Plants including the whole body, specific tissues or organs and plant-derived suspension cells are considered as attractive and competitive platforms for recombinant protein production. Compared with Escherichia coli and mammalian cells, plant-based production systems have been demonstrated to have many advantages. First, plant-based production systems are more flexible, as well as easier to operate, manage, and scale up, and thus can decrease the production cost of recombinant proteins to a great extent. Second, plants, similar to mammalian cells, are able to perform post-translational modification of recombinant proteins such as glycosylation, and thus additional cost for in vitro protein modifications can be avoided (Frenzel et al., 2013; Sil and Jha, 2014; Kolotilin et al., 2015). Third, long-term continuous recombinant protein production can be realized in plant platforms because transgenes can be stably integrated into the nuclear genome of host plants, faithfully inherited, and expressed in later generations. Furthermore, plant-derived recombinant proteins may be safer than those from E. coli or mammalian cells because the risk of contamination with human pathogens, which is always a concern when using mammalian cells as a bioreactor, can be well circumvented by plant-based production systems (Thie et al., 2008; Ni and Chen, 2009; Merlin et al., 2014). Because of these properties, various bioactive pharmaceutical proteins have been produced in plants since first expression of a human growth hormone in transgenic tobacco and sunflower callus tissue (Barta et al., 1986), and expression of antibodies, vaccines, hormones, growth factors, and cytokines (De Muynck et al., 2010; Desai et al., 2010; Xu et al., 2011; Huang and McDonald, 2012).

Monoclonal antibodies (mAbs) are protein complexes containing four subunits with two identical light chains (LC) and two identical heavy chains (HC). MAbs are important in biological research, clinical diagnosis, and recently immunotherapy for various diseases and cancer (De Muynck et al., 2010). Unlike other single chain recombinant proteins, the production of mAbs requires simultaneous expression of two genes coding for both LC and HC, and the correct folding of four polypeptides linked by disulphide bonds. The discovery that plant can efficiently express and correctly assemble functional antibodies (Hiatt et al., 1989) have made plants an alternative antibody production system, and since then, many recombinant antibodies have been produced in various plants, including moss (Decker and Reski, 2008), algae (Franklin and Mayfield, 2005), and higher plants (Stoger et al., 2005; De Muynck et al., 2010; Xu et al., 2011, 2012; Schillberg et al., 2013).

Previously, the LC and HC genes of a mAb were expressed in two different expression cassettes on one T-DNA region of a vector, or expressed separately in individual vectors which were co-transformed to the same plant, or expressed in different transgenic plants which were cross-fertilized to produce the functional antibody (De Muynck et al., 2010; Ko, 2014). Because the two genes are expressed separately, it is difficult to control their relative expression level even though identical regulatory elements are used. In fact, most of the previous reports have produced unbalanced LC and HC in both transgenic plants and mammalian cells (Voss et al., 1995; Law et al., 2006; De Muynck et al., 2010; Chng et al., 2015). A varied LC:HC ratio is usually unfavorable for the folding of functional mAbs, and affects both the level and quality of mAbs (Schlatter et al., 2005; Law et al., 2006; Lee et al., 2009; Ho et al., 2013b). The use of internal ribosome entry site (IRES) to translate two polypeptides (LC and HC) from one mRNA also results in an unbalanced expression because of the lower efficiency of the IRES directed downstream gene expression by a cap-independent translation mechanism (Hennecke et al., 2001; Ho et al., 2012, 2013a,b). In contrast, the use of 2A peptide from the Aphthovirus Foot-and-mouth disease virus (FMDV) for high level mAb expression has been reported in both the human embryonic kidney 293 and the Chinese hamster ovary (CHO) cells (Fang et al., 2005; Chng et al., 2015), but this strategy for mAb expression in transgenic plant system has not been reported so far.

Bevacizumab is a humanized mAb that targets to the vascular endothelial growth factor (VEGF) antigen (Presta et al., 1997; Ferrara et al., 2005), which is widely over expressed in a variety of human solid tumors and plays a key role in tumor angiogenesis (Ellis and Hicklin, 2008; Goel and Mercurio, 2013; Domigan et al., 2015). Bevacizumab neutralizes VEGFs, prevents their interactions with VEGFR-1 and VEGFR-2 receptors, and thus blocks the downstream signal transductions for tumor angiogenesis (Wang et al., 2004). Bevacizumab is derived from the murine VEGF mAb A4.6.1. It has 93% human and 7% murine sequence, and has similar biochemical and pharmacologic properties to the original murine mAb. It neutralizes all isoforms of human VEGF (hVEGF) with high affinity and inhibits VEGF-induced proliferation of endothelial cells and tumor angiogenesis, but with reduced immunogenicity and longer circular half-life as compared to the murine antibody (Gerber and Ferrara, 2005). It has been widely applied in clinical practice of metastatic colorectal cancer, glioblastoma, non-small-cell lung cancer, metastatic kidney cancer, advanced cervical cancer, platinum-resistant ovarian cancer (Giantonio, 2006; Ali et al., 2008; Tonini et al., 2013; Tewari et al., 2014; Liu and Matulonis, 2015; Mason, 2015)¹. Bevacizumab (Avastin) is approved by the US Food and Drug Administration for clinical use in cancer therapy (Bellmunt, 2007; Pal et al., 2014).

Although increasing evidence suggests the great potency of Bevacizumab for various disease therapies, there are still some obstacles which limit its widespread applications (Hayashi et al., 2014). One of the shortcomings is the high cost of Bevacizumab-mediated therapy. The high cost of Bevacizumab is greatly influenced by its production, which is solely dependent on mammalian expression systems (Makino et al., 2011; Spadiut et al., 2013).

In this study, we reported the first plant made Bevacizumab by using transgenic rice callus as a gene expression system. Rice is both a main stuff crop and a model plant for biological studies because of its small genome and the high efficient gene transformation system. Rice seeds and suspension cultures have been reported as good systems for the production of many high value bioactive proteins (Kim et al., 2008; He et al., 2011; Kuo et al., 2013; Corbin et al., 2016; Fujiwara et al., 2016). Codon-optimized Bevacizumab LC and HC genes were designed, synthesized as a polyprotein with a 2A self-cleavage linker peptide from the FMDV, cloned into a plant binary vector under a constitutive maize ubiquitin promoter, and transformed into rice nuclear genome through Agrobacterium-mediated transformation. High levels of biologically functional Bevacizumab were expressed and purified by affinity chromatography. An enzyme-linked immunosorbent assay (ELISA) binding assay showed that the rice expressed Bevacizumab and the commercial Avastin had similar binding affinity to the VEGF target.

In this report, we designed and expressed a biologically functional mAb (Bevacizumab) in transgenic rice callus. The plant expression system has many advantages including high-level productivity, easy management and scaling-up, safety and low cost. To maximize the benefit of the plant expression system, many strategies have been developed to improve the expression and to increase the production. For example, transgenes can be stably integrated into the nuclear genome or transiently expressed using viral or Agrobacterium vectors (Vaquero et al., 1999; Kathuria et al., 2002; Rodriguez et al., 2005; Sainsbury et al., 2008; Villani et al., 2009). In addition, stable expression in cell cultures can produce uniform recombinant proteins in controlled environments by using bioreactors so that good manufacturing practice (GMP) can be applied (Shih and Doran, 2009; Xu et al., 2011; Huang and McDonald, 2012). In this study, we produced bioactive Bevacizumab in stably transformed rice callus and the rice-derived Bevacizumab content reached 242.8 mg/Kg FW (Figure 4). With the development of the rice suspension culture system, it is possible to scale up the production to reach large-scale production by using the bioreactor technology such as the continuous or semi-continuous production system to reduce the production time and manufacturing cost (Tekoah et al., 2015; Yang et al., 2015; Corbin et al., 2016).

The complete mAb consists of two identical LCs and HCs connected by disulfide bonds. Low-level expression of mAbs in previous reports was mostly affected by the expression strategies. Unbalanced expression of the LC and HC was unavoidable in the previous systems when the two genes coding for the LC and HC were expressed separately, or expressed in one bicistronic mRNA separately by the IRES because the downstream open reading frame (ORF) was only expressed at ∼10% level of the upstream one (Minskaia and Ryan, 2013; Minskaia et al., 2013). Unbalanced expression of LC and HC is unfavorable to the yield and quality of mAb production because efficient assembly and modification of mAb seem to require an optimal ratio of LC and HC (Schlatter et al., 2005; Ho et al., 2013b; Chng et al., 2015). However, contradictory results were reported on the optimal ratio of LC and HC in recombinant mAb production. Excessive expression of LC was considered to be beneficial for mAb production because excess LC would allow more efficient mAb folding, assembly, and the clearance of HC, which is prone to aggregation when expressed in excess (Vanhove et al., 2001; Lee et al., 2009). Ho et al. (2013b) reported higher titers of mAb production at a LC:HC ratio of 3.43. Similarly, Schlatter et al. (2005) observed efficient production of three mAbs with LC:HC peptide ratios of 2, 2.5, and 3.1 in CHO cells. On the contrary, separate groups also reported that excessive HC was beneficial to IgG expression (Jiang et al., 2006; Fallot et al., 2009). However, excessive expression of either LC or HC produced LC or HC dimmer and monomer molecules besides the LC₂HC₂ mAb (Ho et al., 2013b). Unlike the other systems, using of 2A self-cleavage peptide can ensure equal molars expression of LC and HC peptides to avoid unbalanced expression, and thus produce high levels of mAbs (Chng et al., 2015).

The FMDV RNA genome contains a single ORF, which encodes a polyprotein. The FMDV 2A peptide between the 2A and 2B regions of the polyprotein is self-cleaved to produce the viral 2A and 2B proteins (Ryan et al., 1991). The self-cleavage of 2A peptide is co-translational by a “Stop-Go” or “Stop-carry on” recoding mechanism, by which the upstream protein translation is paused by the 2A sequence and the peptide backbone is broken between the last two amino acids, Glycine (G) and Proline (P) of the 2A peptide (Atkins et al., 2007; Doronina et al., 2008; Sharma et al., 2011). The 2A peptide also drives a translational recoding of the downstream protein. Unlike other systems, 2A peptide-mediated cleavage is not dependent on cellular factors in different cell types because the self-cleavage activity is only dependent on the ribosomes, which are structurally conserved among all eukaryotic organisms. Thus, the 2A peptide is active in all eukaryotic systems, and has been used widely as a co-expression tool in the expression of two or more proteins (Donnelly et al., 2001; de Felipe et al., 2006). Recently, 2A peptide was applied in stable and high levels expression of mAbs in CHO cells and mice, and was considered to have potential applications of in vivo delivery of mAbs in gene therapy (Fang et al., 2005; Chng et al., 2015). 2A peptide was also applied in the production of a single chain antibody in plants (Smolenska et al., 1998). However, the application of 2A in the production of complete mAb in plants has not been reported so far.

We reported the expression of a mAb (Bevacizumab) LC and HC as a polyprotein with the 2A linker from FMDV in one gene expression cassette driven by a strong constitutive ubiquitin promoter in transgenic rice callus. High levels of functional antibodies were produced (Figure 3). We achieved the expression of more than 200 mg/Kg FW of Bevacizumab mAb in transgenic rice cells, which was higher than most of the reported recombinant mAb expression systems (Gomord et al., 2004; Girard et al., 2006; Villani et al., 2009; De Muynck et al., 2010; Madeira et al., 2015). The high levels of expression could be attributed to the following factors. First, the codon optimization makes the mRNA stable and the protein translation efficient. Second, a strong constitutive ubiquitin promoter increases the transgene transcription level. Third, single copy transgene by the Agrobacterium-mediated transformation reduces the chances of gene silencing. Last but not least, the expression of a polyprotein using 2A self-processing peptide favors the accumulation of functional mAb. The using 2A peptide expression system may affect the mAb yield by the following ways. First, the self-cleavage of 2A linked polyprotein produces equal molecules of HC and LC, which might be an optimal ratio for mAb accumulation. Western blot revealed that the LC and HC of Bevacizumab expressed with 2A polypeptide were at similar levels in most transgenic lines, but unbalanced expressions of LC and HC were observed when they were on different expression cassettes (Figure 3). Although the excess of either LC or HC might speed up the mAb assembly as shown in a mathematical modeling (Gonzalez et al., 2002), a 1:1 LC:HC ratio is more economical because IgG mAbs have equal numbers of LC and HC subunits. Second, the self-cleavage is co-translational, so that both the LC and HC are simultaneously produced and mobilized into ER, where mAb can be immediately assembled and folded. Third, correct and complete assembly of mAb may reduce proteolytic degradation of either LC or HC monomers or dimmers when present in excess because incorrectly or incompletely folded antibodies are susceptible to protease attack (Gardner et al., 1993). Post-translational protein degradation has been a common phenomenon that significantly influences the yield of foreign protein production (Doran, 2006; De Muynck et al., 2010; Desai et al., 2010; Xu et al., 2011; Hehle et al., 2016). But, the degradation of mAb was not observed in our studies based on the Western blot results. This is probably because of the correct and complete assembly of mAb from LC and HC subunits generated from the 2A self-cleavage polyprotein (Figure 2C). The weak lower molecular weight bands observed in Figure 4 were probably degradation products during the process of protein extraction and purification because such bands were not observed before this process (Figure 2C).

Glycosylation is an important property of recombinant mAb that affects its binding affinity and circular half-life, and causes immunogenicity (Saint-Jore-Dupas et al., 2007; Reusch and Tejada, 2015). It has been shown that plant expressed proteins are differently modified as compared to mammalian systems. The plant-specific glycosylation including the β 1, 2-xylose and a core α 1, 3-fucose takes place in the Golgi apparatus. To avoid plant-specific glycosylation, two strategies are generally used: to keep the recombinant protein in the ER with an ER retention signal peptide or to knockout plant genes coding the plant-specific glycosylation enzymes by glycoengineering (Strasser et al., 2008, 2014; Castilho and Steinkellner, 2012; Castilho et al., 2014; Kurogochi et al., 2015). In this report, we tried to target both the HC and LC to ER by adding KDEL signal peptides at the C-termini, but were unsuccessful in preventing the plant-specific glycosylation (Table 2), indicating that the KDEL signals were not functioning properly. One reason for this malfunction of the KDEL signal could be related to results of the self-processing of the 2A peptide linker between the HC and LC. It was reported that the cleavage of the 2A peptide between the last two amino acids, Glycine (G) and Proline (P) would result in the adding of a GSGQLLNFDLLKLAGDVESNPG peptide to the C-terminal of the upstream protein and a P amino acid residue to the N-terminal of the downstream one (Chng et al., 2015). In our design (Figure 1), the adding of the 2A peptide to the C-terminal of the HC would abolish the KDEL signal, which only functions when present at the C-terminal of a protein. The loss of KDEL signal in upstream protein can be solved by adding a furin protease recognition site upstream of 2A, which will remove the residues of the 2A peptide (Chng et al., 2015). Although this strategy can solve the ER retention problem in future studies, the retention of recombinant mAb in ER with the KDEL signal will limit the protein secretion and invariably increase the downstream process cost. Thus, an alternative approach by glycoengineering to silence or knockout the plant-specific glycosylation genes and to introduce mammalian glycosylation genes into plants could be a better solution (Strasser et al., 2008, 2014; Castilho and Steinkellner, 2012).

Glycosylation affects the antibody-dependent cell-mediated cytotoxicity (ADCC) of recombinant mAbs. For example, fucose-free mAb (13F6) produced in glycoengineered tobaccos was demonstrated to have higher binding affinity to Fc_γR and enhanced potency against Ebola virus (Zeitlin et al., 2011). However, Bevacizumab functions through the binding and neutralization of secreted VEGF, but lacks the ADCC (Wang et al., 2004). The biological activity of rice-derived Bevacizumab mAb (including KDEL tagged and non-tagged) was analyzed by an ELISA-based binding assay to its hVEGF antigen (Figure 5). No significant difference was observed between the two rice expressed mAbs and between the rice expressed mAb and the commercial Bevacizumab control, indicating that the rice expressed Bevacizumab mAb is functional. It remains unknown if the rice expressed Bevacizumab mAb have other comparable properties such as the neutralization of hVEGF in vivo, the circular half-life in clinical studies, and most importantly the adverse immunogenicity related to plant expressed recombinant proteins, which need to be characterized in future studies.

In conclusion, we reported an efficient recombinant mAb expression system in transgenic rice callus using the FMDV 2A self-cleavage peptide. The cleavage of 2A peptide linked polyprotein in one gene expression cassette produced equal molecules of HC and LC of a mAb, allowed efficient assembly and folding of functional antibody, and produced recombinant mAb at high levels.

WY and DL conceived, designed, and supervised the study, discussed the results, and wrote the manuscript. LC and XY performed the experiments, analyzed the data, and wrote the manuscript. All authors approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.