Thanks to their potential to bind virtually all types of molecules; monoclonal antibodies are in increasing demand as therapeutics and diagnostics. To overcome the overloading of current production facilities, alternative expression systems have been developed, of which plants appear the most promising. In this review, we focus on the expression of monoclonal IgG or IgM in plant species. We analyse the data for 32 different antibodies expressed in various ways, differing in DNA construction, transformation method, signal peptide source, presence or absence of an endoplasmic reticulum retention sequence, host species and the organs tested, together resulting in 98 reported combinations. A large heterogeneity is found in the quantity and quality of the antibody produced. We discuss in more detail the strategy used to express both chains, the nature of the transcription promoters, subcellular localization and unintended proteolysis, when encountered.
Since the establishment of monoclonal antibody production (Kohler and Milstein, 1975), antibodies have become important tools in research, as well as in therapy and diagnostics. For human therapy, there were initial limitations because of the non-human origin of monoclonal antibodies. Rodent antibodies are immunogenic in humans and show variable ability to initiate antibody-dependent cellular cytotoxicity or complement-dependent cytotoxicity in humans. This problem was overcome thanks to molecular engineering and the progressive introduction of chimeric (Morrison et al., 1984) and humanized (Jones et al., 1986) antibodies. Besides full antibodies, antibody derivatives, such as single-chain, single-domain or bispecific antibodies, minibodies as well as protein- or toxin-conjugated antibodies were also developed (Jain et al., 2007).
In view of the growing demand for recombinant antibodies, existing facilities might rapidly reach saturation, so there is a need for the development of alternative means of production to the classical mammalian expression systems (Stoger et al., 2005b). Because of their complex glycosylation and their assembly from four polypeptides linked by disulphide bridges, antibodies require an appropriate expression host. Plants have appeared as an interesting alternative, even though mammalian and plant-type glycosylations are not identical. Two advantages often mentioned for plant-produced antibodies are their lower cost of production compared to mammalian cells and, for therapeutic purposes, the absence of contamination by mammalian pathogens. Concerning the former, it is not obvious that this statement is true, considering the relatively low expression of antibodies in most plant systems described so far. However, recent progress in expression systems (e.g. transient expression, see in the following paragraphs) might change this view. Moreover, an advantage of plant-produced antibodies might be the possibility of their glycoengineering and, as a result, the production of improved pharmaceutical proteins (bio-betters). Furthermore, if whole transgenic plants are used instead of culture cells, antibody production is only limited by land area, and harvesting can be performed using existing agronomical facilities (Schillberg et al., 2003).
Since the first report of antibody production in N. tabacum plants (Hiatt et al., 1989), antibodies have been expressed in moss (for review, see Decker and Reski, 2008), algae (for review, see Franklin and Mayfield, 2005) and various dicot and monocot species (Table 1 and references herein, Stoger et al. (2005a)). Different expression strategies have been successfully used, i.e. transient or stable expression and the use of vector-dependent (Agrobacterium or viral) or vector-free transformation. However, the wide experience gained of expression of antibodies in plants indicates that there are still problems regarding expression, subcellular localization and proteolytic degradation. Close inspection of the literature shows that these problems vary depending on the antibody, host system or expression system used. In this review, we compare the data obtained for 32 different IgGs or IgMs expressed in either monocot or dicot species. We focused on whole antibodies because they require expression of two genes, an additional problem compared to the expression of single-chain fragments. The majority were expressed in various ways, differing in the gene construction and transformation method used, the nature of the signal peptide, the presence or absence of an endoplasmic reticulum (ER) retention sequence, the host species and the organs tested, resulting in 98 reported combinations (Table 1). Using these data, we discuss in more detail the type of molecular constructs used, the nature of the transcription promoters, subcellular localization and unintended proteolysis, when encountered.
Table 1. Expression of antibodies in plants and the strategy used
Signal peptide/sequence tag H‡
Signal peptide/sequence tag L‡
*m, murine; h, human; κ, kappa light chain; λ, lambda light chain; chimeric: var cγ1 cα2 cα3, chimeric antibody containing a mixed gamma/alpha heavy chain;
†H, heavy chain; L, light chain; when a single promoter is mentioned, it was used to drive transcription of both heavy and light chain genes; CaMV 35S, cauliflower mosaic virus 35S; CaMV 35S+, cauliflower mosaic virus 35S with duplicated enhancer (Kay et al., 1987); NOS, nopaline synthase gene of Agrobacterium tumefaciens; mas1’2’, mannopine synthase genes 1 and 2 of Agrobacterium tumefaciens (organized as divergent transcription units) (Velten et al., 1984); TMV, tobacco mosaic virus; CPMV, cowpea mosaic virus; Pin2, potato proteinase inhibitor II gene (Korth and Dixon, 1997); (OCS)3Mas promoter, promoter of the mannopine synthase gene of Agrobacterium tumefaciens reinforced by a trimer of the upstream activating sequence of the octopine synthase gene (also from Agrobacterium tumefaciens) (Gelvin et al., 1999); 7S, α’ subunit of β-conglycinin (7S) soybean storage protein promoter (Fujiwara et al., 1992); Pact, Arabidopsis thaliana actin 2 promoter; PVX, potato virus X; gt-1, rice glutelin-1 promoter; En2pPMA4, promoter of the Nicotiana plumbaginifolia plasma membrane ATPase4 gene reinforced by two copies of the cauliflower mosaic virus 35S enhancer (Zhao et al., 1999);
‡When a single signal peptide is mentioned, it was used to drive targeting of both heavy and light chains; Sequence tag: KDEL, c-myc, His6, TM (transmembrane sequence), ELP (elastin-like peptides);
§Single T-DNA, T-DNA bearing both heavy and light chain coding sequences; T-DNA H, T-DNA bearing the heavy chain coding sequence; T-DNA L, T-DNA bearing the light chain coding sequence; the arrows refer to the relative orientation of the transgenes; TMV, tobacco mosaic virus; CPMV delRNA-2, cowpea mosaic virus RNA-2 with the region encoding the movement protein and both coat proteins deleted (Canizares et al., 2006); (tr.), transient expression; CP coat protein; CPMV delHTRNA-2, hypertranslatable cowpea mosaic virus RNA-2 lacking AUG 161 and with the region encoding the movement protein and both coat proteins deleted;
¶TP, total proteins; TSP, total soluble proteins; ab, antibody; total ab, antibody extracted from both the biological material and the culture medium; d, day; DW, dry weight; FW, fresh weight; F0, transformed plant; F1 (F3): first (third) generation obtained after sexual crossing; BBI, Bowman–Birk serine peptidases inhibitor.
Because antibodies contain two distinct polypeptides, two possible insertion strategies are conceivable, the insertion of a single DNA construct (usually a T-DNA insert) bearing both the heavy and light chain coding sequences (irrespective of their relative arrangement) or two independent insertion events. In the latter, two different approaches have been used: 1 - the generation of independent plants, each containing one transgene (heavy or light chain cassette) followed by cross-pollination of individuals to obtain plants expressing both the heavy and light chains (described hereafter as the cross-pollination method); 2 - the concomitant co-transformation with both DNA constructs. Concomitant co-transformation with two different plasmids resulted, in some cases, in the insertion of both transgenes in the same locus and, sometimes, in complex insertion patterns involving transgene repeats (De Block and Debrouwer, 1991; De Neve et al., 1997, 1999; Ramessar et al., 2008). This might thus be considered as a single DNA construct-like configuration. It is rather difficult to draw any clear conclusion about the best transgene(s) insertion strategy (Table 1), because the data were obtained for different antibodies, different host plants and different vectors. However, Bouquin et al. (2002) compared the expression of GAN4B.5 using either transformation with a single T-DNA bearing both the heavy and light chain coding sequences in tandem or the cross-pollination method, and, having tested approximately 20 homozygous lines for each construct, found better expression with the latter (Table 2). Other reports showed that the accumulation of mRNAs for heavy and light chains occurred independently, even though they were controlled by identical regulatory elements (Voss et al., 1995; Law et al., 2006) and within a single T-DNA (Voss et al., 1995). It is therefore unlikely to find a plant in which both transgenes are transcribed at a particularly high level. In conclusion, cross-pollination seems to be the most promising method. However, it was not used in most cases reported (Figure 1a), probably because it is more time-consuming. Furthermore, it cannot be used when suspension cells or vegetatively-reproduced plants (e.g. some varieties of potato) are transformed.
Table 2. Influence of the transgene insertion strategy on the expression level
Expression level (% TSP†)
*Single T-DNA, T-DNA bearing both heavy and light chain coding sequences; T-DNA H, T-DNA bearing the heavy chain coding sequence; T-DNA L, T-DNA bearing the light chain coding sequence;
Most of the single T-DNA constructs tested so far were in the tandem orientation. However, although only poorly investigated, the relative arrangement of the two transgenes seems to be of upmost importance (Table 2). The expression of antibody 21C5 was nearly twice as high in lines in which the two transgenes were arranged in a divergent orientation than in lines in which they were in tandem. This difference could not be attributed to the influence of neighbouring plant DNA, as the same result was observed in transient expression assays (van Engelen et al., 1994). The authors suggested that the two cauliflower mosaic virus (CaMV) 35S promoters used in this case acted as transcription enhancers for each other in a distance-dependent way, as shown by Kay et al. (1987). In another study, expression of the antibody LO-BM2 using an inverted convergent configuration of the transgenes was very poor compared to that using the tandem configuration (De Muynck et al., 2009). These authors hypothesized that leaky transcription termination would result in the synthesis of transcripts encompassing both genes, thus generating double-stranded RNA (dsRNA), which might catalyse RNA silencing (De Wilde et al., 2000; De Muynck et al., 2009). Note that the inverted convergent orientation does not always lead to poor results, as exemplified by the yield of 8.2 μg antibody/g root dry weight per day obtained by Komarnytsky et al. (2006). However, in the transformation vector used in their study, the heavy and light chain genes were separated by root proliferation genes. An inverted divergent orientation of two reporter genes had previously been shown to be a better arrangement than an inverted convergent orientation (Padidam and Cao, 2001). The authors attributed this result to transcriptional interference, in which RNA polymerases recruited by the two promoters interfere, because of leaky terminators (Shearwin et al., 2005). Given these results, the orientation of the selectable marker relative to the other transgene(s) should also be taken into account.
While obtaining high expression levels during plant screening is an important factor, maintaining it during subsequent generations is also critical. De Neve et al. (1999) studied instability of antibody production in connection with gene silencing. In particular, they showed that, in three lines obtained after co-transformation and showing transcriptional gene silencing, the transgenes were linked to each other at a single locus and oriented convergently. They also showed that, in two lines in which the transgenes were located at different loci, post-transcriptional gene silencing was somehow triggered during development by the presence of the light chain gene insert. Another striking observation is that plants transformed using Agrobacterium sometimes display sequences originating from outside the right and left borders of the binary vector (Kononov et al., 1997; De Wilde et al., 2000). Some authors (Iglesias et al., 1997; Matzke and Matzke, 1998; De Wilde et al., 2000; Agrawal et al., 2005) suggested that vector backbone sequences, because of their prokaryotic origin (and thus their different nucleotide composition compared to plant DNA), may trigger de novo DNA methylation, resulting in transgene silencing. According to Matzke and Matzke (1998), the apparent tendency of multicopy transgene loci to become silenced might actually reflect a higher probability that binary vector sequences outside the T-DNA are present in complex inserts. An alternative method for overcoming this phenomenon might be biolistic plant transformation using minimal cassettes (i.e. promoter, transgene and terminator in the form of linear DNA, following restriction digestion of the plasmid) (Altpeter et al., 2005).
Whatever the transformation method used, a key feature for high antibody expression might be a clean transformation, i.e. irrespective of the copy number or repeated elements, a transformation event in which the transferred sequences are intact, not rearranged and free of vector backbone (Agrawal et al., 2005). Moreover, Iglesias et al. (1997) showed that out of four independent transgenic tobacco lines, the two that displayed stable expression in the homozygous state over several generations had their insert in the vicinity of the telomeres, whereas, in the other two, the insert was in intercalary and paracentromeric locations. Although the number of transgenic plants studied was not sufficient to draw a general conclusion, this observation underlines the necessity for a detailed study of the insertion locus.
The choice of transcription promoters to express heteromultimeric proteins, such as antibodies, has to take into account not only the high expression level expected, but also the coordinated expression of both the heavy and light chain genes. This latter point explains the widespread use of identical promoters to drive heavy and light chain gene transcription (see Table 1 and Figure 1c). Nevertheless, one should carefully weigh the need for the coordinated expression of both genes against the risk of homology-dependent gene silencing inherent in the use of identical regulatory elements. van Engelen et al. (1994) observed the production of similar amounts of 21C5 heavy and light chain mRNAs using two different promoters in the same T-DNA (see Table 1). Furthermore, they showed that levels of heavy and light chain mRNAs covaried in different independent transformants. In contrast, Voss et al. (1995) and Law et al. (2006) reported different levels of heavy and light chains mRNAs despite using the same transcription promoter. These conflicting results highlight the difficulty in guaranteeing the coordinated expression of the two genes.
Various strategies have been used to further improve the final yield. These include adding enhancers of transcription [e.g. the amplification-promoting sequence isolated from the non-transcribed spacer region of tobacco ribosomal DNA (Borisjuk et al., 2000)] or translation (e.g. the Ω translational enhancer sequence from tobacco mosaic virus) and/or mRNA-stabilizing 5′ and 3′ untranslated regions (UTRs) (e.g. those from the chalcone synthase gene). This latter strategy not only increases the transcription rate, but also favours correct mRNA processing, resulting in increased mRNA stability (Streatfield, 2007).
It is noteworthy that only thirteen different promoters were used to drive heavy and light chain gene transcription of the different antibodies listed in Table 1. This probably reflects the low number of strong plant transcription promoters available. Furthermore, of these thirteen promoters, only six are directly derived from plants, the others being derived from either Agrobacterium tumefaciens T-DNA genes or the CaMV 35S transcript, by far the most widely used (see Figure 1c). Of special interest is the Medicago sativa plastocyanin promoter used by Vezina et al. (2009) to transiently express the C5-1 antibody in Nicotiana benthamiana leaves. Plastocyanin is a protein involved in electron transfer during photosynthesis and is therefore highly expressed in the palisade layer and spongy parenchyma, where initial agroinfection takes place during transient expression. Using this promoter and the corresponding 5′ and 3′ UTRs to drive heavy and light chain transcription in a tandem cassette, as much as 757 μg antibody/g leaves fresh weight (FW) was obtained [antibody retained in the ER and using the co-expressed silencing suppressor HcPro (Vezina et al., 2009)]. Seed-specific promoters offer an interesting option, as seeds are generally considered a stable storage compartment for recombinant antibodies. This is particularly the case if accumulation occurs just prior to desiccation (Streatfield, 2007). Moreover, the seed endosperm has a triploid genome, thus increasing transgene dosage (zygosity level) and thus the antibody content in mature seeds, as reported by Law et al. (2006). In contrast, the number of inserted transgenes following a single transformation event did not correlate with antibody levels (Law et al., 2006). Inducible transcription promoters might also be useful in preventing toxicity of antibody expression (although, to our knowledge, this has never been reported) or transgene silencing. In this respect, the development of new constitutive and inducible strong promoters would probably represent important progress in heterologous expression.
Besides the transcription rate, RNA stability is a major factor in ensuring high expression (Green, 1993). Much less is known about the role of the 5′ and 3′ UTRs. Although such regions taken from plant transcripts have been added to recombinant antibody-coding sequences, their exact role is still unclear, especially because the replacement of the original plant-coding sequences by antibody-coding sequences might considerably modify the 3-D structure of these chimeric transcripts and lead to the loss of the potential positive effects of these UTRs. Although not related to transcription or transcript stability, we should also mention the growing interest in optimizing protein translation. Optimization of the translation initiation codon context is easy, as it involves only a few nucleotides. On the other hand, adapting the whole coding sequence to the host codon usage is much more time-consuming and expensive. Interestingly, Batard et al. (2000) improved the expression of a P450 oxygenase by limiting codon optimization to the first 40 amino acid region. This strategy allowed a fivefold increase in enzymatic activity. Nevertheless, if optimization of the whole sequence is preferred, putative cis-acting sequence motifs (e.g. cryptic splice sites, RNA-destabilizing sequence elements) can be removed at the same time as the codon bias is adapted.
Transient expression classically consists in the Agrobacterium tumefaciens-mediated delivery of binary transformation vector into the plant cell following leaf infiltration. This method was generally used either to test the efficacy of an expression cassette or to produce small amount of proteins. Nevertheless, recent progress such as the development of new promoters (e.g. Vezina et al. (2009), mentioned earlier and Table 1) or the introduction of viral vector-based expression systems have revolutionized the field of transient expression. This method now allows the rapid and high-yield production of heterologous proteins.
Viral vector-based expression systems were developed to express multimeric proteins, such as antibodies (see Table 1 and Figure 1a,c). Initial attempts were made by co-infecting N. benthamiana plants with in vitro synthesized transcripts of two recombinant tobacco mosaic viruses (TMV) modified to encode either the antibody heavy or light chain (Verch et al., 1998). However, Giritch et al. (2006) showed that using two TMV-based vectors to express two different proteins rapidly led to a spatial separation of the two distinct TMV populations in the infiltrated tissues. To overcome this limitation, Giritch et al. (2006) used non-competing viral vectors [TMV and potato virus X (PVX)] to drive antibody expression. This innovative method was used together with magnifection technology (Marillonnet et al., 2005), which is based on the replication of viral vectors delivered to the plant by Agrobacterium tumefaciens infiltration. In this case, the plant promoter placed at the 5′ end of the modified viral genome mainly acts to produce the first viral-like mRNA, as the subsequent replication steps are carried out by the viral RNA-dependant RNA polymerase. Sainsbury et al. (2008) took advantage of the bipartite RNA cowpea mosaic virus (CPMV) to express the C5-1 antibody via agroinfection. In this system, the RNA-1 vector encoding the replicase was co-infiltrated with two modified RNA-2 vectors encoding either the heavy or light chain. While the infiltrated tissues showed co-expression of both chains, the modified RNA-2 molecules segregated during the systemic spread of CPMV. As the advantage of systemic spread was lost, a deleted version of RNA-2 (delRNA-2), in which the region encoding the movement protein and both coat proteins had been removed, but the elements needed for RNA-1-mediated replication were retained, was tested to allow the cloning of larger inserts and avoid the bio-containment problem. Different combinations of full length and deleted versions of RNA-2 molecules encoding heavy and light chains were used to express C5-1 and, in all cases, a higher yield (up to 1.9% of the total soluble protein, corresponding to 74 μg antibody/g FW) was obtained with the deleted version (see Table 1) (Sainsbury et al., 2008). In a more recent version of this method, Sainsbury and Lomonossoff (2008) developed a delRNA-2-based expression system without using RNA-1-mediated replication. Transcription is therefore under the control of a plant promoter, and high expression is conferred by the viral 5′ and 3′ UTR. This vector was further improved by deleting an in-frame AUG codon [referred to as hypertranslatable deleted RNA-2 (delHTRNA-2)]. Up to 325 μg antibody/g FW was obtained (ER-accumulated antibody). Note that, in the CPMV-based expression system reported earlier, a silencing suppressor was co-expressed.
In mammals, both the heavy and light chains are synthesized with a peptide signal, which targets the nascent proteins to the ER lumen, where the antibody is assembled. Complex glycosylation takes place in the ER and the Golgi before the antibody is secreted. Plant and animal glycosylations are similar in the early steps but differ in the later steps. However, plants have been genetically engineered to mimic the typical animal glycosylation pattern and so prevent potential side effects and rapid clearance from the blood stream of plant-made pharmaceuticals (Saint-Jore-Dupas et al., 2007). In recent years, glycoengineering has been addressed in two different ways: inactivating endogenous glycosyltransferases and/or expressing heterologous glycosyltransferases.
Koprivova et al. (2004) disrupted the genes for α1,3-fucosyltransferase (Fuc-T) and β1,2-xylosyltransferase (Xyl-T) in Physcomitrella patens by homologous recombination. Plant growth and morphology were not impaired. Mass spectrometry analysis of N-glycans from the double knockout plant showed complete absence of α1,3-fucosyl and β1,2-xylosyl residues. Moreover, secretion of the human glycosylated growth factor VEGF121 was shown to be as effective in double knockout than in WT plants.
Cox et al. (2006) used a single T-DNA construct to stably express an antibody and an RNA interference cassette aiming at silencing both Fuc-T and Xyl-T in the aquatic plant Lemna minor. GnGn was shown to make up 95.8% of the total N-glycans harboured by the antibody and no α1,3-fucosyl and β1,2-xylosyl residues were found. The high glycosylation homogeneity offered by this system undoubtedly constitutes an advantage compared to cultured mammalian cells that generally produce heterogeneous N-glycosylation. Moreover, the plant glycoengineered antibody was shown to exhibit higher binding affinity for human Fc receptors, resulting in a 20- to 35-fold increase in biological activity compared to its CHO-produced counterpart.
Instead of inactivating Fuc-T and Xyl-T expression, Vezina et al. (2009) rather expressed a chimeric form of the human β1,4-galactosyltransferase (Gal-T). The chimeric protein consisted in a protein fusion between the N-terminal part (cytosolic tail and transmembrane domain) of A. thaliana N-acetylglucosaminyltransferase I (GNTI) and the catalytic domain of human Gal-T. The transmembrane and cytosolic domains of GNTI allowed targeting of the Gal-T activity in the ER and the cis-Golgi apparatus, i.e. upstream from Fuc-T and Xyl-T activity in the plant secretory pathway. Transient co-expression of this chimeric construct together with genes encoding C5-1 antibody heavy and light chains in N. benthamiana leaves resulted in high-yield production of the antibody containing no detectable (<1%) α1,3-fucose and β1,2-xylose.
Beside sensu stricto plant glycoengineering, other strategies were used to alleviate plant-made N-glycosylation limitations in antibody production. These include mutation of N-glycosylation sites to produce aglycosylated antibodies (Rodriguez et al., 2005) and the addition of ER retention signal to prevent the addition of complex-type glycans (see Table 1). Nevertheless, both methods are not desirable as N-glycans structure is determinant in antibody functional activity and half-life (Saint-Jore-Dupas et al., 2007).
To conclude this part, it is clear that glycosylation engineering is effective in adapting plant-made antibodies to therapeutic applications.
In the first attempt to express an antibody in plants, several different constructs were tested: one in which both the heavy and light chain cDNAs retained their native signal peptide sequence, one in which this sequence was removed from both cDNAs and two other combinations in which only one of the two chains retained its signal peptide (Hiatt et al., 1989; Hein et al., 1991). The results showed that, as in mammalian systems, both the heavy and light chains had to be directed to the secretory pathway to allow antibody assembly and accumulation. It was also shown that the native signal peptide could be replaced by a plant signal peptide (During et al., 1990) (i.e. that from barley α-amylase) or a yeast signal peptide (Hein et al., 1991) [i.e. the pre-pro sequence from the Saccharomyces cerevisiaeα-mating factor (Kurjan and Herskowitz, 1982)] without affecting antibody secretion in N. tabacum.
Once in the secretory pathway, antibodies can be retained in the ER if they are provided with a KDEL or HDEL retention signal. ER sequestration generally allows increased antibody accumulation and also prevents the addition of non-mammalian glycosylation types added in the late-Golgi apparatus which might promote an immune response if the antibody had to be used parenterally. However, ER-retained antibodies end up with high mannose structures which are not desirable from a pharmacokinetic viewpoint. Furthermore, in some cases, antibodies without a retention signal have been found in intracellular compartments. We will restrict our discussion to antibodies for which an in situ subcellular localization has been reported (Table 3). We must stress that only antibodies retaining their antigenic epitopes can be identified in this way and that a fraction of the proteins might be partly degraded in some compartments and so escape detection.
Table 3. Subcellular localization of antibodies expressed in plants
CaMV 35S+, Cauliflower Mosaic Virus 35S promoter with duplicated enhancer; En2pPMA4, promoter of the Nicotiana plumbaginifolia plasma membrane ATPase4 gene reinforced by two copies of the cauliflower mosaic virus 35S enhancer; mas1’2’, promoters of the mannopine synthase genes 1 and 2 of Agrobacterium tumefaciens (organized as divergent transcription units); mas2’, promoter of the mannopine synthase genes 2 of Agrobacterium tumefaciens; CaMV 35S, Cauliflower Mosaic Virus 35S promoter; 7S, α’ subunit of β-conglycinin (7S) soybean storage protein promoter; gt-1, rice glutelin-1 promoter.
The subcellular localization of an antibody (B1-8) expressed in plants was first determined by During et al. (1990) using immunogold labelling and electron microscopy. This analysis was performed on callus and stem tissues generated from transgenic tobacco plants by wounding and hormone application. Both chains were provided with the signal peptide of barley α-amylase. The free light chain was found in the cytosol, while the assembled antibody was found in the ER and also in the chloroplasts. A chloroplast localization does not seem to have been reported for any other antibody.
Rademacher et al. (2008) showed, by immunofluorescence and electron microscopy, that 2G12, a human IgG1, accumulated in the prolamin bodies of maize endosperm when placed under the control of the gt-1 promoter and with both chains retaining their native signal peptide and each fused to a SEKDEL ER retention sequence. This result is in agreement with the facts that prolamin bodies are derived from the ER (Chrispeels and Herman, 2000) and that, in maize, prolamins are retained within the ER lumen (Shewry et al., 1995).
Petruccelli et al. (2006) assessed, by immunogold labelling and electron microscopy, the accumulation in leaves or seeds of N. tabacum of 14D9, a mouse IgG1, in which both chains retained their native mouse signal sequence, under the control of two different promoters [CaMV 35S+, i.e. the cauliflower mosaic virus 35S promoter with duplicated enhancer (Kay et al., 1987) or 7S, i.e. the promoter for the gene coding for the α’ subunit of the β-conglycinin (7S) soybean storage protein] and with or without an ER retention signal. When the heavy and light chains were under the control of the CaMV 35S+ promoter and fused to a SEKDEL ER retention signal, 14D9 accumulated in the ER lumen in leaf tissues and in both the protein storage vacuole (PSV) matrix and the apoplast in seeds (Table 3). A similar construct in which the CaMV 35S+ promoter was replaced by the 7S promoter gave identical results, with an accumulation of 14D9 in the seed PSV matrix and apoplast. When both the heavy and light chains lacked an ER retention signal, 14D9 still accumulated in the PSV matrix of the seeds, irrespective of the promoter used. In leaves, the construct with the CaMV 35S+ promoter and without any retention signal on the heavy and light chains led to 14D9 accumulation in the apoplast (note that this result was not assessed by microscopy, but inferred from the glycosylation pattern). Together, these results indicate that the ER retention signal was efficient in leaves, but did not prevent the transport of 14D9 to later Golgi compartments in the seeds of N. tabacum, where it was found in the PSV and apoplast (Petruccelli et al., 2006). It is noteworthy that these authors also showed that the SEKDEL-bearing antibody was immunogenic in BALB/c mice, whereas the secreted antibody was not. As BALB/c mice are not immuno-sensitive to plant glycosylation, this suggests that the SEKDEL tag itself and/or an antibody conformational change induced by the typical ER glycosylation pattern may be responsible for this immunoreactivity (Petruccelli et al., 2006).
Vine et al. (2001) expressed a membrane-bound variant of the Guy’s 13 mouse antibody in N. tabacum leaves. Both the heavy and light chains retained their native murine signal peptide. Only the heavy chain contained a native transmembrane sequence and this was sufficient to obtain a membrane-bound assembled antibody, as both chains were co-localized at the cell surface.
De Muynck et al. (2009) expressed LO-BM2, a chimeric rat/human antibody, in both N. tabacum plants and BY-2 cells. Both the heavy and light chains retained their native (rat) signal peptide. The authors reported that, in BY-2 cells, LO-BM2 was localized to the ER, while, in N. tabacum plants, it was localized to a post-Golgi compartment, most probably a pre-vacuolar compartment, in leaf protoplasts and epidermal cells. This unexpected localization was not totally a surprise, because Irons et al. (2008), using fluorescent reporter proteins fused to the light and heavy chains, showed that an antibody transiently expressed in leaf epidermal cells was located in punctuate structures in close association with the ER and partly overlapping with a pre-vacuolar compartment marker. In BY-2 cells, the ER-localized LO-BM2 antibody, which was expected to be secreted, might represent only part of the total amount synthesized. Indeed, De Muynck et al. (2009) showed that antibody in the extracellular medium was resistant to endoglycosidase H, indicating that it bears complex glycans typical of the late-Golgi. This strongly suggests that at least part of the antibody was secreted.
From this survey, it is clear that some antibodies were found to be localized in the expected subcellular compartment, while others seemed to be, in part, diverted from the expected accumulation site (During et al., 1990; Petruccelli et al., 2006; De Muynck et al., 2009) or at least delayed in their transport (De Muynck et al., 2009) (Table 3). Moreover, the very same antibody, encoded by the same transgene construct, could end up in different subcellular localizations, depending on the tissue in which it is expressed (Petruccelli et al., 2006; De Muynck et al., 2009). At this stage, it is difficult to propose any coherent explanation for these variable results, because they were obtained with different antibodies, different signal peptides, different transcription promoters and different host species and analysed in different tissues (see Figure 1b, c and d). From a practical point of view, heterogeneous subcellular localization might result in heterogeneity and/or reduced recovery of the final product, which represents a problem for the industrial production of large quantities of homogenous antibodies. Thus, effort should be directed towards a better understanding of the trafficking of heterologous proteins in plants to improve this limiting step.
In conclusion, although it is possible to address antibodies to different subcellular localizations, partial mistargeting sometimes occurs, which might result in antibody structural heterogeneity. As discussed later, choosing the final destination should also take into consideration costs of processing as well as the purpose (e.g. therapeutics, diagnostics).
A major drawback to antibody production in plants is the loss of material because of proteolytic degradation. This problem is not specific to this protein-host combination, because it is typically found in many cases of heterologous expression in various organisms. The presence of additional and smaller than expected antibody fragments can be noted in the reports on 22 of the 32 different antibodies listed in Table 1 (for the other 10 antibodies, smaller bands were not reported or no electrophoretic analysis was provided). It is noteworthy that this occurred in all of the analysed tissues (leaf, callus, tuber, hairy root, suspension cell, shooty teratoma, root and seed). As reported by Sharp and Doran (2001a), antibody fragments in plants have been variously explained as antibody assembly intermediates or the consequence of extracellular peptidase activity after secretion or the activity of peptidases released during sample homogenization. Although it is likely that all three phenomena occur in some cases, it has become clear that extracellular peptidase activity is a major factor to consider in antibody production in plants (van Engelen et al., 1994; Sharp and Doran, 2001a,b; Komarnytsky et al., 2006; De Muynck et al., 2009).
Despite the numerous reports on antibody degradation in plants (Table 4), it is not possible to draw any general rules concerning the degradation pattern. This is mainly because of the diversity of the host species and antibodies tested. De Neve et al. (1993), for instance, showed that MAK33 antibody displayed many more degradation products when expressed in N. tabacum than in A. thaliana. Furthermore, it is likely that fragments resulting from a single cleavage might be further processed by exopeptidases into smaller fragments. It is also important to bear in mind that the detection of antibodies by Western blotting may not reveal all generated fragments, depending on the antibody used for detection, whether antigenic epitopes are still present in the fragments and whether the fragment size is large enough to be seen on SDS-PAGE. It is therefore possible that degradation is largely underestimated in some cases.
Table 4. Putative degradation products noted on PAGE of plant-produced antibodies
Organ/ER retention signal (X)
Additional bands under non-reducing conditions
Reactivity of additional bands (non-reducing conditions)
Total antibody levels increased by up to 90% with 0.1% (w/v) nitrate, 4 mg antibody/L in the medium Total antibody levels increased by 20% to 90% in cultures with PVP, 11 mg antibody/L in the medium supplemented with 2 g/L PVP Total antibody levels increased by 14% to 68% with gelatin, 9.5 mg antibody/L in the medium supplemented with 9 g/L gelatin
Recognized by anti-mouse antibodies (suggested to be assembly intermediates or antibody fragments)
90 kDa (biomass and medium)
120 kDa (biomass and medium)
Recognized by anti-mouse antibodies (suggested to be assembly intermediates)
140 kDa (biomass and medium)
Hairy root cultures
40 kDa (es) (biomass)
Recognized by anti-mouse IgG, anti-mouse Fab, anti-mouse IgG-gamma1 heavy chain, and anti-mouse kappa light chain antibodies
33 kDa (biomass)
Recognized by anti-mouse IgG and anti-mouse IgG-gamma1 heavy chain antibodies
Medium supplemented with 1.5 g/L PVP: prevents antibody loss at the end of the culture, maximum total antibody levels of 4% TSP, maximum antibody accumulation in the medium ± 4 times higher than in controls Periodic antibody recovery from the medium using hydroxyapatite: maximum total antibody levels 20% higher than without 150% air saturation in a recirculation bioreactor: biomass antibody content 52% higher at 1.6 mg/g DW (negligible amount in the medium)
Recognized by anti-mouse IgG, anti-mouse Fab, anti-mouse IgG-gamma1 heavy chain, and anti-mouse kappa light chain antibodies, endo H-resistant
80 kDa (biomass and medium)
43 kDa (biomass and medium)
Recognized by anti-mouse IgG and anti-mouse IgG-gamma1 heavy chain antibodies, and biotinylated concanavalin A
120 kDa (biomass and medium)
Recognized by anti-mouse IgG, anti-mouse Fab, anti-mouse Fc, anti-mouse IgG-gamma1 heavy chain, and anti-mouse kappa light chain antibodies, endo H-sensitive
135 kDa (biomass and medium)
Recognized by anti-mouse IgG, anti-mouse Fab, anti-mouse Fc, anti-mouse IgG gamma1 heavy chain, and anti-mouse kappa light chain antibodies, endo H-sensitive (biomass)
Plant-derived shooty teratoma cultures
50 kDa (biomass)
Recognized by anti-mouse IgG antibodies
43 kDa (biomass)
Recognized by anti-mouse IgG antibodies
120 kDa (biomass)
135 kDa (biomass)
135 kDa (biomass and medium)
Recognized by anti-mouse IgG antibodies, reduced intensity after BFA treatment (performed only on biomass), endo H-sensitive (biomass, medium), unaffected by swainsonine (performed on biomass only), unaffected by benzylamine
Medium supplemented with 1.5 g/L PVP: maximum total antibody levels of 12% TSP, maximum antibody accumulation in the medium 2.5 times higher than in controls Negative effect of tunicamycin (10 μg/mL) on antibody content (biomass −35%, medium −49%) No significant effect of castanospermine (100 μg/mL) on antibody levels in the biomass, antibody content 74% lower in the medium of treated cells compared to the controls Medium supplemented with BFA (20 μM): antibody levels in the biomass increased to 1.4–2.7 times that in untreated cultures Periodic antibody recovery from the medium using hydroxyapatite: maximum total antibody levels 21% higher with antibody recovery than without
120 kDa (biomass and medium)
Recognized by anti-mouse IgG antibodies, reduced intensity after BFA treatment (performed only on biomass), endo H-sensitive (biomass only, medium resistant), sensitive to swainsonine (performed only on biomass), unaffected by benzylamine
80 kDa (biomass and medium)
Recognized by anti-mouse IgG antibodies, reduced intensity after BFA treatment (performed only on biomass), endo H-resistant, unaffected by benzylamine
50 kDa (medium)
Recognized by anti-mouse IgG antibodies, endo H-resistant, unaffected by benzylamine
Recognized by anti-mouse IgG1 antiserum
Medium supplemented with 8 g/L gelatin: percentage of intact IgG to total antibody about 4 times higher than in controls
Coomassie blue-stained, binds antigen, recognized by anti-human lamda chain antibodies (suggested to be Fab fragment)
25 kDa (p) (es)
Coomassie blue-stained, degradation products of heavy chain occurring around the hinge region, having either a correctly processed C-terminal peptide (the most N-terminal intact peptide DTLMISR) or a correctly processed N-terminal peptide (the most C-terminal intact peptide STSGGTAALGCLVK), together with some minor degradation products of LC
50–100 mg purified antibody/kg FW when co-expressed with p19 silencing suppressor of artichoke mottled crinkle virus (AMCV)
100 kDa (p)
27 kDa (p) (es)
32 kDa (p) (es)
13 kDa (medium)
Recognized by anti-human IgG (Fc-specific) antibodies
Recognized by anti-human IgG (Fc-specific) and anti-human kappa light chain antibodies
19 kDa (es) (medium)
Recognized by anti-human kappa light chain antibodies
35 kDa (es) (biomass)
Recognized by anti-human IgG antibodies (Fc-specific),
35 kDa (p) (biomass)
Coomassie blue-stained, N-terminal residues KTHTCPPCP (Edman degradation performed on purified antibody)
35 kDa (medium)
Recognized by anti-human IgG antibodies (Fc-specific)
39 kDa (es) (biomass)
Recognized by anti-human IgG antibodies (Fc-specific)
39 kDa (p) (es)
13 kDa (apoplast)
Recognized by anti-human IgG (Fc specific) and anti-human kappa light chain antibodies
35 kDa (es) (apoplast)
Recognized by anti-human IgG antibodies (Fc-specific)
In three cases (all human IgG1 antibodies, but expressed in different plant species), degradation fragments of the heavy chain were analysed by either mass spectrometry (MS) (Ramessar et al., 2008; Villani et al., 2009) or Edman degradation (De Muynck et al., 2009) to localize the cleavage(s) site(s). Villani et al. (2009) identified two major degradation products derived from cleavage occurring close to the hinge region. The exact cleavage point could not be precisely determined because a small region on both sides of the hinge was not identified by MS, reflecting either the absence of this region (and thus its degradation by peptidases) or the failure to detect the corresponding peptides by MS. Furthermore, De Muynck et al. (2009) determined the sequence of the N-terminal part of a heavy chain degradation product as KTHTCPPCP, a sequence localized in the hinge region. These results pinpoint the antibody hinge and closely related regions as the most sensitive to proteolytic activity. However, other susceptible regions must exist, because products other than those resulting from cleavage in the hinge region are found. For instance, Ramessar et al. (2008) established a continuous MS peptide map of a heavy chain degradation product encompassing the CH1, hinge, CH2 and CH3 regions.
Sharp and Doran (2001a) extensively studied the degradation patterns of a mouse IgG1 antibody (Guy’s 13) produced in different N. tabacum expression systems and suggested that antibody degradation in plants is likely to occur in the extracellular medium (apoplast or culture medium) and also during secretion, between the ER and the Golgi. However, it is clear (Table 4) that antibodies with an ER retention KDEL sequence are also often subjected to degradation when expressed in plants. This observation might be explained either by the antibody escaping from the ER or, more puzzling, by the fact that the ER might not be as safe a storage compartment for heterologous proteins as generally accepted. A case study is the 14D9 antibody fused to a KDEL sequence and expressed in N. tabacum, which was correctly retained within the ER in leaves, but further transported to protein storage vacuoles in seeds, where it underwent proteolytic cleavage (Petruccelli et al., 2006).
Attempting to overcome the peptidase issue, Komarnytsky et al. (2006) showed that an antibody expressed in N. tabacum was less subject to degradation and showed higher accumulation (a nearly threefold increase) when co-expressed with a secreted form of the Bowman–Birk serine peptidase inhibitor than when co-expressed with a cytosolic form of this inhibitor. It has also been reported that addition of gelatin as a substitution substrate for peptidases increased total antibody levels by 68% in N. tabacum hairy roots (Wongsamuth and Doran, 1997) or by 300% during N. tabacum rhizosecretion (Drake et al., 2003) (Table 4). Furthermore, De Muynck et al. (2009) showed, using zymography, that the extracellular medium of N. tabacum (i.e. culture medium from suspension cells or leaf apoplasm) contains several peptidases able to degrade human IgG in vitro.
To reduce proteolytic processing of plant-produced pharmaceuticals, other strategies were also used such as confining transgene expression to specific tissues, targeting to cellular organelles or fusing stabilizing partners (Benchabane et al., 2008). Stoger et al. (2000) showed that an scFv antibody expressed in seeds of wheat and rice could be stored for at least 6 months at room temperature without significant lost of amount and activity. Similarly, Artsaenko et al. (1998) showed that half of the amount of an scFv antibody expressed in potato tubers remained detectable after 1.5 years of tuber storage at 4 °C and that the specific activity did not decreased during tuber storage. These results highlight the advantage of specific tissues presenting a reduced metabolic activity as storage compartments for recombinant proteins (Benchabane et al., 2008). Recombinant proteins have been expressed in, or targeted to, different subcellular compartments (for review, see Benchabane et al. (2008); Streatfield (2007)). Nevertheless, full-size immunoglobulins need to be addressed to the secretory pathway for assembly and glycan processing. Fusion of recombinant proteins to a stabilizing partner has been shown to be an effective strategy to improve the final yield as well as to facilitate protein recovery. For instance, Floss et al. (2008) showed that fusion of elastin-like peptides (ELPs) to the C-terminal of the 2F5 antibody enhanced its stability while keeping its binding properties unchanged compared to the CHO-produced counterpart lacking ELPs. Nevertheless, if removal of the fusion partner is mandatory for recombinant protein activity, this step might be either performed in vivo taking advantage of identified recognition sites of endogenous peptidases or in vitro, but therefore adding to downstream processing costs (Streatfield, 2007; Benchabane et al., 2008). Gene knockout or silencing of plant peptidases might also serve to increase recombinant protein stability (Streatfield, 2007; Benchabane et al., 2008). This approach is only feasible if the target peptidases are not essential for plant growth. In addition, it only works when antibody is degraded by a single or few peptidases. This might turn out not to be true considering the number of peptidases found, e.g. in the tobacco apoplasm (Delannoy et al., 2008).
Recently, Hassan et al. (2008) have examined different extraction methods for monoclonal antibodies targeted to different subcellular compartments in transgenic N. tabacum plants. Extraction methods were knowingly chosen as simple as possible because they are more likely to be scalable and financially viable at an industrial scale (Hassan et al., 2008). Antibodies were targeted to the apoplasm, within the ER, or, when fused to a membrane span, to the plasma membrane. Their results show that while a grinding step was necessary for ER-retained and membrane-bound antibody recovery, secreted antibody was efficiently recovered by a simple freeze-thaw technique consisting in freezing the leaves at −20 °C and thawing them for 10 min at room temperature before adding an extraction buffer. This technique is also very convenient as the freezing step allows storage of leaves prior to extraction procedure (Hassan et al., 2008). Together with glycosylation considerations that have already been mentioned in this review, this result brings a new and major point in favour of secreted antibodies, particularly if they are expressed for therapeutic purposes. Secreted antibodies are also appropriate for suspension cells secretion or plant rhizosecretion as antibodies can be purified directly from the liquid culture medium without grinding. Nevertheless, purification in those cases might be made difficult depending on the complexity of the medium used and on the endogenous secreted compounds (e.g. carbohydrates). Expression of antibodies in storage organs such as seeds or tubers was shown to be advantageous from a storage point of view (Artsaenko et al., 1998; Stoger et al., 2000). Nevertheless, this strategy would invariably lead to the need of a grinding step that is economically unfavourable, except if recombinant protein accumulation is large enough to compensate for the extra costs of grinding. The high expression levels obtained recently for antibodies transiently expressed in N. benthamiana leaves (Sainsbury and Lomonossoff, 2008; Vezina et al., 2009) suggest that, to date, this system coupled with the extraction procedure described by Hassan et al. (2008) constitute the most promising way for expressing therapeutic antibodies in plants at an industrial scale. It is noteworthy that expression strategies should always be weighted according to the particular use intended for the recombinant protein. For instance, while it is clear that membrane-bound IgG are not recommended for further purification and therapeutic use, they might be particularly well suited for phytoremediation purposes.
A limited number of antibodies expressed in plants have reached the clinical development stage. For instance, CaroRx, an antibody that specifically binds Streptococcus mutans (the bacteria that causes tooth decay) was expressed in N. tabacum (Ma et al., 1994) and is licensed in Europe. In addition, three cases are expected to soon enter Phase I clinical trials: BLX-301, an anti-CD20 optimized antibody for the treatment of non-Hodgkin’s B-cell lymphoma expressed in the aquatic plant Lemna minor (http://www.biolex.com); MAPP66, a combination of several antibodies expressed in N. benthamiana by the MagnIcon technology and used as a HSV/HIV microbicide (http://www.mappbio.com); and 2G12, an HIV-neutralizing antibody expressed in N. tabacum or maize (http://www.pharma-planta.org). There are also other antibodies developed by several companies that are currently being involved in preclinical development. Finally, CIGB is producing in N. tabacum plants the antibodies CB-Hep.1 and HB-01 (http://www.cigb.edu.cu) that have been used for several years in Cuba for the manufacturing process of a Hepatitis B vaccine.
The survey of several publications dealing with antibody expression in plants indicates that this is an appropriate heterologous expression system. We did not discuss the activity of the antibodies produced in plants, because, although binding activity was shown in most cases, there was usually no quantitative comparison of the binding activity of the plant- and animal-produced antibody, and the biological activity of the antibody was seldom addressed. Nevertheless, it is clear that, when activity was reported in the literature, plant-produced antibodies were functional. The data also clearly show important variations in the yield, subcellular localization and proteolytic degradation profile of antibodies. This is, in part, because of the use of different transcription promoters, the presence or absence of retention signals, the transformation of different host species and the analysis of different organs. There are also indications that different antibodies might not behave in the same way. In this respect, comparing the expression of different antibodies in the same expression system would be very informative.
Two recent advances have moved antibody expression closer to industrial application. Indeed, high antibody expression has been obtained recently using transient expression and this system is expected to become predominant. In addition, glycosylation engineering of antibodies has shown to be functional and represents an interesting tool for adapting the antibody structure to therapeutic use.
The drawback of proteolytic degradation is still particularly annoying, as, besides decreasing the yield of functional antibodies, it also necessitates costly steps to remove the partly degraded antibodies. In this respect, it would be interesting to develop extracellular peptidase-free host plants or to engineer antibodies displaying higher resistance to peptidases while maintaining their activity. In addition to the scientific aspects, there are economical considerations that were not covered here, but are of prime interest when moving from the laboratory to the industrial production. Most of the articles dealing with antibody expression in plants do not take into account the costs and pitfalls of downstream processing (For review, see Basaran and Rodriguez-Cerezo (2008); Stoger et al. (2005b)).
Beside full antibodies that were covered in this review, more and more antibody derivatives are considered for therapeutic applications. A nice illustration is the recent production of scFv idiotypic vaccines in Nicotiana benthamiana plants using a viral tobacco mosaic virus vector (McCormick et al., 2003, 2008). This system allowed the production of personalized vaccines (derived from each patient’s tumour) in even less time than required with some animal cell-based system and showed to be effective and safe to administer to patients (McCormick et al., 2008).
Finally, while plant expression systems are expected to be further improved, there are already several plant-produced therapeutic antibodies that are in the pipeline for medical evaluation. This is probably the best indication that plant-produced antibodies have a future.
The work carried out in this laboratory was supported financially by a grant from the European Community (PHARMA-PLANTA integrated Project), the Région Wallonne, the Inter-university Attraction Poles Program-Belgian Science Policy and the Belgian Fund for Scientific Research. BDM was the recipient of a fellowship from the Fonds pour la Formation à la Recherche dans l’Industrie et dans l’Agriculture (Belgium).