Functional expression of a single-chain antibody to ErbB-2 in plants and cell-free systems

The murine monoclonal antibodies (mAb) W6/800E6 (in short, 800E6) and mAb 100A4, an IgG1 and IgG2a respectively, bind two distinct polypeptide epitopes in the extracellular portion of ErbB-2 [19]. They were used in all flow cytometry experiments at optimal pre-determined dilutions. Hybridoma 800E6 was used to clone Ig sequences. MAb 100A4 was used as a control in some experiments. The mAbs W6/32 and Ep3 [20] recognize class I Major Histocompatibility antigens and a melanoma antigen, respectively, and were also used as controls. ErbB-2 transfectants and neoplastic cell lines (Table 1) were previously characterized by others [21] and ourselves [19] for ErbB-2 expression. Monovalent Fab' fragments were prepared by papain (Sigma-Aldrich St. Louis, MO, USA) digestion.

The cloning of Variable Light (VL) and Variable Heavy (VH) chain Ig sequences from the 800E6 hybridoma into the pEMBL-ScFv800E6 and pHEN (for prokaryotic expression) vectors has been described [13]. The pEMBL-ScFv800E6 plasmid was used to generate all the remaining constructs, depicted in figure 1. For stable plant expression, a Hind III/Eco RI fragment was cloned into pBG(dAbs)BIN (Fig. 1A). For transient plant expression, a Hind III/Xho I fragment was filled in by the Polymerase fragment unit (Pfu, Stratagene), and blunt end-ligated into Sal I-linearized pP2C2S (Fig. 1B). For in vitro expression in transcription-translation systems, ScFv800E6 sequences from pEMBL-ScFv800E6 were PCR-amplified and subcloned into pIVEX 2.1 and pIVEX 2.2 vectors (Roche Applied Science), designed for the introduction of Strep II (Strep) tags at either the N-terminus or C-terminus (Fig. 1C and 1D). Fragments from pIVEX 2.1 and pIVEX 2.2 were excised and cloned into pIVEX 2.3d and pIVEX 2.4d to express polyhistidine (6XHis)-tagged ScFvs, (Fig. 1F and 1G). The two clones with C-terminal tags (Fig. 1C and 1E) were then linearized with Xho I, blunted, and re-circularized by ligation to bring the tag in frame with the open reading frames. A construct with a 27 residue-long spacer arm between the N-terminal His-tag and the coding sequence was produced by transferring the insert from pEMBL-ScFv800E6 into the polylinker of pIVEX 2.4d using Not I-Hind III adapters (Fig. 1G). The resulting ScFvs are shown (Fig. 1H). A control ScFv with irrelevant specificity (to Citrus Tristeza Virus, ScFvαCTV) was expressed in bacteria [22].

Plant biology protocols were carried out as described [5], according to standard procedures [23]. Stable expression of ScFv800E6 in plants has been described [13]. Briefly, bacterial cultures of the A. tumefaciens GV3101 strain harboring pBG(dAbs)BIN-ScF800E6 were used to transform leaf disks from Nicotiana tabacum, and transgenic leaf disks selected in the presence of kanamycin. One shoot per leaf disk was grown in vitro in a climatic chamber, and plant RNA was analyzed by RT-PCR for the expression of VH sequences. Positive transgenic plants and their progeny were grown in a containment greenhouse, leaf tissues were homogenized [4], and total proteins were analyzed by Western blot.

Nicotiana benthamiana plants were grown up to the 6 leaves state in a controlled greenhouse. In vitro transcripts generated from 1 μg of Spe I-linearized pP2C2S-ScFv800E6 were used for infection by rubbing leaves dusted with celite [24]. Tissues were collected seven days later, frozen in liquid nitrogen, and proteins were extracted in 0.05 M Tris-HCl pH 8.0/0.3 M NaCl/0.01 M PMSF/0.005 M ascorbic acid (5 ml per g of frozen tissue), homogenized, sonicated at 100 Watts (3 times, 10" each), ultra-filtered and concentrated (20×).

The different pIVEX-ScFv800E6 proteins (Fig. 1H) were expressed using the RTS 100 E. coli, a newly developed E. coli cell-free expression system for disulfide bonded proteins, according to the manufacturer's protocol (Roche Applied Science), in a ProteoMaster instrument [25]. This is based on the development, extensively described by Kim and Swartz [18], of a transcription-translation system involving several major novelties: (a) the inactivation of disulfide-reducing activities contained in a standard E. coli S30 extract, (b) the use of a glutathione redox buffer, (c) pH optimization, (d) the addition of GroE chaperones, and (e) a semi-continuous exchange dialysis format to achieve longer expression reactions. A conventional, reducing cell-free expression system was also used in control experiments. The UK construct and the E. coli chaperone DnaK are also from Roche. Brij 35 is from Calbiochem/EMD Biosciences, San Diego, CA. His-tagged ScFvs were purified by metal-chelate affinity chromatography on Ni-NTA agarose columns (Qiagen).

All ScFv preparations were tested for their ability to bind ErbB-2⁺ cells by flow cytometry. The binding of ScFvs and mAbs was revealed using fluorescein isothiocyanate (FITC)-labeled rabbit antibodies to whole murine Ig (Cappel/ICN, Aurora, OH) at 50 μg/ml in the second step, unless noted otherwise. Where noted, antimurine Ig from a different vendor (Dako, Glostrup, Denmark) was also used. Phycoerythrin (PE)-conjugated Strep-Tactin was from IBA (Goettingen, Germany). An ELISA binding assay was carried out as follows: 96 well PVC microtiter plates (Corning, Acton, MA) were coated overnight with purified mAb 800E6, Fab' 800E6 and ScFv800E6 (approximately equimolar concentrations of 6.0, 2.0 and 1.0 μg/ml, respectively, in 0.1 M NaHCO₃ pH 9.5). Following 3 washes with NaCl-tween (saline containing 0.05% Tween 20), adsorbed mAbs and fragments were incubated for 1 h with FITC-labeled rabbit antibodies to whole murine Ig (50 μg/ml, e.g. the concentration used in immunofluorescence). After washing with NaCl-tween, binding of the FITC-labelled antibody was revealed by 1 h incubation with peroxidase-conjugated goat anti rabbit Ig (Cappel/ICN), followed by washing and color development using O-phenylenediamine (Sigma-Aldrich) as substrate. Cells were metabolically labeled by incubation for 18 h in ³⁵[S]-methionine (3.7 MBq/ml) containing medium, solubilized by the nonionic detergent NP40, and immunoprecipitated with protein A-sepharose 4B immunoadsorbents (Amersham) pre-loaded with rabbit antimurine Ig and either ScFvs or mAbs. Equilibrium binding studies were performed by incubating affinity purified antibodies and recombinant ScFvs (radiolabeled to a specific activity of 185 MBq/mg with ¹²⁵I by the Chloramine T method) with target cells (1 × 10⁵ per well in 100 μl volumes) in membrane-sealed 96 well plates (Millipore, Billerica, Ma) allowing instantaneous removal of free ligands by vacuum manifold filtration. Values of bound and free ligands were plotted according to the linear transform of the Law of Mass equilibrium, and the best fit of experimental data determined by regression analysis. All these procedures and Scatchard plot analysis are described in detail elsewhere [20].

Human breast carcinoma specimens were obtained in the course of ablative surgery, in compliance with informed consent procedures. Four μm frozen sections, fixed in cold acetone for 10 min., were immunostained using a biotinylated anti Ig secondary antibody, and a streptavidin-biotin detection kit (Super Sensitive Detection Kit, Menarini, Florence, Italy), and the samples were counterstained with Mayer hematoxylin.

Preliminary experiments were performed using crude bacterial lysates to compare the binding of ScFv800E6 and its parental antibody, mAb 800E6. Both reagents bound ErbB-2 transfectants but not parental ErbB-2-negative cells, as expected, although the former was 7–10 times weaker than the latter. An irrelevant ScFv to Citrus Tristeza Virus (ScFvαCTV) did not stain either cell line (Fig. 2A and 2B). In spite of the different binding intensities, the two reagents concordantly estimated ErbB-2 surface expression in a panel of breast carcinoma cell lines known [21] to express a wide range of ErbB-2 levels (Table 1). ScFv800E6 was titratable upon serial dilution (Fig. 2C), further supporting its binding specificity. In addition, ScFv800E6 and mAb 800E6 immunoprecipitated an identical 185 kD band from soluble extracts of metabolically radiolabeled SK-BR-3 cells (Fig. 2D).

Upon stable plant transformation, RT-PCR of putative transgenics revealed that 56% of them expressed ScFv800E6 transcripts. Extracts from selected positive plants were analyzed by Western blotting and estimated to contain ScFv polypeptides at a concentration of 0.6–0.8 μg/g of plant tissue (not shown). Similarly, ScFv800E6 was detected by Western blot in extracts from both directly and systemically infected leaves of transiently modified Nicotiana benthamiana plants. Much higher yields (800 μg/g of plant tissue) were obtained from leaves exhibiting systemic symptoms (data not shown). Extracts (50 μl) containing 0.14 μg/ml and 160 μg/ml of ScFv800E6 from stable and transient transgenic plants, respectively, were used to stain SK-BR-3 cells in indirect immunofluorescence. A representative flow cytometry analysis at the widely different antibody concentrations noted above revealed that the two preparations similarly bound SK-BR-3 cells (Fig. 2E). Binding intensities were comparable to (or slightly lower than) the intensities seen with bacterial extracts on the same target cells (compare panels C and E), indicating that saturating concentrations of ScFv had been reached. Quite puzzling, ScFv binding could not be detected using an antibody to a c-Myc tag present on the ScFvs produced in bacteria and plants (not shown), hampering purification. Based on these findings, before embarking on the production of additional transgenic plants, we resorted to in vitro transcription-translation systems to rapidly obtain a set of modified ScFv800E6 proteins to be used for the identification of optimal folding conditions and for the evaluation of tagging and purification strategies.

ScFv800E6 constructs were prepared in 4 distinct pIVEX vectors, each of which Strep- or His-tagged at the N- or C-terminus (Fig. 1C–F), plus a fifth construct with a N-terminal His-tag on the tip of a 27 residue-long spacer arm (Fig. 1G). The Strep tag was selected because it is particularly useful for flow cytometry detection, since it is recognized with particularly high affinity by fluorescent, recombinant Strep-Tactin. Because the ScFvs contain two disulfide bonds, the five constructs were expressed not only in a conventional E. coli-based cell-free transcription-translation system, but also in a newly developed transcription-translation system for disulfide-bonded proteins. The control urokinase (UK) protein, that contains 6 disulfide bonds, was transcribed-translated in parallel. Supernatants from representative His(6x)-ad-N-ScFv800E6 transcription-translation mixtures were run on a SDS-PAGE gel, and either stained by Coomassie blue (Fig. 3A) or Western blotted by peroxidase-conjugated antibodies to the His-tag and murine Ig (Fig. 3B). His(6x)-ad-N-ScFv800E6 (solid black arrowheads: lanes 3, 4, 5, 15, 16 and 17) and UK (open white arrowheads: lanes 2 and 13) displayed the expected electrophoretic mobilities (approximately 32 kD in both cases), and were absent in lanes loaded with mock-transcribed-translated mixes (lanes 1, 12, and 18). Neither the nonionic detergent Brij 35 (1%) nor the E. coli chaperone DnaK significantly increased ScFv yield (compare lanes 3 and 4 to 5; 15 and 16 to 17; 21 and 22 to 23). In contrast, His(6x)-ad-N-ScFv800E6 was barely detectable when translated in a conventional, reducing system (lanes 14 and 20). Under these conditions the ScFv could be mainly recovered in the insoluble fractions (data not shown).

To assess the activity and fine specificity of the ScFvs produced in the transcription-translation system, we tested the ability of the parental mAbs and His(6x)-ad-N-ScFv800E6 to inhibit each other in flow cytometry. Pre-incubation of SK-BR-3 cells with a wide range of mAb 800E6 concentrations, but not mAb 100A4 to a distinct ErbB-2 epitope [19], proportionally (and up to 95%) inhibited the binding of His(6x)-ad-N-ScFv800E6 (Fig. 4A), with a clear prozone effect (a paradoxical reduction of inhibition at high mAb concentrations) at 1 mg/ml. Four selected points are also analytically displayed in panel B. Lesser, but still significant, levels of competition for the 800E6 epitope were seen when ScFv800E6 was used in the pre-incubation step (not shown). Thus, the binding of the disulfide-bridged ScFv can be competed by, and retains the fine epitope specificity of, the parental antibody.

All the 5 tagged pIVEX-ScFvs were transcribed-translated in disulfide mixtures, and tested by flow cytometry at 5 different dilutions using the same FITC-labeled secondary antibody to murine Ig. ErbB-2 was detected on SK-BR-3 cells with comparable efficiency and with a similar prozone effect (Fig. 4C). The binding of all the ScFvs displayed the highest mean florescence intensity (m.f.i.) value (140 approximately) at similar concentrations (in the microgram per ml range). This concentration is equivalent to 20 ng per 5 × 10⁵ target cells. Two pIVEX-ScFvs bearing a suitable tag (Strep-N-ScFv800E6 and Strep-C-ScFv800E6) were also tested in flow cytometry using PE-conjugated Strep-Tactin and PE-conjugated streptavidin as secondary reagents (Fig. 4D). Four observations were made: (a) the C-terminally tagged ScFv performed much better than the N-terminally tagged one; (b) the fluorescence intensity conveyed by Strep-Tactin was in each case much stronger than that conveyed by conventional streptavidin, to the extent that the weak binding of Strep-N-ScFv800E6 was essentially undetectable by the latter; (c) the optimal combination (Strep-C-ScFv800E6 followed by Strep-Tactin) resulted in a substantial improvement in m.f.i. values as compared to the same ScFv followed by antimurine Ig secondary reagents conjugated with either FITC or PE (m.f.i of 700 vs. 140; compare with Fig. 4C); (d) these m.f.i. values were closer to those typical of the parental mAb 800E6 followed by conventional secondary antimurine Ig reagents conjugated with either FITC or PE (m.f.i. values > 1300 in all the experiments that were performed; see Fig. 2 and data not shown).

Similar to the strep II tag, a preference for C-terminal tagging was also seen with His-tagged ScFvs and anti His secondary antibody, but the introduction of a spacer arm, as in His(6x)-ad-N-ScFv800E6, restored the availability of the N-terminal epitope tag (not shown). It can be concluded that: (a) all the variants of ScFv800E6 can be produced in a cell-free system at concentrations and amounts similar to those recommended for flow cytometry and immunohistochemistry with most monoclonal antibodies, (b) the type and position of the tags have no detectable effect on ScFv binding (i.e. tags do not interfere with the antigen binding site), whereas the tag position is important in order to make the tag available for secondary reagents, and (c) Strep-Tactin outperforms streptavidin. Thus, optimization in the position and type of tag, as well as selection of appropriate secondary reagents, largely compensate for the reduced performance of ScFvs in comparison to their parental mAb.

To directly address (a), we tested the FITC-labeled antibody to whole mouse Ig used in flow cytometry for its ability to bind mAb 800E6, its monovalent Fab' fragment (as a control), and affinity-purified His(6x)-ad-N-ScFv800E6, taking advantage of the ELISA assay described in methods. The binding of FITC-labeled secondary antibodies (from two different commercial sources, see methods) to mAb, Fab and ScFv adsorbed on PVC plates in equimolar amounts displayed the following ratios: 10:5:1.

These results are consistent with the removal of some and most framework Ig epitopes from the Fab' and ScFv, respectively, and quantitatively agree with the 7–10 fold difference in binding intensities between the ScFv and the parental mAb, routinely detected in the above flow cytometry studies. Having estimated their reactivity with FITC-labeled secondary antibodies, the three reagents were again compared, in the same flow cytometry experiment, for their ability to bind SK-BR-3 cells. Optimal working dilutions of the three reagents were determined as shown in the experiments displayed in figures 2C and 4C to rule out the possibility of prozone effects, and staining was performed at the indicated, selected equimolar concentrations. Representative results with mAb, one of two preparations of Fab', and ScFv, followed by one of the two FITC-labeled secondary antibodies that were employed (Fig. 5A) revealed an approximately 10:1:1 ratio in the binding of the three reagents, with the Fab and the ScFv displaying extremely similar binding intensities. The ten-fold drop in the ScFv performance was confirmed, and correlated with the ten-fold reduced recognition of ScFv vs. mAb detected by the secondary antibody in the ELISA binding assay. In contrast, the ten-fold drop in Fab' performance exceeded the two-fold reduced recognition by the secondary antibody.

These results altogether provide further support for (a), and suggest that (b), e.g. reduced binding due to monovalent interactions, is also involved, although a more direct method is needed to account for the different extent of this reduction in the case of monovalent Fab' and ScFv. Then, we directly measured the equilibrium binding affinities of ¹²⁵I-labeled mAb800E6, Fab' 800E6, and His(6x)-ad-N-ScFv800E6, and determined the valence of the parental mAb. Equilibrium binding studies demonstrated a high binding affinity (Ka in the 10⁸ M^-1 range) for all the three ligands (Fig. 5B), and detected the expected drops in affinity of the ScFv and Fab' as compared to the parental antibody. These were estimated to be approximately 3-fold and 4.5-fold, respectively. A double number of ScFv and Fab' as compared to mAb binding sites (intercept on the abscissa) was consistent with mAb 800E6 being the only reagents capable of bivalent interactions. Altogether, the limited range of variation in the observed association binding constants, and the detection of monovalent vs. divalent binding demonstrated that the absence of Fc epitopes (a) and monovalent binding (b) largely explain the apparent low staining efficiency of ScFv800E6 when using conventional secondary anti Ig reagents, whereas alterations and/or mis-folding (c) of the ScFv play a minor, if any, role in reducing its binding performance. Finally, a comparison of Fab' and ScFv binding abilities in different assays indicates that enzymatic fragmentation of the natural antibody is more deleterious in our hands than cloning and expression in a recombinant form of ScFv800E6 sequences.

To provide evidence that recombinant ScFvs and phytoantibodies may be useful in a diagnostic setting, semi-serial cryostat sections of breast carcinoma lesions were stained with different ScFv800E6 preparations. Representative results (Fig. 6) demonstrate that all of them clearly and reproducibly discriminated ErbB-2⁺ from ErbB-2^- breast carcinoma lesions.