Intradiabodies, Bispecific, Tetravalent Antibodies for the Simultaneous Functional Knockout of Two Cell Surface Receptors*

The specific and high affinity binding properties of intracellular antibodies (intrabodies), combined with their ability to be stably expressed in defined organelles, provides powerful tools with a wide range of applications in the field of functional genomics and gene therapy. Intrabodies have been used to specifically target intracellular proteins, manipulate biological processes, and contribute to the understanding of their functions as well as for the generation of phenotypic knockouts in vivo by surface depletion of extracellular or transmembrane proteins. In order to study the biological consequences of knocking down two receptor-tyrosine kinases, we developed a novel intrabody-based strategy. Here we describe the design, engineering, and characterization of a bispecific, tetravalent endoplasmic reticulum (ER)-targeted intradiabody for simultaneous surface depletion of two endothelial transmembrane receptors, Tie-2 and vascular endothelial growth factor receptor 2 (VEGF-R2). Comparison of the ER-targeted intradiabody with the corresponding conventional ER-targeted single-chain antibody fragment (scFv) intrabodies demonstrated that the intradiabody is significantly more efficient with respect to efficiency and duration of surface depletion of Tie-2 and VEGF-R2. In vitro endothelial cell tube formation assays suggest that the bispecific intradiabody exhibits strong antiangiogenic activity, whereas the effect of the monospecific scFv intrabodies was weaker. These findings suggest that simultaneous interference with the VEGF and the Tie-2 receptor pathways results in at least additive antiangiogenic effects, which may have implications for future drug developments. In conclusion, we have identified a highly effective ER-targeted intrabody format for the simultaneous functional knockout of two cell surface receptors.

The specific and high affinity binding properties of intracellular antibodies (intrabodies), combined with their ability to be stably expressed in defined organelles, provides powerful tools with a wide range of applications in the field of functional genomics and gene therapy. Intrabodies have been used to specifically target intracellular proteins, manipulate biological processes, and contribute to the understanding of their functions as well as for the generation of phenotypic knockouts in vivo by surface depletion of extracellular or transmembrane proteins. In order to study the biological consequences of knocking down two receptortyrosine kinases, we developed a novel intrabody-based strategy. Here we describe the design, engineering, and characterization of a bispecific, tetravalent endoplasmic reticulum (ER)-targeted intradiabody for simultaneous surface depletion of two endothelial transmembrane receptors, Tie-2 and vascular endothelial growth factor receptor 2 (VEGF-R2). Comparison of the ER-targeted intradiabody with the corresponding conventional ER-targeted single-chain antibody fragment (scFv) intrabodies demonstrated that the intradiabody is significantly more efficient with respect to efficiency and duration of surface depletion of Tie-2 and VEGF-R2. In vitro endothelial cell tube formation assays suggest that the bispecific intradiabody exhibits strong antiangiogenic activity, whereas the effect of the monospecific scFv intrabodies was weaker. These findings suggest that simultaneous interference with the VEGF and the Tie-2 receptor pathways results in at least additive antiangiogenic effects, which may have implications for future drug developments. In conclusion, we have identified a highly effective ER-targeted intrabody format for the simultaneous functional knockout of two cell surface receptors.
Antibodies can bind almost any molecule with high specificity and affinity, providing powerful biotechnological tools for diagnostic and therapeutic applications. Advances in recombinant DNA technology have facilitated the manipulation of the antibody genes, so that design, cloning, expression, and use of single-chain antibodies have become routine procedures in protein engineering. The potential of single-chain antibody fragments (scFv) 1 for intracellular applications, termed "intrabodies," has been exploited in a number of laboratories (1)(2)(3)(4)(5)(6)(7)(8). To date, intrabodies have been utilized for targeting proteins in a singular fashion. Bispecific and tetravalent antibody fragments could improve and expand the inhibitory potential of intrabodies by exhibiting increased apparent affinity for their antigen and by being more efficient at inhibiting protein function or intracellular trafficking (9). Intrabodies present a potent alternative to methods of gene inactivation that target at the level of DNA or mRNA, such as antisense (10), zinc finger proteins (11), targeted gene disruption, or the relatively new RNA interference (12). Operating at the posttranslational level, intrabodies can be directed to relevant subcellular compartments and precise epitopes on target proteins (2,13,14), potentially blocking only one out of several functions of an expressed protein. Numerous studies have reported the development of engineering antibodies that are both multispecific and multivalent (15,16). Applying this knowledge, the goal of our study was to develop an ER-targeted intrabody format for the simultaneous downregulation of two independent cell surface receptors, in order to investigate the biological consequences of knocking down two receptor-tyrosine kinases. To accomplish this, an ER-targeted tetravalent antibody construct, with dual specificity, was generated. Using a recombinant adenovirus as gene delivery system, we could show that this intrabody construct, termed here an "intradiabody," was expressed in the ER and able to trap both targeted proteins in the same compartment. The intradiabody targets the endothelial transmembrane receptors Tie-2 and VEGF-R2, which are essential for angiogenesis. The interplay of VEGF, VEGF-R2, Tie-2, and Ang-1 and -2 has been suggested as a key modulator in the onset of tumor angiogenesis (17). According to this model, Tie-2 is constitutively engaged with Ang-1 in quiescent blood vessels. The Tie-2/Ang-1 complex stabilizes quiescent blood vessels by promoting their interaction with surrounding perivascular cells, smooth muscle cells, and the extracellular matrix. The constitutive Tie-2/ Ang-1 complex is antagonized by Ang-2, which is up-regulated in endothelial cells that are proximal to the tumor. By competing with Ang-1 for Tie-2 binding, Ang-2 destabilizes the interaction of endothelial cells and their microenvironment. This is thought to sensitize the endothelial cells to VEGF signaling. Thus, Ang-2 produced by endothelial cells promotes tumor angiogenesis in concert with VEGF produced by the tumor. Comparison of the effect of the intradiabody (targeting Tie-2 and VEGF-R2) with the corresponding conventional scFv intrabodies (targeting Tie-2 or VEGF-R2 alone) revealed a remarkable superiority of the intradiabody, as represented by a complete and extended surface depletion of both Tie-2 and VEGF-R2. This finding can be attributed to the extended halflife of our intradiabody, as determined by pulse-chase studies. In addition, we show that the intradiabody strongly inhibits endothelial tube formation beyond that seen with inhibitors of either receptor-tyrosine kinase alone, thus confirming its antiangiogenic properties. Our results confirm that the inhibition of the VEGF receptor pathway cannot be compensated by the Tie-2 pathway, nor vice versa (18), but also demonstrates that targeting both pathways simultaneously results in additive antiangiogenic effects in vitro. Combining the specific and high affinity binding properties of our intradiabody with its ability to be stably expressed in the ER, we identified a very effective intrabody format for the simultaneous functional knockout of two cell surface receptors.
Library Generation and Selection-The generation and selection of rabbit/human chimeric Fab libraries, as well as the characterization of positive Fabs, were done essentially as described (19,20). In brief, rabbit V-, V-, and VH-encoding sequences were amplified from first strand cDNA and fused to human C-and C H 1-encoding sequences, respectively, followed by assembly of chimeric rabbit/human light chain-and Fd fragment-encoding sequences and by asymmetric SfiI cloning into phagemid vector pComb3X. Libraries were panned against VEGF-R2/VEGF complex immobilized on Costar 3690 96-well ELISA plates (Corning, Acton, PA). Four rounds of panning (19,21) were carried out using 500 ng of VEGF-R2/VEGF complex in the first round, 250 ng in the second round, and 100 ng of VEGF-R2/VEGF complex in the third and fourth rounds. To eliminate the selection of clones that bind to the human IgG1 Fc part of the recombinant human VEGF-R2/Fc fusion protein, 2.5 mg/ml human IgG (Pierce) was added to the phage preparations during selection. After the final round of panning, 10 isopropyl-1-thio-␤-D-galactopyranoside-induced clones from each library were analyzed for binding to 100 ng of immobilized VEGF-R2, human IgG, and bovine serum albumin by ELISA using a rat anti-HA monoclonal antibody conjugated to horseradish peroxidase for detection. Positive clones were analyzed by DNA fingerprinting and sequencing as described before (20). Soluble Fab were expressed from gene III fragment-depleted phagemid vector pComb3X and purified using goat anti-human F(abЈ) 2 N-hydroxysuccinimide resin columns as described (19). Library generation and selection for Tie-2 clones is described elsewhere (20). For a control construct, designated T2V2, one clone was selected from each generated library (VEGF-R2 and Tie-2) prior to selection. ELISA was done the same way as for detection of binders after panning, determining no binding to VEGF-R2 or Tie-2.
Surface Plasmon Resonance-Surface plasmon resonance for the determination of association (k on ) and dissociation (k off ) rate constants for the interaction of chimeric rabbit/human Fab with Tie-2 was performed on a Biacore instrument (Biacore AB, Uppsala, Sweden). A CM5 sensor chip (Biacore AB) was activated for immobilization with N-hydroxysuccinimide and N-ethyl-NЈ-(3-dimethylaminopropyl)carbodiimide according to the methods outlined by the supplier. Recombinant human VEGF-R2/Fc fusion protein was coupled at a low density (500 -1000 resonance units) to the surface by injection of 10 -20 l of a 10 ng/l sample in 20 mM sodium acetate (pH 3.5). Subsequently, the sensor chip was deactivated with 1 M ethanolamine hydrochloride (pH 8.5). Binding of chimeric rabbit/human Fab to immobilized human VEGF-R2 was studied by injection of Fab at five different concentrations ranging from 50 to 150 nM. PBS was used as the running buffer. The sensor chip was regenerated with 20 mM HCl and remained active for at least 20 measurements. The k on and k off values were calculated using Biacore AB evaluation software. The equilibrium dissociation constant K d was calculated from k off /k on .
Conversion of a VEGF-R2 and a Tie-2-specific Fab into a Singlechain Antibody Fragment (scFv)-Specific oligonucleotide primers (19,22) were used to amplify VH and VL gene segments from purified phagemid DNA of Fab VC06 and Fab 1S05. VL of VC06 and 1S05 were amplified with extompseq (5Ј-GCG GAG GAG CTT GCT AGC TGC GAG AAG ACA GCT ATC GCG ATT GCA TGT) and RJO-BL. VH of VC06 and 1S05 were amplified with RSCVH4 or RSCVH3, respectively, and HSCG1234-B. Overlap extension PCR was done using primers ext and RSC-B. The resulting overlap-PCR product encodes an scFv in which the C-terminal VL region is linked to the N-terminal VH region through a peptide linker (SSGGGGSGGGGGGSSRSS). Control construct JC7U (against integrin ␣ v ␤ 3 ) was constructed as described (22). The scFv encoding sequences were cloned into phagemid vector pComb3X using asymmetric SfiI sites and binding activity of the expressed scFv was confirmed by ELISA.
Assembly of Intrabody Constructs in pAdTrackCMV and Generation of Adenoviral Plasmids by Homologous Recombination-Intrabody coding regions were initially assembled in pBabePuro essentially as described (23). In these constructs, the scFv coding regions are flanked by a human light chain leader sequence at the 5Ј-end and a sequence encoding the HA tag (YPYDVPDYA) and the ER retention signal (KDEL) at the 3Ј-end. The intrabody coding regions were then excised by digestion with BamHI and SalI and ligated into pAdTrackCMV (24) digested with BglII and SalI. This adapter fragment contains also compatible SfiI sites, which were used for cloning the different intrabodies (intradiabody against Tie-2/VEGF-R2, control intradiabody T2V2, scFv intrabody against Tie-2, and scFv intrabody against VEGF-R2) into the adenovirus vector. The generation of recombinant adenoviruses was done essentially as described (24). High titer viral stocks were produced and purified by CsCl banding. All virus preparations were GFP-corrected (25).
Immunocytochemical Analysis of Antigen/Intrabody Colocalization-For analysis of Tie-2, VEGF-R2, and intrabodies on GFP-positive HU-VEC, cells were seeded on collagen-coated Lab-Tek coverglasses and infected with 50 MOI of adenovirus encoding intradiabody or control intrabody (JC7U). Forty-eight hours postinfection, HUVEC were washed with copious amounts of PBS, followed by incubation in a humidifying chamber at room temperature for 1 h with a primary antibody mixture of rat anti-HA monoclonal antibody (5 g/ml; Roche Applied Science) and biotinylated goat anti-human Tie-2 or anti-human VEGF-R2 polyclonal antibodies (R&D Systems). The cells were then stained for 1 h at room temperature with the mixture of Cy5-conjugated FIG. 2. Intrabody and Tie-2/VEGF-R2 colocalization on HUVEC. Maximal projections of three-dimensional data sets show the total cellular distribution of Tie-2 (a-f) and VEGF-R2 (g-l). HUVEC cells were infected with an MOI of 50 with the intradiabody against Tie-2 (a-c) and VEGF-R2 (g-i) and a control intrabody against integrin The left panels show the target antigen in red; middle panels show intrabodies in blue; right panels show overlapping antigen (red) and intrabody (blue) signals in purple and nuclei in green. Antigen/intrabody colocalization was observed in intradiabody-infected cells (c and i, purple). By contrast, a homogenous antigen distribution was observed in control intrabody-infected HUVEC (f and l, red and blue).
donkey anti-rat IgG polyclonal antibodies and streptavidin/rhodamine red-X (both from Jackson Immunoresearch, West Grove, PA) diluted to 1:100 in FACS buffer, 0.1% saponin. Finally, the cells were covered with SlowFade Antifade reagent. Three-color (GFP, rhodamine red-X, and Cy5) three-dimensional data sets were collected with a DeltaVision system (Applied Precision, Issaquah, WA); this consisted of an Olympus IX-70 fluorescence microscope, a motorized high precision xyz stage, a 100-watt mercury lamp, and a KAF1400 chip-based cooled chargecoupled device camera. Exposure times were 0.2-0.5 s (2-binning), and images were obtained with a ϫ60 oil objective. Three-dimensional reconstructions were generated by capturing 150-nm serial sections along the z axis. Images were deconvolved (based on the Agard-Sadat inverse matrix algorithm) and analyzed with softWorX version 2.5.
Pulse-chase for Determination of Half-life-HUVEC cells were seeded at a density of 1 ϫ 10 6 cells in T175 flasks and infected 24 h later with recombinant adenoviruses using an MOI of 10. After infection for 24 h, cells were washed with Hepes-buffered saline solution and trypsinized. Cells were starved for 2 h in 10 ml of serum-free, methionine-free, cysteine-free minimal essential medium at 37°C and swirled periodically. Samples were then labeled with Tran 35 S-label medium (50 Ci/ml; ICN, Aurora, OH) for 2 h at 37°C and subsequently chased with EGM medium, containing a 40-fold excess of methionine and 20-fold excess of cysteine for various time points (0, 4, 8, 16, 24, and 48 h). At each time point, cells were washed once with ice-cold PBS containing 1 mM phenylmethylsulfonyl fluoride and lysed (Promega lysis buffer, containing complete protease inhibitor mixture). Supernatants were collected and stored at Ϫ80°C. Samples were precipitated using protein G and monoclonal anti-HA antibody (Covance) for scFv or protein G alone for the diabody. Immunocomplexes were washed twice with lysis buffer and twice with TBS, boiled in 1.5ϫ SDS loading buffer, and separated by a 4 -20% gradient SDS-PAGE gel under reducing conditions. The gels were stained, dried, and exposed to autoradiography as well as quantitatively analyzed with a PhosphorImager.
Endothelial Cell Tube Formation Assay-6 ϫ 10 4 HUVEC cells were infected with recombinant adenovirus using an MOI of 20 in an Eppendorf tube for 45 min and were then transferred to 6-well plates. Three days after infection, cells were washed with Hepes-buffered saline solution, trypsinized, and counted. 2 ϫ 10 4 cells/well (volume 50 l) in complete EGM medium were seeded in triplicates for each virus in a 96-well plate coated with Matrigel Basement Membrane Matrix (BD Bioscience, Bedford, MA) and incubated for 15.5 h. Cells were then stained and fixed with the Diff-Quik® staining set (DADE BEHRING Inc., Newark, DE). For this, cells were fixed with Diff-Quik Fixative, followed by Diff-Quik Solution I and II, each with 100 l for 2-3 min. Cells were then washed four times with distilled H 2 O, and pictures were taken under an inverted light microscope at ϫ2 magnification. The number of tube branches for each virus was counted in triplicates to calculate the average Ϯ S.D.
Cell Proliferation Assay-HUVEC cells were seeded (5 ϫ 10 4 ) in 96-well plates in complete EGM medium and infected with different recombinant adenoviral constructs with an MOI of 10 and 50. Three days later, the cells were trypsinized, stained with trypan blue, and counted in triplicate by a hemacytometer. Cell proliferation was also measured by adding ␤-counter (Beckman Coulter). The background was defined by uninfected cells. The inhibition was calculated according to the following formula: (background Ϫ infected cell count)/background ϫ 100%.

Selection of Rabbit Fabs
Binding to VEGF-R2-Using phage display, several New Zealand White rabbit Fabs against VEGF-R2 were selected by panning (19) chimeric rabbit/human antibody libraries (20, 21) on a human VEGF-R2⅐VEGF complex. For further analysis, chimeric rabbit/human Fab VC06 was selected and produced as soluble Fab in Escherichia coli and purified by affinity chromatography using goat antihuman F(abЈ) 2 N-hydroxysuccinimide resin columns. Fab VC06 demonstrated a strong binding to human VEGF-R2 in ELISA and to HUVEC in flow cytometry (not shown). Thus, the VEGF-R2 epitope recognized by VC06 is displayed by native VEGF-R2 expressed on the cell surface and is an accessible target for antiangiogenic therapy. Surface plasmon resonance studies of VC06 revealed that it possessed a monovalent dissociation constant of ϳ1 nM to human VEGF-R2, whereas Fab 1S05, which we described earlier (20), bound with a dissociation constant of 14 nM to human Tie-2 ( Table I).
Generation of an Intradiabody against Tie-2 and VEGF-R2 and scFv Intrabodies against Tie-2 or VEGF-R2 Alone-Fabs were converted into scFv, in which the VL and VH fragments were covalently linked with a peptide linker consisting of 18 amino acids. Preserved binding to their respective antigens was confirmed for both scFv VC06 (VEGF-R2) and scFv 1S05 (Tie-2) by ELISA (not shown). Next, scFv 1S05 and scFv VC06 were linked through the second and third heavy chain constant domains of human IgG1, resulting in a scFv-CH2-CH3-scFv expression cassette. As a key feature, the scFv-CH2-CH3-scFv expression cassette provides for the production of a bifunctional tetravalent antibody construct from a single polypeptide.
Through the homophilic interaction of CH3, two scFv-CH2-CH3-scFv molecules associate to form a 150-kDa dimer, which displays both the N-terminal and C-terminal scFv module bivalently as the intradiabody (Fig. 1). The scFv against Tie-2 or VEGF-R2 and the scFv-CH2-CH3-scFv genes were cloned into a modified adenovirus shuttle vector pAdTrackCMV (24) using two asymmetric SfiI sites. In these constructs, the scFv coding regions are flanked by a human light chain leader sequence at the 5Ј-end, and a sequence encoding the HA tag (YPYDVPDYA) and the ER retention signal (KDEL) at the 3Ј-end. The generation of replication-deficient recombinant adenoviruses was done essentially as described (24). The resulting recombinant adenoviruses were purified by CsCl banding, and final yields were between 4.3 ϫ 10 11 particles/ml and 1.2 ϫ 10 12 particles/ml.
Colocalization of Intradiabody and Targeted Proteins in the ER-To investigate whether our intrabody constructs were expressed in the ER and able to trap the targeted proteins in the same compartment, we infected HUVEC with an MOI of 50 and verified the endoplasmatic reticulum localization of intrabodies and proteins by deconvolution microscopy on day 3 after infection (Fig. 2). As a control, HUVEC were infected with an intrabody against integrin ␣ v ␤ 3 (scFv JC7U). To visualize the intrabodies in the ER, intracellular staining was carried out with saponin and rat anti-HA monoclonal antibody, followed by donkey anti-rat Cy5 polyclonal antibodies. Both intrabody constructs were found to be expressed in the ER, as indicated by a staining of the characteristic tubular network. To confirm that the targeted proteins were trapped in the ER, another staining protocol using biotinylated Tie-2 and VEGF-R2 polyclonal antibodies, followed by streptavidin-rhodamine red-X polyclonal antibodies was performed. As shown in Fig. 2  body against Tie-2/VEGF-R2 was found to be co-localized with both targeted proteins in the ER, indicated by the purple color in the merged picture. On the other hand, the control intrabody (JC7U), although expressed in the ER, did not retain either Tie-2 or VEGF-R2.
Extended Surface Depletion of Tie-2 and VEGF-R2 with Intradiabody-To compare the effect of the intradiabody with that of the scFv intrabodies on the surface expression of Tie-2 and VEGF-R2, HUVEC were infected with an MOI of 10 and analyzed by flow cytometry over a period of 15 days. The surface expression of human Tie-2 and VEGF-R2 on HUVEC infected with the intradiabody, was specifically blocked up to 57% (Tie-2) and 78% (VEGF-R2), respectively, on day one. On day three after infection, blockade of Tie-2 and VEGF-R2 was 98% complete. Fifteen days after infection, surface expression remained efficiently blocked, 97.5 and 96% for Tie-2 and VEGF-R2, respectively. In comparison, the surface expression of HUVEC infected with the scFv intrabodies, were blocked with 80% (Tie-2) and 83% (VEGF-R2) efficiency on day one and 84% (Tie-2) and 90% (VEGF-R2) on day three postinfection. However, on day 15, surface depletion was only 68 and 5% for Tie-2 and VEGF-R2, respectively (Fig. 3, a and b). No downregulation of integrins ␣ v ␤ 3 and ␣ v ␤ 5 was observed during 15 days of the experiment (data not shown).
Extended Half-life of the Intradiabody-To determine whether the remarkable superiority of the intradiabody with respect to effectiveness and duration of surface depletion of the targeted proteins can be attributed to an extended half-life, pulse-chase studies were performed. In these experiments, we compared the intradiabody with the 1S05 scFv intrabody, which had been found to be more efficient than the VC06 scFv intrabody. In brief, 10 6 HUVEC cells were infected with an MOI of 10 with each virus for 24 h. Proteins were labeled with [ 35 S]methionine/cysteine, and HUVEC were lysed following different time points of chase. Following immunoprecipitation, proteins were separated on a 4 -20% gradient SDS-PAGE gel under reducing conditions, visualized by autoradiography, and quantified using a PhosphorImager. Pulse-chase experiments revealed a half-life of the intradiabody beyond 48 h (t ϳ230 h), whereas the half-life of the scFv intrabody 1S05 was ϳ22 h (Fig. 4, a and b). The half-life of the scFv intrabody VC06 was even shorter than 22 h (data not shown).
Antiangiogenic Effect of Intradiabody-To test whether our intradiabody can inhibit angiogenesis in vitro, we used a threedimensional capillary tube formation assay. This in vitro assay has been used as a model of the early organization of new blood vessels and is consistently found to recapitulate to a large extent the process of angiogenesis in vivo (26,27). HUVEC were infected with an MOI of 20 of each virus. Cells were transferred to the Matrigel-coated 96-well plate 3 days after infection to achieve maximum surface depletion of the targeted proteins. All cells were incubated in the presence of exogenous growth factors and serum (complete EGM medium). Controls consisted of mock-infected cells (empty vector) and a control intradiabody without specificity for any known target (T2V2). These controls were used to determine possible toxic or growth inhibition (antiproliferative) effects of either the vector or the intradiabody format. As shown in Fig. 5, formation of capillaries in Matrigel was inhibited 90 Ϯ 3% with the intradiabody (VEG-R2/Tie-2), 44 Ϯ 4.3% with the scFv intrabody VC06 (VEGF-R2), and 11 Ϯ 5.8% with the scFv intrabody 1S05 (Tie-2). Significantly, cells infected with the empty vector and the control intradiabody (T2V2) were able to form capillaries to the same extent as uninfected cells.
Intradiabody Does Not Affect Cell Viability-In order to determine that the antiangiogenic effects of the intradiabody are not related to cytotoxic effects on HUVEC, cell proliferation and survival were assessed via [ 3 H]thymidine incorporation assay and trypan blue staining. HUVEC were infected with an MOI of 10 and 50 of each virus. Cell viability was assessed 3 days after infection. All cells were incubated in complete EGM medium. As shown in Fig. 6, cell survival and proliferation of cells infected with the intradiabody are similar as for the scFv intrabodies or the control intradiabody. We demonstrate hereby that accumulation of the bispecific intradiabody in the ER does not affect cell viability to a greater extent than scFv intrabodies or the control intradiabody and exclude that the antiangiogenic effects are related to cytotoxic effects caused by ER stress. DISCUSSION The use of scFv fragments as intrabodies has received considerable attention over the past 10 years. Intrabodies have been utilized to neutralize the function of endogenous target proteins using several different strategies (13,28,29). Among these, the misdirected localization of the target to another subcellular region features as the most popular methodology employed. We hypothesized that an ER-targeted intrabody with dual specificity and valency could be generated by fusing two scFv modules with unique specificity to the N terminus and C terminus of an Fc domain, respectively. Although there are reports of presenting functional bispecific antibodies in scFv-Fc formats (30,31), none of those antibody constructs have been studied as functional intrabodies targeted to the ER. Furthermore, these antibody constructs bind in a monovalent fashion to each of the target proteins. As a prototype and proof of concept, we demonstrate here the development of an ER-targeted intradiabody format for the simultaneous down-regulation of two independent cell surface receptors using an adenovirus-mediated gene delivery system. Our prototypical intradiabody targets the endothelial transmembrane tyrosine kinase receptors Tie-2 and VEGF-R2. Both of these receptors have been shown to be key in angiogenesis.
The endothelial cell receptor-tyrosine kinases, VEGF-R2 and Tie-2, as well as their ligands, VEGF and angiopoietin-1 and -2, respectively, play key roles in tumor angiogenesis. A model for the interplay of VEGF, VEGF-R2, Tie-2, and Ang-1 and -2 in tumor angiogenesis is discussed by Holash et al. (17). From a therapeutic perspective, the Tie-2/Ang-2 complex and the VEGF/VEGF-R2 complex, which are formed on the surface of proliferating endothelial cells, are attractive targets for antiangiogenic agents. Several reports (32,33) have been published showing the inhibition of tumor growth by interference with the VEGF or Tie-2 receptor pathway by means of antibodies and soluble or dominant-negative receptor domains (34,35). As of yet, no reports have been published based on the use of an ER-targeted intrabody to inhibit expression of the two endothelial transmembrane receptors Tie-2 and VEGF-R2, either separately or simultaneously. Our results demonstrate the feasibility of expressing functional, bispecific, and tetravalent antibodies intracellularly and the possibility of inhibiting the transit of two integral membrane proteins simultaneously through the ER by means of diabody expression. The potential for inhibiting angiogenesis by this approach should have implications for the treatment of cancer, since VEGF-R2 and Tie-2 are essential for tumor angiogenesis. In vitro angiogenesis assays using Matrigel showed that the intradiabody significantly reduces the number of capillary tubes formed, demonstrating its potential as an antiangiogenic drug. Interestingly, monospecific scFv intrabodies against VEGF-R2 or Tie-2 inhibited capillary tube formation to a significantly lesser extent. Moreover, the toxicity of the bispecific intradiabody and the control intradiabody (T2V2) are comparable. The relative efficiency of the intradiabody over the traditional intrabody-based approach could be explained by several reasons: first, the bispecificity of the intradiabody versus the monospecificity of the scFv intrabodies; second, simply their monovalency versus the bivalency of the intradiabody. A third possibility is that the relative stability of the proteins could also account for these findings. Fourth, the bivalent display of the ER retention signal may lead to better retention of the targeted proteins. We found that the superiority of the intradiabody with respect to effectiveness and duration of surface depletion of the targeted proteins can be attributed to an extended half-life. This was demonstrated in pulse-chase studies, which revealed a half-life of the intradiabody beyond 48 h (t ϳ230 h), whereas the halflife of the scFv intrabody 1S05 was ϳ22 h and the half-life of scFv intrabody VC06 was even shorter (data not shown). The extended half-life of the intradiabody correlates well with its more efficient and durable surface depletion of the targeted proteins compared with the scFv intrabody. A plausible explanation for the extended half-life of the intradiabody format is the bivalent display of the ER retention signal (KDEL) through the homodimerization of the constant domains (Fig. 1). This may increase the efficiency of recovery of the intradiabody from the Golgi by the KDEL receptor (36). In addition, fusion to constant domains per se has been suggested as a means of increasing the stability of intrabodies (37). Although the increased valency of the intradiabody, which results in a higher apparent affinity (avidity) for the antigen, is probably an important factor for the efficient surface depletion we observed, other factors might contribute to the superiority of the intradiabody. In particular, it is known that intrabodies with extended half-lives achieve higher steady-state expression levels, which increases effectiveness in target molecule inactivation. In fact, intracellular stability can be more important than affinity for the antigen (28,38,39).
A major application for intrabodies lies in functional genomics (3,5,6). Since the number of proteins with unknown function and their interactions with other proteins steadily increases, bispecific and tetravalent intrabodies could provide an attractive proteomic tool to analyze these networks. This novel tool provides for the simultaneous surface depletion of two independent receptors while improving both the efficiency and duration of the resulting phenotypic knockout. Intradiabodies may be used to investigate linked regulatory pathways or block redundant pathways. In addition, intradiabodies may extend the therapeutic applicability of intrabodies. For example, our prototype developed in this study provides a precise tool for the simultaneous silencing of two independent signaling pathways essential for angiogenesis (18).