R. Fischer, Institut für Biologie I (Botanik/Molekulargenetik), RWTH Aachen, Worringerweg 1, D-52074 Aachen, Germany. Fax: + 49 241871062, Tel.: + 49 241806631, E-mail: email@example.com
We describe the expression of the bispecific antibody biscFv2429 in transgenic suspension culture cells and tobacco plants. biscFv2429 consists of two single-chain antibodies, scFv24 and scFv29, connected by the Trichoderma reesi cellobiohydrolase I linker. biscFv2429 binds two epitopes of tobacco mosaic virus (TMV): the scFv24 domain recognizes neotopes of intact virions, and the scFv29 domain recognizes a cryptotope of the TMV coat protein monomer. biscFv2429 was functionally expressed either in the cytosol (biscFv2429-cyt) or targeted to the apoplast using a murine leader peptide sequence (biscFv2429-apoplast). A third construct contained the C-terminal KDEL sequence for retention in the ER (biscFv2429-KDEL). Levels of cytoplasmic biscFv2429 expression levels were low. The highest levels of antibody expression were for apoplast-targeted biscFv2429-apoplast and ER-retained biscFv2429-KDEL that reached a maximum expression level of 1.65% total soluble protein in transgenic plants. Plant-expressed biscFv2429 retained both epitope specificities, and bispecificity and bivalency were confirmed by ELISA and surface plasmon resonance analysis. This study establishes plant cells as an expression system for bispecific single-chain antibodies for use in medical and biological applications.
Antibodies are essential tools in medicine, biology and biochemistry. Their high affinity and specificity make them invaluable for diagnostic and therapeutic applications in medicine and human healthcare. To broaden the potential use of antibodies in medical and biological applications, bispecific antibody molecules, which have two independent binding sites for two epitopes are often desirable. Bispecific antibody molecules are unique therapeutic agents with their ability to crosslink two different antigens, which can be exploited in cancer therapy for the recruitment of cytotoxic T cells to a tumor cell [1,2].
Various strategies have been used to generate bispecific antibodies. The first bivalent bispecific full-size antibodies or F(ab′)2 fragments were produced by in-vitro chemical cross-linking of two different antibodies [3–5]. In addition, bispecific F(ab′)2 fragments have been created by heterodimerization of Escherichia coli-expressed Fab fragments through cysteine residues  or leucine zippers . Bispecific antibodies have been produced in vivo using the hybrid hybridoma (quadroma) technology , but a limitation of this procedure is the low yield of bispecific antibody of the desired dual specificity. Protein engineering has permitted the design of even smaller bispecific fragments based on single-chain Fv fragments (scFv) fragments, such as the scFv heterodimer diabody, which is formed in vivo by noncovalent association of two single-chain fusion products . Alternatively, the two different binding specificities can be combined in a single polypeptide using a flexible linker peptide to form a bispecific single-chain antibody (biscFv). Functional biscFvs were first expressed in E. coli in 1994 . Since then several biscFv antibody molecules have been successfully expressed in microbes [11–13] and mammalian cells [14,15]. However, expression levels are often low, and expression in E. coli led to the production of inclusion bodies containing misfolded recombinant antibodies which require labor-intensive in-vitro refolding procedures to produce functional biscFv molecules [16,17].
To explore alternative expression systems, we evaluated the production of a biscFv in plant cell suspension culture and in transgenic plants. To date full-size antibodies, Fab-fragments, scFv and heavy-chain variable domains have been successfully expressed in plants, reaching production levels of up to 6.8% of total soluble protein [18–22], but there are no reports on biscFv expression in plants. As a model biscFv for establishing proof of principle, we selected biscFv2429, which contains the well-characterized TMV-specific scFv fragments scFv24 and scFv29, that recognize epitopes present only on the surface of intact TMV particles and TMV coat protein monomers, respectively [23,24]. Our data demonstrate that functional biscFv2429 could be expressed in different plant cell compartments (cytoplasm, apoplast, ER) with a maximum yield of 1.65% of total soluble protein when expressed in the ER of transgenic plants. Affinity-purified biscFv2429 was bispecific for both epitopes, and had similar binding characteristics and affinities compared to its parental single-chain antibodies.
Plasmids, bacteria, suspension cultures, plants
The following plasmids, bacterial strains and plants were used throughout this study: plasmids, pUC18  and pSS ; bacterial strains, E. coli SCS110 (Stratagene, Heidelberg, Germany), Agrobacterium tumefaciens GV3101 (pMP90RK, GmR KmR RifR) ; suspension-culture, Nicotiana tabacum L. cv. bright yellow 2 (BY-2); plants, N. tabacum L. cv. Petite Havana SR1.
Vector design and construction
To combine scFv24 and the CBHI linker with scFv29 in a bispecific single-chain antibody, a cassette arrangement was chosen with restriction sites at the 5′ and 3′ ends of the two scFv and linker sequences. First, the scFv29 was subcloned into the EcoRI and SalI restriction sites of a pUC18 derivate, containing a His6 sequence (pUC18-scFv29-his). The plasmid pML2 (kindly provided by Dr T. Teeri, VTT Biotechnical Laboratory, Espoo, Finland) containing the cDNA of the CBHI-linker was used in conjunction with the forward primer CBH-CLA 5′-GCG GAA TTC GTA ATC GAT CCC GGG GGT AAC CGC GGT ACC-3′ and backward primer CBH-MOD 5′-GCG GAC GTC GCT ATG AGA CTG GGT GGG CCC-3′ to introduce an EcoRI and ClaI (5′ end) or an AatII (3′ end) restriction site(s) (underlined) by PCR. The EcoRI and AatII-restricted PCR fragment was subcloned into pUC18-scFv29-his (CBHI-scFv29-his).
EcoRI and NcoI restriction sites were integrated at the 5′ end of scFv24  by PCR using the primer SCA25: 5′-G CGG AAT TCG GCC ACC ATG GCC CAA ATT GTT CTC ACC CAG TCT-3′ and a 3′ClaI site using the primer SCA26: 5′-GCG ATC GAT TGC AGA GAC AGT GAC CAG AGT-3′. Cloning of the EcoRI–ClaI fragment upstream of the CBHI linker in the vector pUC18-scFv29-his gave the biscFv2429 construct pUC18-biscFv2429.
For targeting biscFv2429 to different plant cell compartments, the 5′EcoRI–StuI fragment of pUC18-biscFv2429, containing the 5′ end of scFv24, was exchanged with its corresponding region from pscFv24CM and pscFv24CW  containing the 5′ UTR of the chalcone synthase (CHS 5′-UT)  and either the original mouse leader sequence of the light chain cDNA or no leader sequence. The subsequent ligation of both EcoRI–XbaI fragments into the plant expression vector pSS, containing a double-enhanced 35S promoter and the cauliflower mosaic virus (CaMV) termination sequence, resulted in the final expression constructs biscFv2429-cyt and biscFv2429-apoplast (Fig. 1), which were used for biscFv2429 expression in the cytoplasm or apoplast, respectively.
A third construct (biscFv2429-KDEL) was generated by replacing the C-terminal His6 sequence of biscFv2429-apoplast with the ER retention signal KDEL, which was introduced by PCR using the primer KDEL: 5′-ACG CTC TAG AGC TCA TCT TTC TCA GAT CCA CGA GAA CCT CCA CCT CCG TCG ACT GCA GAG ACA GTG ACC AGA GTC CC-3′ to generate pUC18-biscFv2429-KDEL. The construct was then subcloned into the pSS plant expression vector .
Transformation of A. tumefaciens, suspension-cultured tobacco cells and tobacco plants
Plant expression constructs were transferred into A. tumefaciens GV3101 by N2 transformation . Transgenic N. tabacum cv. Petite Havana SR1 were generated by the leaf disc transformation method with recombinant agrobacteria and transgenic T0 plants were regenerated from transformed callus . N. tabacum L. cv. BY-2 cells were cultured in Murashige and Skoog basal salt with minimal organics (MSMO+) (S. Schillberg et al., unpublished data). BY-2 cells were transiently or stably transformed by cocultivation with recombinant agrobacteria . Selection of kanamycin-resistant transformants was performed on MSMO+ agar medium supplemented with 75 µg·mL−1 kanamycin and 100 µg·mL−1 claforan.
Generation of anti-mAb24 and anti-mAb29 polyclonal antibodies
Anti-mAb24 and anti-mAb29 polyclonal antibodies were raised in rabbits (Charles River Wiga, Hannover) and affinity purified, as described .
Protein extraction and analysis
Extraction of total soluble proteins from BY-2 cells and tobacco leaves and analysis of antibody fragments by ELISA and Western blot were performed essentially as described . A Fab fragment of the mAb24 and a F(ab′)2 fragment of the mAb29  were used as ELISA standards. To demonstrate simultaneous binding of the biscFv2429 to both antigens, TMV coat protein monomers were coated on the M129B microtiter plate (Dynatech, Burlington, MA, USA). After blocking with 1% (w/v) BSA serial dilutions of plant extracts containing biscFv2429 were added. Bound biscFv2429 was detected using intact 100 µg·mL−1 TMV virions as the second antigen followed by the mouse anti-TMV monoclonal antibody 75 and an alkaline phosphatase conjugated goat anti-(mouse Fc) secondary antibody (Jackson Immuno Research Laboratories, West Grove, PA, USA), prior to p-nitrophenylphosphate substrate addition.
Protein concentrations were determined with the BioRad Protein Assay using BSA as standard.
Surface plasmon resonance
Biomolecular interaction analyses were carried out in HBS buffer (150 mm NaCl, 3.4 mm EDTA, 0.05% (v/v) surfactant P20, 10 mm Hepes, pH 7.4) using the BIACORE® 2000 (BioSensor, Uppsala, Sweden). TMV was immobilized on a CM5-rg sensorchip (BioSensor) using the Amine Coupling Kit (BioSensor). The surface of the sensorchip was activated with 70 µL 100 mmN-ethyl-N′-(dimethylaminopropyl)-carbodimide-hydrochloride/400 mmN-hydroxy-succinimide buffer using a flow rate of 10 µL·min−1. For immobilization of the virus, 200 µg of TMV in 100 µL 10 mm formic acid pH 3.0 were applied (flow rate: 5 µL·min−1). Subsequently, the sensorchip was deactivated with 70 µL 1 m ethanolamine hydrochloride pH 8.5 (flow rate: 10 µL·min−1) and conditioned with 10 µL 100 mm HCl (flow rate: 5 µL·min−1). Between sample injections the surface was regenerated with 10 µL 30 mm HCl (flow rate: 30 µL·min−1).
BY-2 suspension cells expressing biscFv2429 were disrupted by sonication in extraction buffer (50 mm Tris/HCl pH 8.0, 100 mm NaCl, 20 mm dithiothreitol, 5 mm EDTA) (S. Schillberg et al. unpublished data). The cell extract was clarified and filtered (0.2 µm). A glutathione S-transferase (GST)-fusion protein containing the mAb29 TMV coat protein monomer epitope (A. Holmzem et al. unpublished data) was coupled to activated glass beads (ProSep9-CHO, BioProcessing, Consett, UK). GST-TMV epitope immobilized glass beads were packed into a 5 × 200-mm column (1.6-mL bed volume) and equilibrated with extraction buffer. The BY-2 cell extract was passed through the column at a flow rate of 900 cm·h−1, nonspecifically bound material was removed by washing and the biscFv2429-KDEL was eluted with 100 mm citrate pH 3.0.
Small pieces of transgenic tobacco leaves were embedded at low temperature for immunogold labeling . Immunogold labeling of thin sections on plastic-filmed gold grids was carried out as previously described .
Construction of biscFv2429
Our intent was to evaluate plants as an expression system for bispecific single-chain antibodies (biscFv) and to examine whether targeting the biscFv2429 to different compartments of plant cells influenced the expression levels of functional bispecific antibody. Construction of a model bivalent biscFv for testing expression and activity in plant cells was carried out using two well-characterized scFvs binding to two independent TMV epitopes [23,24]. Two scFvs, scFv24 recognizing neotopes on intact TMV virions and scFv29, recognizing a cryptotope on the TMV coat protein monomer, were connected by a linker peptide present in a naturally secreted fungal cellulase protein. The flexible linker peptide of Trichoderma reesi cellobiohydrolase I (CBHI) is an effective interdomain linker for both scFv  and biscFv  expression in E. coli. The CBHI-linker was PCR amplified to introduce the restriction sites ClaI and AatII, to permit construction of the bispecific antibody biscFv2429 in a single expression cassette.
For expression of biscFv2429 in the cytoplasm of tobacco cells, this cassette was cloned into the plant expression vector pSS  as shown in Fig. 1 (biscFv2429-cyt). A second construct to target biscFv for secretion to the apoplast contained the original mouse leader peptide sequence of the parental monoclonal Ab24 (biscFv2429-apoplast) . To retain the biscFv in the lumen of the ER, a C-terminal KDEL-signal [34,35] was added to biscFv2429-apoplast construct (biscFv2429-KDEL).
Transient expression in N. tabacum cv. BY-2
Prior to stable transformation of tobacco plants with the three biscFv2429-constructs, expression of recombinant proteins was analyzed by transient expression in suspension cultured plant cells. N. tabacum cv. BY-2 suspension cells were cocultivated with recombinant agrobacteria and expression of functionally active biscFv2429 was analyzed by ELISA. After 2–3 days of cocultivation, significant expression of biscFv2429-apoplast and biscFv2429-KDEL was detected (Fig. 2). The ER-retained biscFv2429-KDEL was only detectable in cell extracts whereas biscFv2429-apoplast could be detected both in the culture supernatant and in cell extracts. Although the major fraction of biscFv2429-apoplast was detectable in the cell extracts (56–75%), the presence of functional biscFv2429 in the culture supernatant indicated that the antibody fragment was able to cross through the plant cell wall. In contrast, expression of functional biscFv2429-cyt in the cytoplasm was significantly lower and at the ELISA detection limit.
Stable expression in N. tabacum cv. BY-2
The tobacco suspension cell line BY-2 was stably transformed with biscFv2429-apoplast and biscFv2429-KDEL. biscFv2429-apoplast could be detected both in the culture supernatant (≈ 30–70%) and in the cell extracts, whereas biscFv2429-KDEL was retained within the cells (Fig. 3). Saturation of ER retrieval within the secretory pathway was not observed, based on the absence of detectable biscFv2429-KDEL in the culture supernatant. Levels of functional biscFv2429 antibody fragments containing the C-terminal KDEL retention signal reached 0.064% of total soluble protein (Table 1). This level of ER-targeted recombinant protein was 10-fold higher than that of biscFv2429-apoplast secreted into the apoplast and culture medium.
Table 1. Stable expression of biscFv2429-apoplast or biscFv2429-KDEL inN. tabacumcv. BY-2 cells and transgenic tobacco plants. Total soluble protein was isolated from N. tabacum cv. BY-2 cells or from leaves of transgenic N. tabacum cv. Petite Havana SR1 T0 plants producing either biscFv2429-apoplast or biscFv2429-KDEL. Levels of functional scFv, expressed as part of biscFv2429, were quantitated by ELISA using anti-mAb24 antisera and are indicated as percentage of total soluble protein.
No. tested cultures/plants
No. cultures/plants expressing functionally active biscFv2429
Average expression (% of total soluble protein)
Highest expression (% of total soluble protein)
Transgenic N. tabacum
To verify that plant-expressed apoplast and ER-targeted biscFv2429 retained both epitope specificities for TMV, a dual epitope ELISA was used. TMV coat protein was immobilized on a microtiter plate and used as the capturing antigen to bind biscFv2429 molecules via the scFv29 domain. TMV virions were added and bound to the immobilized biscFv2429 via the scFv24 domain. Bound TMV virions were subsequently detected using anti-TMV mAb75 antibody, followed by an alkaline-phosphatase conjugated goat anti-mouse antibody. The ELISA data demonstrated simultaneous binding of both scFv domains in biscFv2429-apoplast and biscFv2429-KDEL to their corresponding antigens, intact TMV particles and TMV coat protein monomers (Fig. 3).
Surface plasmon resonance and Western blot analysis of biscFv2429
The simultaneous binding of biscFv2429 to both epitopes was also confirmed by surface plasmon resonance-based, real-time, biomolecular interaction analysis . The dextran matrix of a CM5-rg sensorchip was coated with intact TMV virions using standard amine coupling chemistry. After surface stabilization, a protein extract from a BY-2 suspension culture expressing biscFv2429-KDEL was injected and binding was observed, as the neotope-specific scFv24 domain of the biscFv2429 was captured by the TMV virions (Fig. 4, peak I). After injection of the second antigen, dissociated TMV coat protein monomer, binding of the monomer to the second antigen-binding scFv29 domain in the bispecific antibody was observed (Fig. 4, peak II). The specificity of the signals was further confirmed by additional binding of an scFv24-specific polyclonal rabbit anti-mAb24 serum to the complex (Fig. 4, peak III). No binding was measured on a control surface lacking immobilized virions, or when using an extract from a nontransformed control BY-2 suspension culture. Also, no binding was detected when TMV coat protein monomers were applied to scFv24 bound to the TMV-chip (data not shown).
biscFv2429-KDEL was purified by affinity chromatography using a GST-fusion protein containing the epitope recognized by scFv29 coupled to activated glass beads. Western blot analysis of affinity-purified biscFv2429-KDEL demonstrated that the biscFv2429 had the expected size (52 kDa) and could be detected with polyclonal anti-mAb24 or anti-mAb29 antisera. This showed that the bispecific antibody contained both scFv fragments (Fig. 5) and that affinity purification was a suitable method to obtain intact biscFv2429.
Stable expression in N. tabacum cv. Petite Havana SR1
Transgenic tobacco plants expressing biscFv2429 were regenerated after leaf disc transformation of N. tabacum cv. Petite Havana SR1. Twelve plants expressing biscFv2429-apoplast and 11 plants expressing biscFv2429-KDEL were obtained. Levels of functionally expressed recombinant protein, as determined by ELISA, reached up to 0.056% of total soluble protein (corresponding to 7.9 µg biscFv2429 per gram leaf tissue, wet weight) for the apoplast-targeted biscFv2429 and 1.65% of total soluble protein (corresponding to 79.4 µg biscFv2429 per gram leaf tissue, wet weight) for the ER-retained biscFv2429 (Table 1). The maximum biscFv2429 protein levels in transgenic plants were ninefold (biscFv2429-apoplast) and 26-fold (biscFv2429-KDEL) higher than those obtained in transgenic suspension cultures.
Analysis of transgenic BY-2 cells demonstrated that biscFv2429-KDEL was not secreted into the culture medium but was completely retained within the cells. Immuno-electron transgenic microscopy confirmed the subcellular localization of biscFv2429-KDEL in transgenic N. tabacum cv. Petite Havana SR1. In young leaves of biscFv2429-KDEL-expressing transgenic tobacco plants, biscFv2429 was localized to the ER (Fig. 6A). The number of gold particles found in other plant cell compartments was consistent with background labeling of a control plant (Fig. 6B).
In this study, we demonstrate high-level expression of a functional bispecific antibody in transgenic tobacco suspension cells and transgenic plants. To our knowledge, this is the first report on the expression of a functional biscFv in plant cell suspension culture or in transgenic plants.
To assess the feasibility of using plants as an expression system for biologically active or therapeutic bispecific antibodies, we engineered biscFv2429 as a model bispecific antibody, which contains two genetically linked scFv24 and scFv29 antibodies. Functional biscFv2429 was expressed in different plant cell compartments, but with significant differences in accumulation levels. Cytoplasmic levels of biscFv2429 were low, in agreement with our studies on cytoplasmic expression of scFv24 and scFv29  (S. Schillberg et al., unpublished data). This may result from the absence of specific chaperones and protein disulfide isomerase in this compartment resulting in scFv misfolding and enhanced proteolysis . In contrast, biscFv2429 targeting to the apoplast and ER was significantly higher, reaching a maximum level of 1.65% of the total soluble protein in the ER of transgenic tobacco plants. Maximum apoplastic and ER retarded biscFv2429 protein levels in transgenic plants were nine and 26 times, respectively, higher when compared to transgenic BY-2 suspension cultures producing the same biscFv2429 proteins.
The highest levels of recombinant bispecific antibody were obtained when the biscFv2429 was targeted to the ER using a murine leader peptide and carried a C-terminal KDEL-tag [34,35]. Targeting of heterologous proteins to the lumen of the ER of eukaryotic cells is advantageous, because the ER contains enzymatic catalysts that allow proteins to fold and mature, resulting in an increased protein stability and a greatly enhanced level of protein accumulation [38–41].
The high levels of biscFv2429 produced in plants provide sufficient quantities for large-scale purification of recombinant biscFv antibodies, which has several major advantages over prokaryotic or mammalian tissue-culture expression systems. Immunoglobulin synthesis in plants uses a similar pathway to mammalian cells and plants are inexpensive to maintain as either cell suspension cultures, or as whole plants on a laboratory or field scale. Furthermore, the use of plants as ‘bioreactors’ for production of recombinant proteins is gaining interest as antibodies produced in plants are essentially identical to those derived from hybridomas.
As we demonstrate here, plants are particularly applicable to the production of biscFvs, which have numerous applications in tumor diagnosis and therapy . In addition, we have shown that it is possible to affinity purify plant-derived biscFv2429 in an intact and functional form. Therefore, downstream processing procedures for biscFv antibody molecules should be feasible at lower cost with reduced effort. Consequently, expression in plants offers an attractive alternative to traditional animal and microbial routes of production. Our long-term goal is to use plants as bioreactors for the large-scale production of diagnostic and therapeutic molecules, such as a tumor-specific scFv connected to a CD8-specific scFv, and these antibodies will be powerful tools for the management and treatment of human disease.
We have demonstrated that biscFv2429 could be functionally expressed in different compartments of plant cells. This will allow analysis of structure–function relationships of two targets in planta simultaneously and aid engineering of pathogen-resistant plants through biscFv antibody molecules that can be targeted to the compartment where the pathogen is most vulnerable and where they are most effective [19,23,43,44]. Bispecific single-chain antibodies binding to two different conserved epitopes of pathogen-derived proteins, such as viral replicase and movement proteins, will be suitable for generating transgenic plants with a more durable and effective form of antibody-mediated pathogen resistance.
We would like to thank Dr Neil Emans for critical reading of the manuscript and Achim Holzem for technical assistance. For providing N. tabacum cv. BY-2 cell line we thank Prof. Anne-Marie Lambert (Institut de Biologie Moléculaire des Plants, Strasbourg). This work was supported by a grant from the European Commission (FAIR CT95-0905).
*Note: Both authors contributed equally to this work.