER-targeted EPO. Importantly, plant-derived EPO had enhanced receptor-binding affinity and was able to protect kidney epithelial cells from cytokine-induced death in vitro. These findings demonstrate that tobacco plants may be an attractive alternative for the production of large amounts of biologically active EPO.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Human erythropoietin (EPO) is produced in the kidneys and is the principal hormone responsible for maintaining the circulating erythrocyte mass (Lin etal., 1985). Recombinant human EPO (rhEPO) is an important biopharmaceutical that is used extensively in anaemia caused by renal failure, chemotherapy and acquired immunodeficiency syndrome (AIDS). In addition, rhEPO had a market value of $12 billion in 2006, clearly a product of considerable value to the pharmaceutical industry (Tucker and Yakatan, 2008).
EPO is a 166-amino-acid glycoprotein existing as a heterogeneous mixture of glycoforms, with 40% of the 34–39-kDa molecular weight consisting of three N-linked [asparagine-24 (Asn24), Asn38 and Asn83] and one O-linked [serine-126 (Ser126)] carbohydrate chain (Egrie etal., 1986). Studies have shown that EPO's glycocomponent is important for the protein's stability, solubility, biosynthesis, secretion and in vivo bioactivity, but is not required for the hormone's in vitro interaction with its receptor (Wasley etal., 1991; Yamaguchi etal., 1991). In particular, the terminal sialic acids located on the N-linked oligosaccharides are not required for the in vitro bioactivity, but their presence is essential for the in vivo bioactivity of EPO, because they prolong the protein's circulating half-life in the bloodstream (Spivak and Hogans, 1989; Takeuchi etal., 1989).
EPO exerts its haematopoietic activity through interaction with the homodimeric EPO receptor [(EPOR)2] located on erythroid progenitor cells in the bone marrow (Matthews etal., 1996). However, recently, the tissue expression of EPOR has been determined to be widespread, including neurones, endothelial cells, kidney cells and cardiomyocytes, suggesting an alternative and perhaps more important biological activity for EPO (Juul etal., 1998; Masuda etal., 1999). Consistent with this, EPO has recently been shown to possess remarkable tissue-protective activities in preclinical injury models of the spinal cord, kidney, retina, brain and heart. EPO's broad efficacy in a wide range of injury models is related to multiple cytoprotectant pathways, which are activated during disease or tissue injury. These include the inhibition of programmed cell death (apoptosis), attenuation of inflammatory responses, stimulation of angiogenesis and direct recruitment of stem cells (reviewed by Ghezzi and Brines, 2004). Since the chronic use of rhEPO as a tissue-protective therapeutic is likely to cause undesirable side-effects, such as increased blood pressure and thrombosis because of its haematopoietic activity (Stohlawetz etal., 2000), recent studies have identified EPO derivatives that are tissue protective, yet lack haematopoietic activity. Carbamylated EPO (CEPO), desialylated EPO (asialoEPO) and an EPO-R103E mutant have all shown tissue-protective activity without exerting a haematopoietic effect (Erbayraktar etal., 2003; Leist etal., 2004); however, these modifications would lead to substantial production costs that would probably prevent their practical use as a tissue injury prevention therapeutic.
Recombinant EPO has been made in a variety of systems, including bacteria (Lee-Huang, 1984), yeast (Hamilton etal., 2006), insect cells (Kim etal., 2005) and the milk of transgenic goats (Toledo etal., 2006). However, all rhEPO used therapeutically is currently generated in mammalian cell culture (Goto etal., 1988), which is technically complex and expensive. Limitations imposed by the mammalian cell culture system reduce the quantities of EPO that can be produced and limit its therapeutic potential. The utilization of plants as production systems may lower that barrier because of the potential for low production costs and the rapid scalability associated with plant-made biopharmaceuticals. Plants also offer other advantages over conventional expression systems for the production of recombinant proteins, such as the absence of human pathogens, the ability to properly fold and assemble complex multi-subunit proteins, and the potential for direct oral administration of unprocessed or partially processed plant material (Ma etal., 2003).
EPO has been produced in tobacco BY2 cell lines (Matsumoto etal., 1995), tobacco and Arabidopsis plants (Cheon etal., 2004) and the moss Physcomitrella patens (Weise etal., 2007). Matsumoto etal. (1995) were the first to demonstrate that plant-derived EPO exhibited in vitro haematopoietic activity, but found that it lacked in vivo activity, most probably because of the absence of the sialic acids required for this function. However, our interest is focused on the tissue-protective activity of EPO and the plant's natural ability to produce asialylated EPO, which has already been shown to possess tissue-protective activity without excessive haematopoiesis (Erbayraktar etal., 2003).
Many approaches have been used to increase the concentration of biopharmaceuticals in plant tissues. Targeting the recombinant protein to the appropriate plant tissue and subcellular compartment within the cell appears to be critical in reaching suitable expression levels (reviewed by Streatfield, 2007). The subcellular location is also important because of its impact on the glycosylation profile of the recombinant protein, which may affect its potential immunogenicity when administered to humans. Human and plant cells produce the same high-mannose-type N-glycans within the endoplasmic reticulum (ER), whereas complex-type N-glycans, which are attached to proteins as they pass through the late Golgi apparatus of the secretory pathway, are structurally different in plants and animals (reviewed by Gomord etal., 2005). These differences lead to complex α(1,3)-fucose and β(1,2)-xylose glycan motifs on plant glycoproteins that are immunogenic when administered to rodents (Cabanes-Macheteau etal., 1999; Bardor etal., 2003).
The availability of large quantities of biologically active EPO will be essential for the cost-effective, widespread use of EPO as a therapeutic in tissue injury prevention. In this article, we evaluate a number of strategies to enhance the yield of EPO in plant tissue. These strategies include the plant optimization of the human EPO gene, the use of human vs. plant signal peptides, the removal of glycosylation sites and the use of ELP translational fusions. In addition, the effects of subcellular location on EPO yield and glycosylation profile are examined. Finally, we demonstrate the increased receptor-binding affinity of plant recombinant EPO (prEPO) relative to rhEPO and, for the first time, we show the tissue-protective biological activity of prEPO using an in vitro model of kidney epithelial cell death.
Transient expression of EPO in various tobacco subcellular compartments
Three plant expression vectors were constructed that targeted prEPO to the apoplast, ER and chloroplast (Figure 1a). The human endogenous EPO signal peptide (SPEPO) was fused to the mature native EPO sequence (EPONat) to direct the protein into the secretory pathway (secEPO) and, finally, into the apoplast. Retention to the ER (SPEPO·EPONat) was achieved by adding an ER retrieval signal (KDEL) to the C-terminus of the apoplast-targeted secEPO construct. For chloroplast targeting, the transit peptide from the tobacco small subunit RuBisCo gene was fused to EPO (chEPO). To aid in the detection and processing of the recombinant protein, a tobacco etch virus (TEV) protease site, StrepII purification tag and c-Myc detection tag were attached to the C-terminus of the EPO coding sequence. All coding sequences were introduced into the plant binary expression vector pCaMterX and placed under the control of the enhanced cauliflower mosaic virus (CaMV) 35S promoter, a tCUP translational enhancer and the nopaline synthase (nos) terminator.
These constructs were agroinfiltrated into tobacco leaves, and the concentration of prEPO was quantified using a sandwich enzyme-linked immunosorbent assay (ELISA). Of the three subcellular compartments tested, prEPO from the ER-retained construct (SPEPO·EPONat) accumulated to the highest level [88 ng/mg total soluble protein (TSP)], which was seven times higher than the apoplast-targeted construct (secEPO) and 185 times higher than the chloroplast-targeted construct (Figure 1b). Clearly, the subcellular location of prEPO greatly affects its accumulation, with the ER being the best of the compartments tested.
Western blot analysis also showed the presence of different prEPO forms in the various subcellular compartments. When using a polyclonal EPO antibody, the apoplast-, ER- and chloroplast-targeted prEPO were detected as 28-, 31- and 19-kDa bands, respectively (Figure 1c). Occasionally, a 38-kDa Agrobacterium-specific band was detected (Figure 1c, lanes 1, 3, 5) in the plant extracts after agroinfiltration, but this band was not observed in wild-type tobacco tissue (lane 4). The larger size of SPEPO·EPONat relative to secEPO can be attributed to the presence of additional C-terminal amino acids (c. 3.7 kDa), whereas chEPO was considerably smaller, probably because of the absence of glycans. In general, prEPO was significantly smaller than rhEPO produced in Chinese hamster ovary (CHO) cells, which migrates as a broad band (c. 33–39 kDa) because of its microheterogeneity of glycosylation (Egrie and Browne, 2001). Differential glycosylation between plants and humans, particularly the lack of sialic acids on plant glycoproteins, is probably responsible for the major size discrepancy between prEPO and rhEPO.
Transient expression of EPO in the ER of tobacco leaves
To achieve high levels of EPO protein translation, a synthetic version of the human EPO gene (EPOOpt) was constructed to reflect the codon usage and optimal codon context of highly expressed tobacco genes (Campbell and Gowri, 1990; Chiapello etal., 1998). Furthermore, sequences encoding potential cryptic splice sites, polyadenylation signals, destabilizing AT-rich regions, strong secondary structure and TATA-box-like elements were removed (Koziel etal., 1996; Gutierrez etal., 1999). Consecutive strings of A + T and C + G were also avoided, together with CG and TA dinucleotides and regions with strong transcript secondary structure (Figure S1a, see online ‘Supporting Information’). In its entirety, the optimization of the EPO gene resulted in changes to 24% of the nucleotides in 59% of the codons and a decrease in the G + C content from 59% to 48% (Figure S1b).
A collection of ER-targeted EPO expression vectors was constructed in order to evaluate the effect of a plant and mammalian signal peptide [SPTob (plant) and SPEPO (human)], four EPO variants (EPONat, EPOOpt, EPOR103, EPOAgly) and the presence of an ELP fusion partner on prEPO accumulation and biological activity (Figure 2a). The EPO mutant EPOR103 was created by changing a single amino acid (R103E) of the EPOOpt gene with the purpose of separating EPO's cytoprotective bioactivity from its haematopoietic bioactivity. In addition, aglycosylated EPO (EPOAgly) was created by inactivating all glycosylation sites of EPOOpt by substitution mutagenesis of the target asparagines with lysine, and the target serine with valine. EPOAgly provides an alternative approach to separate EPO's biological activities, whilst also reducing the risk of plant glycan immunogenicity when administered to humans.
Agrobacterium-mediated transient assays and quantitative ELISA were performed on the series of ER-targeted EPO constructs. For each combination of signal peptide and ELP fusion, the EPONat variant tended to have the highest concentration of prEPO, followed by EPOOpt, EPOR103 and EPOAgly (Figure 2b). The optimization of the human EPO sequence had no effect on the accumulation of prEPO in the absence of an ELP fusion partner. In the presence of an ELP fusion partner, the optimized gene had a detrimental effect on the prEPO concentration relative to the native gene. On average, the removal of glycosylation sites resulted in prEPO concentrations that were 1/250th of their glycosylated counterparts. In the case of EPONat, the presence of ELP increased the concentration of prEPO twofold when combined with the tobacco signal peptide (SPTob). Without ELP, the native human signal peptide (SPEPO) produced a higher level of EPONat when compared with the tobacco signal peptide. However, the two positive effects were not additive in the case of SPEPO·EPONat·ELP. For the EPOOpt, EPOR103 and EPOAgly variants, comparable levels of prEPO accumulation were observed, irrespective of the signal peptide and ELP fusion arrangement.
To confirm the integrity of transiently expressed prEPO, Western blot analysis was performed on the protein extracts. The aglycosylated expression constructs were not included because of their extremely low levels of accumulation. Many of the EPO protein bands were over-exposed to allow for the detection of less concentrated EPO proteins, thus preventing a quantitative comparison between these expression constructs via Western blots. As shown in Figure 2c, the EPO anti-sera reacted with a single protein, yielding bands of the expected size. Longer exposures of the same blots revealed slightly smaller immunoreactive bands (data not shown), which could represent either degradation products or the partial glycosylation of prEPO. However, all recombinant bands shifted to a smaller single band after deglycosylation (data not shown), suggesting that prEPO exists as a mixture of different glycoforms in the plant.
Generation of transgenic tobacco plants expressing EPO
To complement the transient expression analyses of prEPO, stable transgenic tobacco plants were generated using Agrobacterium-mediated transformation. Four-hundred and five independent transgenic tobacco lines were regenerated from 13 EPO expression constructs. As expected, the prEPO concentration varied significantly among transgenic plants (Figure 3a), which was attributed to the chromosomal position effects associated with random gene insertion (Hobbs etal., 1990; Krysan etal., 2002). The level of prEPO approached 500 ng/mg TSP (i.e. 0.05% of TSP) in the highest expressing lines. The range of EPO accumulation within a given population expressing the same genetic construct was greatest for SPEPO·EPONat at 210-fold, and lowest for SPTob·EPOOpt·ELP at ninefold, with an average difference of 80-fold across all individual populations. For comparison, the mean of the five transformants from each population with the highest concentration of prEPO was used to demonstrate the effects of signal peptides, EPO variants, subcellular localization and presence of an ELP fusion on prEPO protein production (Figure 3b).
The trends observed during the transient analyses correlated well with those observed in the stable transgenic plants, with some minor discrepancies. For each combination of signal peptide and ELP fusion, the EPOOpt variant tended to produce the most prEPO protein, followed by EPONat and EPOR103. Thus, tobacco optimization of the human EPO sequence had a more positive effect in stable plants than for the transient analyses, but still did not have any practical impact on prEPO protein accumulation in tobacco plants. The alteration of the EPO primary sequence (EPOR103) decreased significantly the EPO concentration by more than twofold. To evaluate the effect of glycosylation on EPO accumulation, 29 transgenic SPTob·EPOAgly plants were generated. The top five plants for this construct produced prEPO at a concentration of 0.12 ng/mg TSP, which is an 850-fold decrease relative to their glycosylated counterparts (data not shown). The retention of EPO in the ER of transgenic plants (SPEPO·EPONat) produced seven times more prEPO than in the apoplast-targeted construct (secEPO), which was consistent with the transient analysis. The presence of an ELP fusion partner increased significantly the amount of prEPO (3.5-fold) for all EPO variants utilizing the tobacco signal peptide, which contrasts with the transient analysis, where an increase was observed only for EPONat. The ELP tag also doubled the level of EPOR103 utilizing the human signal peptide. The human signal peptide (SPEPO) had a positive effect (3.5-fold) on the accumulation of EPONat and EPOOpt; however, the presence of ELP did not increase significantly the prEPO concentration for these constructs any further, which was consistent with the transient analysis.
Western blot analysis was also performed on the prEPO proteins from stable transgenic lines in order to verify their sizes and to allow for comparison with their counterparts from transient expression. In all instances, single immunoreactive bands of the expected size were detected from the transgenic plants (Figure 3c), although partially glycosylated forms of prEPO could be detected in longer exposures of the blots (data not shown). The prEPO produced in stable transgenic tobacco plants is similar in size to that obtained from agroinfiltrated tobacco tissue.
Glycosylation analysis of plant-derived EPO
To characterize the glycosylation of prEPO, transgenic plant extracts were treated with various glycosidases, followed by Western blot analysis. The apoplast-targeted prEPO (secEPO) was fully resistant to digestion by endoglycosidase H (EndoH), suggesting that secEPO acquires a complex glycan structure when transported through the Golgi apparatus (Figure 4a, lanes 1 and 2). Conversely, treatment with EndoH resulted in a mobility shift for SPEPO·EPONat and its ELP fusion partner (SPEPO·EPONat·ELP), consistent with the glycosylation pattern of ER-retained glycoproteins. As expected, rhEPO accumulated in an EndoH-resistant form, suggesting the presence of complex glycans. After peptide N-glycosidase F (PNGaseF) treatment, no mobility shift was observed for secEPO (Figure 4b, lanes 1 and 2), indicating the presence of a core α(1,3)-linked fucosyl residue, which is expected for secreted plant glycoproteins. After PNGaseF digestion, SPEPO·EPONat migrated as a smaller band of about 23 kDa (includes 4 kDa of additional C-terminal amino acids). This is nearly identical to that of N-deglycosylated rhEPO protein standard (c. 20.5 kDa), which still contains an O-linked oligosaccharide chain (c. 2 kDa). Partial deglycosylation demonstrated that all three potential glycosylation sites on prEPO were occupied (data not shown). From these results, the smaller size of prEPO can be attributed to smaller plant-specific N-glycans relative to mammalian-produced EPO. Furthermore, enzymatic treatment with α-neuraminidase or O-glycosidases demonstrated that prEPO lacks sialic acid residues and O-linked oligosaccharides (data not shown). Taken together, these results demonstrate that apoplast-targeted EPO (secEPO) contains complex-type N-glycans, with a core α(1,3)-linked fucose residue. In contrast, the KDEL-tagged prEPO proteins were sensitive to EndoH and PNGaseF, suggesting a high-mannose oligosaccharide structure indicative of ER localization with efficient retention/retrieval from the cis-Golgi.
Subcellular localization of recombinant EPO in tobacco leaves
To verify the subcellular localization of prEPO, green fluorescent protein (GFP) was used as a C-terminal fusion, and the resulting constructs (Figure 5a) were agroinfiltrated into tobacco leaves and examined by confocal laser scanning microscopy. Weak secEPO·GFP fluorescence was detected in the apoplast of epidermal plant cells (Figure 5b), which was confirmed by FM4-64 labelling of adjacent plasma membranes (Figure 5b, inset). The addition of a C-terminal KDEL sequence to secEPO·GFP resulted in bright SPEPO·EPONat·GFP fluorescence, which resembled a typical reticulate pattern (Boevink etal., 1996), consistent with ER localization (Figure 5c). Furthermore, fluorescence was absent from the apoplastic space (Figure 5c, inset). SPTob·EPONat·GFP (Figure 5d) showed a very similar localization pattern to SPEPO·EPONat·GFP, demonstrating that the plant and human signal peptides target EPO to the secretory pathway in a similar fashion. To allow for easier visualization, the laser intensity gain was increased for the apoplast-targeted proteins relative to the ER-targeted proteins. As controls, the same three localization constructs were synthesized without the EPO genetic component, and their localization was indistinguishable from the experimental constructs (data not shown).
EPOR-binding analysis of plant-produced EPO
To examine prEPO's ability to bind human EPOR, an indirect ELISA was developed. Briefly, the microtitre plates were coated with anti-EPOR antibody and then incubated with the same set of plant extracts and standards as used when determining the concentration of prEPO (Figures 1b and 2b); however, a saturating amount of EPOR was added to the buffer used for sample dilutions. Anti-EPO antibodies were then utilized for the detection of the bound EPOR:EPO complexes. The EPOR-binding assay demonstrated that the accumulation of prEPO in planta (Figure 6a) followed the same construct-dependent expression trends as observed in Figures 1b and 2b. However, the absolute values of prEPO accumulation detected by the EPOR-binding assay were significantly different from the quantitative ELISA data, suggesting that prEPO has a different binding affinity than rhEPO to human EPOR. The standard curve comprising rhEPO was used as the common reference point in both the EPOR-binding assay and the quantitative EPO ELISA, allowing for a comparison between the two assays. Every EPO variant (EPONat, EPOOpt, EPOR103, EPOAgly and secEPO) demonstrated a very consistent enhancement or reduction in EPOR binding, regardless of the signal peptide utilized or the presence of an ELP fusion partner (Figure 6b). Therefore, the presence of an ELP fusion partner appears to have no effect on the receptor binding of prEPO. The four EPONat and EPOOpt expression constructs showed twofold higher binding affinity than rhEPO for EPOR. The successful engineering of the EPOR103 variant to possess a lower affinity for the haematopoietic receptor was achieved, as a single amino acid change decreased its affinity for EPOR by threefold, resulting in an overall sixfold decrease from its unmodified plant-produced counterpart. In the aglycosylated EPOAgly variants, the protein bound EPOR with a fourfold higher affinity than did the mammalian-produced rhEPO. Finally, the apoplast-targeted secEPO had a 1.5-fold higher affinity for EPOR. Together, these results clearly show the efficient binding of tobacco-produced EPO to human EPOR, which is an important step towards demonstrating the biological activity necessary for therapeutic applications.
Biological activity of plant-derived EPO in an in vitro model of tissue injury
Although prEPO has been shown to bind human EPOR with a high affinity, it is critical to assess the recombinant protein's in vitro biological activity. To enrich for the prEPO protein, which started at a concentration of 0.02% of TSP, and to remove potentially inhibitory substances from the plant extracts, prEPO was purified to 20% of TSP by StrepII affinity chromatography prior to biological activity analysis. The purified SPEPO·EPONat protein's ability to prevent the death of renal tubular epithelial cells (TECs) was compared with that of commercial rhEPO standards. As shown in Figure 7, a basal level of cell death (c. 20%) occurred in the untreated medium control as a result of serum removal during the 24-h culture period. The addition of the pro-inflammatory cytokine interferon-γ (IFN-γ) greatly increased the level of cell death to approximately 45% in TECs supplemented with purified wild-type tobacco eluate. However, equivalent amounts of prEPO or rhEPO were able to reduce significantly the level of TEC death to near baseline values (24% and 21%, respectively). These results demonstrate that plant-derived EPO is biologically active and could function as a general tissue-protective cytokine.
In addition to its haematopoietic effects on the bone marrow, EPO acts as a general tissue-protective cytokine that has the potential to treat numerous diseases and injuries, including stroke, myocardial infarction, spinal cord injury and acute kidney failure (Ghezzi and Brines, 2004). In this study, for the first time, prEPO has been shown to protect kidney cells from inflammatory-induced death.
To enhance the level of EPO accumulation in planta, the human EPO gene was optimized for tobacco expression and targeted to several subcellular compartments. Furthermore, the efficiency of two signal peptides was compared and the presence of an ELP fusion partner was evaluated. Agrobacterium-mediated transient expression in Nicotiana tabacum was used as a convenient method to rapidly test many different expression constructs (Wroblewski etal., 2005) prior to the resource-intensive process of generating transgenic plants. Although, many novel production systems rely on the transient expression of recombinant proteins in plants (Huang and Mason, 2004; Marillonnet etal., 2005), our goal was the generation of stable transgenic plants for field production (Brandle, 2004). Our results show that the experiments conducted using transient analyses were predictive of those conducted using stable transgenic plants. The use of a rigorous replicated experimental design can be credited for improving the utility of transient analyses as a predictor of performance in stable plants.
EPO was targeted to the apoplast, ER and chloroplast in order to assess the optimal subcellular location for high-level accumulation of prEPO. Although secretion to the apoplast has been demonstrated as a means of EPO production in plants (Matsumoto etal., 1995; Cheon etal., 2004; Weise etal., 2007), Matsumoto etal. (1995) found that the apoplast is not a suitable storage environment because EPO remains attached to the plant cell wall which, when combined with the high level of proteolytic activity in the apoplast (Fiedler etal., 1997), could explain the limited accumulation of EPO in this compartment. The addition of a carboxy-terminal KDEL motif to a secreted protein in plants results in its accumulation in the ER lumen (Schouten etal., 1996). Our results correspond with previous reports demonstrating that secreted recombinant proteins are more stable and accumulate to higher levels when targeted to the ER lumen (Huang etal., 2001; Menassa etal., 2001). The ER provides the appropriate environment for complex post-translational modifications, such as glycosylation and disulphide bond formation, to occur (Hwang etal., 1992), which are both important for EPO's stability. We decided to target EPO to the chloroplasts because they have been shown to accumulate high levels of recombinant proteins (Van Molle etal., 2007). As an additional benefit, the chloroplasts should generate an aglycosylated form of prEPO, which eliminates potentially immunogenic plant glycans. However, the chloroplast-targeted EPO accumulated to very low concentrations, probably a result of a lack of glycosylation (Wasley etal., 1991).
After concluding that the ER was the best location for prEPO accumulation, we optimized the human EPO gene sequence for tobacco expression. Many bacterial and human gene sequences have been successfully optimized for increased plant expression. However, our tobacco-optimized EPO gene did not enhance the level of prEPO accumulation when compared with the native EPO gene. Our results are supported by examples demonstrating that the optimization of heterologous eukaryotic genes for expression in plants (Rouwendal etal., 1997; Lonsdale etal., 1998) is less impressive than the optimization of bacterial genes for plant expression (Perlak etal., 1991; Kang etal., 2004).
Although it is common practice to replace the human signal peptide with a plant signal peptide when producing secreted human proteins in plants, it is probably not essential, as the recognition of N-terminal signal peptides is highly conserved amongst eukaryotes (Blobel etal., 1979; Rapoport etal., 1996). The exchange of human for plant signal peptides has been shown to enhance significantly the accumulation of certain proteins (Sijmons etal., 1990; Schaaf etal., 2005); however, our data suggest that the endogenous human EPO signal peptide performed better than the tobacco signal peptide for EPO expression. Other examples of human signal sequences functioning more efficiently than plant signal sequences in plants also exist (Ma etal., 2005).
The fusion of an ELP partner has been shown to increase the yield of several recombinant target proteins in transgenic tobacco leaves (Patel etal., 2007; Floss etal., 2008) and seeds (Scheller etal., 2006). In this study, the fusion of ELP to EPO had a negligible effect on many of the prEPO expression constructs, although the presence of an ELP fusion partner significantly enhanced the accumulation of specific EPO constructs. However, the enhancement of EPO accumulation when utilizing an ELP fusion was much lower than for other recombinant proteins, which showed up to a 100-fold increase in the concentration of the target protein (Patel etal., 2007). The beneficial effect of ELP on recombinant protein accumulation is probably protein specific but, in our case, the limited effect of ELP fusion on EPO accumulation may be a result of the potential toxicity of EPO-ELP to the plant (Cheon etal., 2004). Within 4 days of agroinfiltration, complete tissue necrosis could be observed within regions of the infiltrated leaf panels. Moreover, the production of EPO-ELP in transgenic tobacco plants retarded their vegetative growth and resulted in malformed leaves with necrotic lesions. As a result, plant cells expressing high levels of EPO-ELP during transient analysis or transgenic plant generation simply die, thus limiting the concentration of prEPO in planta. The C-terminal ELP tag had no effect on the receptor-binding affinity of EPO, demonstrating that ELP should have no effect on the biological activity of recombinant proteins, as was recently demonstrated for antibody–ELP fusions (Floss etal., 2008).
The size of prEPO was significantly smaller than that of mammalian CHO cell-derived rhEPO because of differences in glycosylation patterns. The N-glycans of EPO are important for the stability, solubility, transport and biological activities of the protein (Yamaguchi etal., 1991). Differences in glycosylation between plants and humans could be a potential barrier to the therapeutic use of prEPO for humans because of the presence of immunogenic α(1,3)-fucose and β(1,2)-xylose motifs on plant glycoproteins (Tekoah etal., 2004). We chose to remove the glycosylation sites of EPO in consideration of the potential immunogenicity issues, and to evaluate the role of N-linked glycans on the accumulation and receptor-binding affinity of the aglycosylated form. Our aglycosylated prEPO was made by introducing lysine residues at each of the N-linked glycosylation sites and by changing the serine at the O-glycosylation site to valine. These substitutions have been shown previously to increase the stability and solubility of aglycosylated EPO, without affecting the protein structure (Cheetham etal., 1998; Narhi etal., 2001). However, the absence of plant glycans dramatically reduced the level of prEPO accumulation in plant leaves, suggesting that glycosylation is important for correct folding, biosynthesis, stability and/or secretion of EPO in planta. Regardless, the binding affinity to EPOR actually increased. Our work differs from other reports, which demonstrate that the removal of N-glycans has a limited effect on the stability, folding, assembly and activity of plant-derived antibodies (Nuttall etal., 2005; Rodriguez etal., 2005). Therefore, glycosylation may be more important for the accumulation of cytokines than for the accumulation of antibodies in plants. As an alternative approach to prevent the addition of immunogenic complex plant glycans to prEPO, we fused an ER retention signal to EPO in order to restrict glycosylation to exclusively high-mannose-type N-glycans (Triguero etal., 2005; Petruccelli etal., 2006). The ER-retained prEPO accumulated in an EndoH- and PNGaseF-sensitive form, indicating the absence of complex N-glycans, including the immunogenic α(1,3)-fucose and β(1,2)-xylose motifs known to be attached to glycans when passing through the medial or trans-Golgi compartments (Pagny etal., 2000; Sriraman etal., 2004). Moreover, the presence of the ER retention signal was solely responsible for efficient ER retrieval, as untagged secreted prEPO was both EndoH and PNGaseF resistant, thus demonstrating EPO's ability to be fully secreted. In addition, no significant difference in glycosylation pattern was observed between prEPO produced by transient expression or stable transformation. The absence of O-glycan on prEPO should be of little concern, as O-glycosylation has no essential role in either the in vitro or in vivo biological activity of the cytokine (Higuchi etal., 1992). However, terminal sialic acids on the oligosaccharide chains of EPO are essential for in vivo haematopoietic bioactivity, as they prevent the hepatic elimination of the protein and extend the circulating half-life of EPO (Fukuda etal., 1989). Therefore, all commercially available rhEPOs for the treatment of anaemia are produced in mammalian cell cultures (Jacobs etal., 1985; Lin etal., 1985; Fibi etal., 1995), which are able to synthesize sialyated glycoproteins. However, reports have shown that asialoEPO, made by the enzymatic desialylation of rhEPO, provides tissue protection without inducing red blood cell production (Erbayraktar etal., 2003). A potential explanation for this effect is that prolonged periods of high circulating levels of EPO are needed to induce haematopoiesis, but brief exposure to EPO is sufficient to initiate tissue-protective programmes (Erbayraktar etal., 2003). Therefore, an inability to produce sialic acid may be a beneficial feature of EPO production in plants, as a simple and straightforward means of separating EPO's tissue-protective activity from its haematopoietic activity.
We developed an EPOR-binding ELISA which allowed us to compare the relative binding affinities of prEPO and rhEPO to human EPOR. The aglycosylated prEPO has a fourfold higher receptor-binding affinity than fully glycosylated rhEPO, which is similar to the increased receptor-binding activity shown for aglycosylated Escherichia coli-produced EPO (Delorme etal., 1992). In addition, our prEPO lacking sialic acid residues exhibited a twofold increase in binding affinity, which has also been demonstrated previously for desialylated EPO (Elliott etal., 2004a). As the EPO molecule is positively charged and EPOR is negatively charged at physiological pH, evidence suggests that the glycans may shield the attractive electrostatic interactions present between EPO- and EPOR-binding surfaces (Darling etal., 2002). This is particularly true in the case of sialic acid, which is highly negatively charged, thus decreasing the affinity of mammalian-produced EPO for the negatively charged EPOR via charge repulsion (Elliott etal., 2004b). Therefore, prEPO without sialic acid residues should bind more efficiently than rhEPO to human EPOR. Leist etal. (2004) have demonstrated that a cytoprotectant EPO analogue lacking haematopoietic activity can be designed by mutating amino acid 103 (i.e. R103E). This amino acid does not interfere with the protein's conformation, but is critical for binding to the haematopoietic receptor (Grodberg etal., 1993). As expected, the ability of our prEPO(R103E) mutant to bind EPOR was reduced by sixfold, although its biological activity has yet to be tested. This prEPO(R103E) mutant could also be used to further separate EPO's tissue-protective activity from its haematopoietic activity when produced in plants.
Matsumoto etal. (1995) and Weise etal. (2007) have demonstrated that plant-derived EPO induces the proliferation of EPO-sensitive cell lines; however, in the report by Matsumoto etal. (1995), in vivo haematopoietic activity was absent, and was attributed to the rapid removal of prEPO from the circulation because of a lack of sialic acid residues. To our knowledge, this is the first report demonstrating that prEPO possesses tissue-protective bioactivities by reducing the susceptibility of renal TECs to cytokine-induced cell death. TECs comprise more than 75% of renal parenchymal cells, and their susceptibility to injury directs the long-term function of allografts, as tubular injury can be a primary cause of nephron loss. Augmenting the endogenous capacity of TECs to resist injury would be a useful strategy in renal transplantation, as well as in other forms of organ injury in which epithelial cells play a prominent role (Du etal., 2004). As there are no current therapeutic agents that specifically protect epithelial cells from inflammation-induced injury or death, there has been growing interest in the use of novel strategies for tissue injury attenuation, including EPO. Therefore, recent studies have demonstrated that EPO protects the kidney against injury caused by ischaemia–reperfusion by utilizing its anti-apoptotic action, which may translate into improved long-term function of the kidney (reviewed by Johnson etal., 2006a). Therefore, prEPO could be administered to renal allograft recipients or to the perfusion solutions of kidney organs to attenuate renal allograft injury caused by ischaemia, toxins and immunological rejection (Johnson etal., 2006b; Spandou etal., 2006). EPOR is also expressed by gut epithelial cells (Juul etal., 1999). Furthermore, EPO has been shown to decrease apoptosis of gut epithelial cells (Cuzzocrea etal., 2004; Guneli etal., 2007), suggesting that prEPO may have a benefit in inflammatory bowel disease by direct oral delivery, which has been shown for other plant-produced cytokines (Menassa etal., 2007). Although current levels of EPO expression in plants are relatively low, we believe that tobacco can produce an economically feasible amount of prEPO, as only low doses of EPO are needed to exert a therapeutic response.
In summary, this study demonstrates that higher accumulation levels of EPO can be achieved in the ER than in the apoplast or chloroplasts. Optimization of the human EPO gene for plant expression has no effect, and the human EPO signal peptide is processed more efficiently than the tobacco signal peptide. Glycosylation is very important for the stability of EPO in planta, whereas the presence of an ELP fusion has a limited effect on EPO concentration. The transient and stable expression analyses are well correlated, and the maximum concentration of EPO approaches 0.05% of TSP in tobacco leaves. The apoplast- and ER-targeted EPO proteins are correctly targeted to their respective subcellular compartments, and complex immunogenic glycans are present only on apoplast-targeted EPO and absent from ER-targeted EPO. Most importantly, our prEPO binds to human EPOR with an increased affinity, and also has anti-apoptotic activity, which may reflect a capacity to prevent tissue injury in human disease. These results suggest that tobacco plants can provide an efficient and less expensive means of producing a tissue-protective EPO analogue that does not possess the potentially harmful side-effects associated with excessive haematopoietic activity. Future work will focus on increasing the concentration of EPO in planta and on the evaluation of prEPO's biological activity in various in vivo models of tissue disease and injury.
Construction of plant expression vectors
A plant-optimized gene (SPTob·EPOOpt) with a PR1b secretory signal peptide from tobacco (Cutt etal., 1988) was designed to mimic the codon usage of highly expressed N. tabacum genes, whilst avoiding all potentially deleterious processing signals and destabilizing motifs. Briefly, the entire synthetic gene was constructed using a combined ligase chain reaction/polymerase chain reaction (LCR/PCR) approach (Au etal., 1998) with a set of overlapping oligonucleotides designed by the web-based program Gene2Oligo (Rouillard etal., 2004). This approach was also employed to synthesize the native human EPO cDNA sequence with its own signal peptide (SPEPO·EPONat), the chloroplast transit peptide (TPRuB) and the ELP and GFP fusion partners. The minor nucleotide changes necessary to generate SPTob·EPOR103 and SPTob·EPOAgly were performed by substituting one or more oligonucleotides in the LCR mix, followed by reassembly. To assist in subsequent cloning steps, BamHI and EcoRI restriction sites were included at the 5′ and 3′ ends of the completed constructs, respectively. In addition, a KasI site was added to the 3′ end of the various EPO genes to create a two-amino-acid linker (glycine-alanine) and to allow for in-frame ligation with various C-terminal fusions, protease cleavage sites, ER retention signals and purification tags. The secEPO and chEPO constructs were synthesized by removing the appropriate 3′ sequence from SPEPO·EPONat by PCR amplification. The remaining expression constructs were created by interchanging previously synthesized construct components using overlap extension PCR combined with restriction enzyme digestion of C-terminal elements, followed by in-frame ligation. Once completed, the constructs were digested with BamHI and EcoRI, cloned into pBluescript for sequencing and then moved into the plant binary expression vector pCaMterX (Laurian Robert, Agriculture and Agri-Food Canada, Ottawa, pers. commun.). The coding sequences were under the control of the dual-enhancer CaMV 35S promoter (Kay etal., 1987), a tCUP translational enhancer (Wu etal., 2001) and the nos terminator. The expression constructs were electroporated into Agrobacterium tumefaciens strain EHA105 (Hood etal., 1993) and then used for plant transformation. All expression vector sequences have been deposited in GenBank (Figure S2, see online ‘Supporting Information’).
Agrobacterium-mediated transient expression assays and tobacco transformation
For transient expression, the Agrobacterium strains were used to infiltrate the leaves of 10–14-week-old N. tabacum plants. Agrobacterium was prepared for leaf infiltration as described previously (Kapila etal., 1997; Yang etal., 2000). Briefly, the Agrobacterium suspensions were adjusted to a final optical density at 600 nm (OD600) of 2.4 and then injected into the abaxial air spaces of intact leaves just under the epidermal surface using a 1-mL syringe. To account for plant to plant variability, leaf to leaf variability and position on a leaf, comparably sized leaves from six different plants of similar age were agroinfiltrated for each expression construct. In addition, the agroinfiltrated panels were systematically distributed across the leaf surface. After infiltration, the plants were maintained in a controlled growth chamber at 22 °C with a 16-h photoperiod for 4 days, and the individual infiltrated panels were collected and analysed separately.
For the generation of stable transgenic plants, expression constructs were introduced into low-alkaloid tobacco (cv. 81V9) (Menassa etal., 2001) using Agrobacterium-mediated transformation of leaf discs, as described by Miki etal. (1999). Primary transformants were grown in a glasshouse for 4–6 weeks following transfer to soil. The first four true leaves were sampled once they had reached 25 cm in length, and were used to represent the concentration of recombinant protein in the whole plant. Seeds were collected from the transgenic tobacco lines with the highest EPO concentration, and were used to form subsequent generations by self-fertilization and the selection of homozygous lines.
Plant protein extraction
For each sample, TSP was extracted from four 7-mm leaf discs (approximate fresh weight, 25 mg) of transgenic and wild-type plants by homogenization with a Mixer Mill MM 300 (Retsch, Haan, Germany). The resulting frozen powdered leaves were then resuspended at 4 °C in 300 µL of extraction buffer [phosphate-buffered saline (PBS), pH 7.4, 0.1% Tween-20, 2% polyvinylpolypyrrolidone (PVPP), 1 mm ethylenediaminetetraacetic acid (EDTA), 100 mm ascorbic acid, 1 mm phenylmethylsulphonylfluoride (PMSF) and 1 µg/mL leupeptin]. The extract was clarified twice by centrifugation at 20 000 g for 10 min at 4 °C. The TSP concentration was measured according to the method of Bradford (1976) using Bio-Rad reagent (Bio-Rad, Hercules, CA, USA), with bovine serum albumin as a standard.
Quantification of EPO protein levels by ELISA
The quantification of prEPO in tobacco leaf extracts was achieved by sandwich ELISA. Nunc-Immuno MaxiSorp surface plates (Nalge Nunc, Rochester, NY, USA) were coated with 2 µg/mL of mouse anti-EPO monoclonal antibody (01350; Stem Cell Technologies, Vancouver, BC, Canada) diluted in disodium phosphate buffer (0.1 m, pH 9.0), and incubated overnight at 4 °C. The wells were blocked with 2.7% ELISA Blocking Reagent (Roche, Mannheim, Germany) in PBS for 1 h at room temperature. Plant extracts were serially diluted in blocking buffer (PBS containing 2.7% ELISA Blocking Reagent and 0.05% Tween-20) and incubated on the plate overnight at 4 °C. The plate was then incubated with 4 µg/mL of rabbit anti-EPO antibody (E2531; Sigma, St. Louis, MO, USA) diluted in blocking buffer for 1.5 h at room temperature. Next, the plates were incubated with a 1 : 1000 dilution of horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (170-6515; Bio-Rad) diluted in blocking buffer for 1 h at room temperature. The plates were washed five times between incubation steps with PBS containing 0.05% Tween-20. The plates were developed by the addition of 2,2′-azino-bis(3-ethylbenzthiazoline)-6-sulphonic acid (ABTS) substrate (A-1888; Sigma), and the absorbance was measured at 405 nm with a Bio-Rad 550 microplate reader. To generate a standard curve, recombinant human EPO (CRE600B; Cell Sciences, Canton, MA, USA) was diluted with blocking buffer to concentrations between 0.125 and 8 ng/mL, and processed as described above. Two individual samples were taken from every transgenic plant or agroinfiltrated leaf panel and treated independently during ELISA.
Western blot analysis
For Western blot analysis, plant extracts were resolved on a 12% sodium dodecylsulphate (SDS) polyacrylamide gel and then transferred to a nitrocellulose membrane by semi-dry electroblotting. The membranes were blocked with 1% Western Blocking Reagent (Roche) in Tris-buffered saline (TBS, 50 mm Tris, 150 mm NaCl, pH 7.5) overnight at 4 °C. The membranes were incubated with a 1 : 500 dilution of rabbit anti-EPO antibody (E3455-10; USBiological, Swampscott, MA, USA) for 1 h at room temperature with gentle shaking. The primary antibody was detected with a 1 : 5000 dilution of HRP-conjugated goat anti-rabbit IgG (Bio-Rad) and visualized using an ECL kit (GE Healthcare, Mississauga, ON, Canada), according to the manufacturer's instructions. The membranes were washed four times between each step with TBS containing 0.05% Tween-20, and all antibodies were diluted in TBS with 0.5% Western Blocking Reagent.
Enzymatic deglycosylation of EPO
Total plant protein extracts and recombinant EPO were deglycosylated with PNGaseF (New England Biolabs, Ipswich, MA, USA) or EndoH (Sigma) for 24 h at 37 °C, according to the manufacturer's instructions. PNGaseF cleaves all high-mannose, hybrid and complex-type oligosaccharides from N-linked glycoproteins, except for those glycans containing a core α(1,3)-linked fucose residue. EndoH is able to remove high-mannose N-linked glycans, but not complex-type glycans, from glycoproteins. In addition, the Enzymatic Protein Deglycosylation Kit (Sigma) was employed to remove any potential O-linked glycans and sialic acids from the protein samples, as described by the manufacturer. Control samples were treated the same, except that no enzyme was added. Finally, the samples were analysed by sodium dodecylsulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotted as described in the previous section.
For transient expression of the GFP constructs, the appropriate Agrobacterium strains were infiltrated into tobacco leaves as described above. For staining of the plasma membrane, a 5-µg/mL solution of FM4-64 (Invitrogen, Burlington, ON, Canada) was injected into the abaxial surface of the leaf, and the resulting tissue was excised and mounted in FM4-64 solution. A Leica TCS SP2 confocal laser scanning microscope equipped with a 63× water immersion objective (Leica Microsystems, Wetzlar, Germany) was used to examine the subcellular localization of GFP fluorescence and FM4-64 staining. For the simultaneous imaging of GFP and FM4-64, the two fluorophores were excited with a 488-nm argon laser line and fluorescence was detected at 500–525 nm and 635–700 nm, respectively.
The binding capacity of prEPO to human EPOR (E0643; Sigma) was assessed by modifying the above quantitative ELISA protocol. Briefly, microtitre plates were coated with 4 µg/mL of mouse anti-EPOR antibody (MAB3071; R&D Systems, Minneapolis, MN, USA) in disodium phosphate buffer and incubated overnight at 4 °C. Subsequently, the wells were blocked with 2.7% ELISA Blocking Reagent in PBS for 1 h at room temperature. Soluble plant extracts and EPO standards were serially diluted in blocking buffer containing 200 ng/mL of EPOR and incubated on the plate overnight at 4 °C. The remainder of the procedure was performed as for the quantitative ELISA described above.
Purification and analysis of biological activity of plant EPO
For the purification of prEPO, 1 kg of frozen tobacco leaves was homogenized in 3 L of cold extraction buffer. The homogenate was filtered through two layers of Miracloth (EMB Biosciences, Mississauga, ON, Canada), and the extract was clarified twice by centrifugation at 15 000 g for 15 min at 4 °C. To enrich for prEPO, the supernatant was filtered through a PLHK 100-kDa and a PLGC 10-kDa Prep/Scale TFF cartridge (Millipore, Burlington, MA, USA). The resulting 10-kDa retentate was further concentrated using a Jumbosep spin column with a 10-kDa cut-off (Pall Corporation, Mississauga, ON, Canada), and then passed through a 0.22-µm membrane filter. Soluble proteins were then applied to a Strep-Tactin MacroPrep column (IBA, St. Louis, MO, USA) and prEPO was eluted according to the manufacturer's instructions. The biological activity of prEPO was assessed by its ability to prevent the cellular death of renal TECs, according to Du etal. (2004) with minor modifications. TECs (2.5 × 105) were added to 24-well plates in triplicate and cultured overnight in complete K1 medium containing serum. After confluence was reached, the medium was replaced with fresh K1 medium without serum or growth factors to arrest cell division. TECs were then pretreated for 5 h with 70 ng/mL of prEPO or commercial recombinant EPO (Amgen, Thousand Oaks, CA, USA) diluted in K1 medium. To promote cell death, IFN-γ (BD Biosciences, Mississauga, ON, Canada) was added to the medium, together with additional EPO (70 ng/mL), and the TECs were cultured for an additional 24 h. TEC monolayers were released from the plates by a brief incubation with trypsin-EDTA solution (Sigma), and then incubated with Annexin-V conjugated with fluorescein isothiocyanate (FITC) and 7-aminoactinomycin D (7-AAD) (BD Biosciences). The level of cell death was determined with a FACSCalibur flow cytometer and analysed by CellQuest software (BD Biosciences).
The authors wish to thank Laura Slade for technical assistance and Alex Molnar for assistance with the preparation of the figures. Thanks are due to Dr Jussi Joensuu, Dr Rima Menassa, Dr Patrick Telmer and Alex Richman for critical comments on the manuscript and helpful discussions. This research was supported by the Agriculture and Agri-Food Canada Matching Investment Initiative Programme. We thank the Natural Sciences and Engineering Research Council (NSERC) Postgraduate Scholarship for providing financial support to A.J.C.