|Home | About | Journals | Submit | Contact Us | Français|
Here we present a bispecific antibody (bsAb) format in which a disulfide-stabilized scFv is fused to the C-terminus of the light chain of an IgG to create an IgG-scFv bifunctional antibody. When expressed in mammalian cells and purified by one-step protein A chromatography, the bsAb retains parental affinities of each binding domain, exhibits IgG-like stability and demonstrates in vivo IgG-like tumor targeting and blood clearance. The extension of the C-terminus of the light chain of an IgG with an scFv or even a smaller peptide does appear to disrupt disulfide bond formation between the light and heavy chains; however, this does not appear to affect binding, stability or in vivo properties of the IgG. Thus, we demonstrate here that the light chain of an IgG can be extended with an scFv without affecting IgG function and stability. This format serves as a standardized platform for the construction of functional bsAbs.
While monoclonal antibodies have shown success in the clinic for a variety of diseases (Hudson and Souriau, 2003), multi-specific antibodies, with an ability to bind to more than one target, may further improve clinical efficacy via novel mechanisms. Multi-specific antibodies have been engineered for a variety of applications including enhanced antibody-dependent cell-mediated cytotoxicity (ADCC) (Garcia de Palazzo et al., 1992; McCall et al., 2001), tumor surface-receptor blocking and downregulation (Lu et al., 2005), simultaneous binding to two soluble effector molecules (Wu et al., 2007) and pretargeting tumor cells for the subsequent capture of radionuclides (Boerman et al., 2003), drugs (Ford et al., 2001) and prodrugs (Bagshawe, 2006).
Early efforts to produce bispecific antibodies (bsAbs) included chemical conjugation of two antibodies or fragments thereof (Graziano and Guptill, 2004) or co-expression of two antibodies with different specificities through the hybrid hybridoma technique (Menard et al., 1989). Unfortunately, the conditions required for chemical conjugation can inactivate, unfold or aggregate the bsAb, while the hybrid hybridoma technique not only produces the desired bsAb but also undesired products from mispairing necessitating complex purification schemes. In 1996, Carter and colleagues described a ‘knobs into holes’ method, wherein different but complementary mutations introduced into the CH3 domains favor heterodimerization. In the past decade, several other formats of multi-specific antibodies have been synthesized by recombinant methods to produce scFv fusions or diabodies, scFv Fc fusions and single variable domain IgGs, among others (Coloma and Morrison, 1997; Kontermann, 2005; Marvin and Zhu; 2005, Shen et al., 2007; Wu et al., 2007). In addition, Goldenberg et al. (2008) have developed a ‘dock-and-lock’ method for creating multi-specific antibodies, in which one antibody fragment is fused to a peptide regulatory subunit of cAMP-dependent protein kinase and another antibody fragment to a peptide-anchoring domain of A kinase anchor protein, where the two peptides have natural affinity for each other (Rossi et al., 2006). A recent format that appears to possess good stability is the dual-variable-domain IgG that extends both the heavy chain and the light chain with the N-terminal addition of a second set of variable domains; however, the physical proximity of the two binding sites may be sterically problematic for certain pairs of antigen-binding domains (Wu et al., 2007). In addition, the authors mention that significant construct optimization is often required to preserve the parental affinities as both the orientation of the variable domains and the linkers between them appear to be critical to function and expression.
We present here a novel bsAb design as a C-terminal fusion of a disulfide-stabilized scFv to the light chain of an IgG. It can be expressed in mammalian cell culture and purified to homogeneity by protein A chromatography. We are interested in developing bsAbs for pretargeted imaging and radioimmunotherapy applications. Here, we test three different versions of this new bsAb construct with specificity to different cell surface protein targets and small molecule haptens. Simultaneous binding, affinity and in vitro stability are assessed, as are in vivo blood clearance and tumor targeting.
The bispecific format was designed as an scFv fusion to the C-terminus of the light chain of an IgG. The heavy chain is the same as that of human IgG1 and was subcloned into the mammalian expression vector gwiz, purchased from Aldevron (Fargo, ND). The light chain is constructed as leader-FLAG-VL-Cκ-(Gly4Ser)2-scFv-cmyc, where VL is the variable light domain, Cκ is the kappa light chain constant domain and FLAG and cmyc are the N- and C-terminal epitope tags, respectively. It was cloned into a separate gwiz plasmid. Both plasmids were transiently co-expressed in HEK293 cells (cat. no. R790-07) purchased from Invitrogen (Carlsbad, CA). HEK293 cells were grown in flasks on an orbital shaker platform rotating at 140 rpm at 37°C, 5% CO2 and subcultured following the manufacturer's protocol. Co-transfection was performed with polyethyleneimine (PEI) as the transfection reagent. Briefly, HEK293 cells were subcultured to a cell density of 0.5–0.7 × 106 cells/ml 24 h before transfection. Immediately before transfection, cell density was adjusted to 1 × 106 cells/ml. Five hundred micrograms of each purified plasmid (1 mg/ml) was added to 19 ml Optipro (Invitrogen). Two milliliters of 1 mg/ml PEI, pH 7.0 (molecular weight (MW) of 25 000) purchased from Polysciences (Warrington, PA) dissolved in water was added to 18 ml Optipro. Both the solutions were incubated at room temperature for 5 min. The DNA/Optipro solution was added to the PEI/Optipro solution and incubated for 10 min at room temperature and added drop wise to 1 l HEK293 culture. The supernatant was collected 6–8 days after transfection. Antibodies were purified by protein A chromatography (Thermo Fisher Scientific, Rockford, IL) following the manufacturer's instructions.
Specific constructs were made by overlap extension PCR and site-directed mutagenesis. The Sm3e/C825 bsAb was cloned and produced as described above using the variable heavy (VH) and VL domains from the affinity-matured anti-carcinoembryonic antigen (CEA) Sm3e scFv (Graff et al., 2004) and the disulfide-stabilized C825 scFv with picomolar affinity to metal chelates of 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) (Orcutt et al., submitted for publication) affinity matured from the wild-type 2D12.5 antibody (Corneillie et al., 2003). The Sm3e/4m5.3 bsAb substituted a 4m5.3 scFv that is a femtomolar fluorescein binder (Boder et al., 2000) disulfide stabilized by introducing two cysteine residues, S43C in the VL domain and Q105C in the VH domain (Reiter et al., 1996). The A33/4m5.3 bsAb uses the VH and VL domains from an A33 humanized Fab fragment (Rader et al., 2000) for the IgG-binding domains. Sm3e IgG and A33 IgG plasmids were produced by introducing two stop codons in the light chain immediately following the Cκ sequence via QuickChange (Stratagene, La Jolla, CA). The C-terminus of the A33 IgG light chain was extended by an 18 amino acid peptide ((G4S)2LPETGGSG), to make the construct A33 IgG + peptide. A33 IgG + peptide was disulfide stabilized by introducing two different pairs of cysteine residues, VL G100C and VH G44C (ds1) and VL V43C and VH Q105C (ds2) (Reiter et al., 1996). The supplemental material contains sequence information for all constructs.
Culture expression media was analyzed by western blot using a horseradish peroxidase (HRP) conjugated goat anti-human Fc antibody (Thermo Fisher Scientific) and an HRP-conjugated goat anti-human κ light chain antibody purchased from AbD Serotec (Raleigh, NC). Purified bsAb and IgG were analyzed by SDS-PAGE and quantified by absorbance at 280 nm.
DOTA-biotin was synthesized by adding p-SCN-Bn-DOTA (S-2-(4-isothiocyanatobenzyl)-1,4,7,10-tetraazacyclododecane-tetraacetic acid) purchased from Macrocyclics (Dallas, TX) to Amine-PEG3-Biotin (Thermo Fisher Scientific) in dimethyl sulfoxide (DMSO) purchased from Sigma Aldrich (St. Louis, MO) with a 10-fold molar excess of triethylamine (TEA) purchased from VWR (West Chester, PA). The reaction mixture was vortexed at room temperature for 3 h, and purified by high-performance liquid chromatography (HPLC) on a C-18 reverse-phase column (Agilent Model 1100 HPLC, 1 × 25 cm, buffer A = 0.05% trifluoroacetic acid (TFA), buffer B = 0.0425% TFA in 80% acetonitrile, 2–100% B gradient for 98 min). Flow through was monitored by absorbance detection at 280 nm. Fractions containing DOTA-biotin were confirmed using matrix-assisted laser desorption/ionization–time of flight (MALDI-TOF) mass spectrometry (Applied Biosystems Model Voyager DE-STR).
DOTA-647 was synthesized by adding 1 mM p-SCN-Bn-DOTA to 1 mM Alexa Fluor 647 cadaverine (Invitrogen) in DMSO with 40 mM TEA overnight at room temperature, with rotation. DOTA-647 was purified by HPLC as described above. Yttrium complexes of DOTA conjugates were prepared by incubating a molar excess of yttrium nitrate hexahydrate (Sigma Aldrich) with the DOTA conjugates in 0.4 M sodium acetate pH 5.2 buffer overnight at room temperature. The pH was adjusted to 7.0 with 10 M sodium hydroxide and the solution was diluted with PBS with 0.1% bovine serum albumin (PBSA).
Fluorescein-647 (Fl-647) was synthesized by adding 1 mM fluorescein-5-EX, succinimidyl ester (Invitrogen) to 1 mM Alexa Fluor 647 cadaverine in DMSO with 40 mM TEA and rotating overnight at room temperature and used without further purification.
105 trypsinized LS174T cells were washed with 500 µl PBSA and incubated with 50 nM bsAb or IgG for 1 h at room temperature. Cells were subsequently incubated with 100 nM fluorescein (Fl) purchased from Invitrogen, 100 nM DOTA-biotin chelated with yttrium (DOTA-Y-biotin), 50 nM bsAb or 100 µl PBSA, followed by incubation with 20 nM DOTA-Y-647 or FITC-647 and analysis by flow cytometry. All incubations were performed for 30 min on ice followed by washing once with PBSA unless otherwise indicated.
CEA-binding affinities for the Sm3e/C825 bsAb and Sm3e IgG were measured using fixed LS174T cells incubated with varying concentrations of bsAb or IgG overnight at 37°C. Cells were subsequently incubated with a 1:200 dilution of protein A Alexa Fluor 647 (Invitrogen) and analyzed by flow cytometry. The affinity of the Sm3e/C825 bsAb for DOTA-yttrium (DOTA-Y) was measured by incubating 100 nM bsAb on the surface of fixed LS174T cells for 1.5 h at room temperature. Cells were incubated with varying concentrations of DOTA-647 loaded with yttrium (DOTA-Y-647) for 2 h at 37°C before analysis by flow cytometry. Yeast expressing the C825 scFv on their surface were grown and induced as described (Chao et al., 2006). Cells were incubated with varying concentrations of DOTA-Y-647 for 2 h at 37°C and analyzed by flow cytometry. All affinities are reported as mean ± standard deviation (SD) calculated from at least two replicates.
IgG purified from human plasma (Sigma Aldrich), and purified Sm3e IgG and Sm3e/C825 bsAb were analyzed by FPLC size exclusion chromatography (Pharmacia Biotech Superdex, 200 column) and monitored by absorbance at 280 nm and data were normalized.
IgG and bsAb were incubated in PBSA for various times at 37°C. Antigen-binding activity was analyzed with a magnetic bead flow cytometry assay. Twenty nanomolar of IgG or bsAb was incubated with Protein A beads (Invitrogen) at room temperature for 1 h. Beads were washed and incubated with 20 nM biotinylated CEA (Graff et al., 2004) or 620 nM A33-HIS6 followed by incubation with 1:200 dilution of anti-his Alexa Fluor 647 (Qiagen, Valencia, CA) or streptavidin Alexa Fluor 647 (Invitrogen), and analyzed by flow cytometry. A33-HIS6 was synthesized by transient HEK cell transfection as described above with 1 mg A33-HIS6 gwiz plasmid transfected per liter of culture. A33-HIS6 was purified by TALON metal affinity resin (Clontech, Mountain View, CA) following the manufacturer's protocol. The A33-HIS6 gwiz plasmid was constructed by cloning the A33 signal sequence and gene from a baculovirus plasmid (Joosten et al., 2004) and inserting a C-terminal hexahistidine.
Sm3e IgG and Sm3e/C825 bsAb were incubated in either PBSA or 50% mouse serum (Invitrogen) in PBSA for various times at 37°C. Binding activity was analyzed with a magnetic bead flow cytometry assay, where 8.1 × 106 biotin binder beads (Invitrogen) were incubated with 20 nM biotinylated CEA in 300 µl PBSA overnight at 4°C with rotation. 105 CEA-coated beads were washed and incubated with 50 nM IgG or bsAb for 1 h at room temperature followed by incubation with 20 nM DOTA-Y-647 or 1:100 dilution of chicken anti-cmyc (Invitrogen) followed by incubation with 1:200 dilution of goat anti-chicken Alexa Fluor 647 (Invitrogen). Beads were washed and analyzed by flow cytometry.
IgG and bsAb protein were conjugated to p-SCN-Bn-DTPA (Macrocyclics) as described (Cooper et al., 2006). Concentrated DTPA-labeled protein was incubated with ~1 mCi 111InCl3 (Cardinal Health, Dublin, OH) for 30 min at room temperature. The protein was diluted with 500 µl saline and concentrated to ~50 µl using Vivaspin 5000 MWCO spin columns (Sartorius Stedim Biotech, Aubagne, France). The 111In-labeled protein was diluted with 500 µl saline and concentrated twice more.
All animal handling was performed in accordance with Beth Israel Deaconess Medical Center Animal Research Committee guidelines. LS174T human colorectal carcinoma cells (CL 188) were obtained from American Type Culture Collection (Manassas, VA) and C6 rat glioma cells were obtained from Brian W. Pogue (Dartmouth Medical School). Both cell lines were maintained under standard conditions and confirmed to be negative for mycoplasma and mouse pathogens by the Yale Virology Laboratory. Xenografts were established in 5–6 week-old male NCRU-nu/nu mice (Taconic Farms, Hudson, NY) by subcutaneous injection of 1–2 × 106 LS174T cells into the left flank and 1–2 × 106 C6 cells into the right flank of each mouse. After 8–10 days, tumors were 0.1–0.5 g in mass. Five hundred of micrograms of 111In-labeled protein was injected intravenously in 100 µl saline. Blood samples of 10–15 µl were collected from the tail vein at various times and counted on a model 1470 Wallac Wizard (Perken Elmer, Wellesley, MA) 10-detector gamma counter. Mice were euthanized by intraperitoneal injection of pentobarbital, a method consistent with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association. Tumors were resected, washed in PBS, weighed and counted as described above.
We have designed a bsAb as a C-terminal fusion to the light chain of an IgG (Fig. 1). The heavy chain is identical to that of an IgG. The light chain is constructed by extending an IgG light chain with a C-terminal (Gly4Ser)2 linker followed by an scFv. In this study, the light chain is constructed with an N-terminal FLAG tag and C-terminal cmyc tag for characterization purposes. The fully assembled bsAb contains two heavy and two light chains, and is tetravalent with two IgG-binding domains and two scFv-binding domains.
We synthesized a bsAb of this format that binds to the CEA and to complexes of the metal chelator, DOTA. The Sm3e/C825 bsAb was constructed from an Sm3e IgG and a DOTA-binding scFv, C825, by cloning VH and VL domains from the picomolar affinity Sm3e scFv (Graff et al., 2004) into a plasmid containing human IgG1 constant heavy domains and kappa constant light chain domain. C825, an affinity matured DOTA-binding scFv that binds with picomolar affinity to yttrium, lutetium and gadolinium chelates (Orcutt et al., submitted for publication) was subsequently cloned into the light chain plasmid immediately following the C-terminus of the Cκ gene. The heavy and light chain expression plasmids were transiently co-transfected into HEK293 mammalian cells. Secreted antibody was purified from cell culture supernatants by protein A chromatography. Yields of both Sm3e IgG and Sm3e/C825 bsAb were ~5–7 mg/l.
Non-reducing SDS-PAGE of purified bsAb displays a species with an MW of ~200 kDa (Fig. 2A). Under reducing conditions, the bsAb gives rise to two bands, both at around 50 kDa, as the scFv fusion increases the MW of the light chain to ~50 kDa. Size exclusion chromatography of purified bsAb shows a single dominant peak with a small amount of higher MW species, similar to that observed for recombinant IgG and for IgG purified from human plasma (Fig. 2B).
bsAb function was tested using CEA-expressing LS174T cells incubated with Sm3e/C825 bsAb followed by Alexa Fluor 647-conjugated DOTA loaded with non-radioactive yttrium (DOTA-Y-647). These samples exhibited a positive 647 signal, demonstrating bsAb simultaneous binding of both cell-surface CEA and soluble DOTA-Y-647. Control samples in which CEA or hapten epitopes were blocked showed no significant fluorescence (Fig. 3A). Similarly, LS174T cells incubated with monospecific Sm3e IgG followed by DOTA-Y-647 exhibited no significant fluorescence.
We produced two additional bsAbs of the same format. Sm3e/4M53 binds to CEA and fluorescein and A33/4M5.3 binds to the A33 antigen, a colon cancer immunotherapy target (Welt et al., 1990), and fluorescein. The 4M5.3 scFv was constructed by disulfide stabilizing the femtomolar fluorescein binding scFv 4M5.3 (Boder et al., 2000). Introduction of the disulfide bond does not significantly affect the binding affinity of the scFv (data not shown). The A33/4M5.3 bsAb was made by fusing an A33 antibody to ds4m5.3. The A33 IgG was constructed with VH and VL domains from an A33 humanized rabbit Fab (Rader et al., 2000). Both the Sm3e/4M5.3 (Fig. 3B) and A33/4M5.3 (Fig. 3C) bsAbs exhibited simultaneous binding to tumor surface antigen and soluble hapten.
It was observed that when both IgG and bsAb were boiled prior to western blot analysis, the molecules dissociate and exhibit a laddering effect with bands observed at MWs that correspond to the fully assembled molecule with two heavy chains and two light chains (H2L2) and partially assembled molecules (H2L and H2) (data not shown). The laddering of the bsAb is more significant compared with that of the IgG, likely due to more incomplete disulfide bond formation between the light and heavy chains of the bsAb due to the fusion of the scFv domain to the C-terminal Cκ cysteine.
To test this theory, we synthesized the A33 IgG with a light chain C-terminal Gly4Ser-based 18 amino acid peptide, which exhibited the same laddering effect as the bsAb under boiling conditions (Fig. 4). In order to restore covalent linkage between the heavy and light chains, a disulfide bond was introduced between the A33 IgG VL and VH domains (the IgG portion of the bsAb). Two different pairs of cysteine mutations were tested (Reiter et al., 1996). This alternative site of disulfide stabilization significantly reduced the laddering effect such that it was similar to that observed for the IgG and suggests that variable domain disulfide stabilization of the IgG allows formation of a covalent bond between the heavy and light chains that can substitute for the normal position at the Cκ C-terminus.
The affinity of IgG- and scFv-binding domains for the Sm3e/C825 bsAb was measured using cell-binding assays. The bsAb-binding affinity for CEA expressing LS174T cells was similar to that of the Sm3e IgG and the scFv-binding affinity for soluble DOTA-Y-647 was similar to that of the scFv alone, expressed on the surface of yeast (Table I).
Likewise, the affinity of the A33/4M5.3 bsAb to A33 antigen expressing LS174T cells was measured to be 647 ± 108 pM, within 2-fold to that measured for the A33 IgG, 303 ± 9 pM.
The thermal stability of the bsAb construct was tested by incubation at 37°C in PBSA. Both Sm3e/C825 and Sm3e/4M5.3 bsAbs exhibited similar CEA-binding activity as the Sm3e IgG after 7 days (Fig. 5A). A33/4M5.3 also exhibited similar stability as the A33 IgG (Fig. 5B).
Serum stability of the Sm3e/C825 bsAb was tested by incubation at 37°C in 50% mouse serum and compared with bsAb incubated in PBSA. After 7 days in PBSA, bsAb bound to CEA-coated beads showed no loss of the C-terminal cmyc tag and retained about 90% of its DOTA-binding activity (Figs 5C and and5D),5D), indicating thermal stability of the scFv and retention of assembled bsAb with simultaneous binding function. After 7 days in serum, the bsAb retained approximately 90% of the cmyc tag, and about 60% of initial DOTA-binding activity, indicating proteolytic stability but lower scFv stability in serum.
111In-labeled Sm3e/C825 bsAb was injected intravenously into nude mice bearing both LS174T and C6 tumors. LS174T cells express ~4 × 105 CEA antigen per cell (Thurber and Wittrup, 2008). C6 tumors have been used previously as an internal CEA negative control (Kenanova et al., 2005). Blood samples were taken at various times over 24 h and the activity was measured by gamma counting. The blood half-life of the Sm3e/C825 bsAb is very similar to that of the Sm3e IgG (Fig. 6), indicating in vivo stability of the bispecific construct and that the addition of the scFv does not interfere with FcRn binding (Olafsen et al., 2006). The tumor uptake of the Sm3e/C825 bsAb and Sm3e IgG were also measured at 24 h and found at be very similar to 19 ± 3 and 15 ± 3%ID/g for the CEA positive LS174T tumors, respectively. The tumor uptake in the CEA negative C6 tumors was 5.3 ± 1.7 and 4.0 ± 1.6%ID/g for the bsAb and IgG, respectively.
We have engineered a novel bsAb construct as an scFv fusion to the C-terminus of the light chain of an IgG. Fusing the scFv in this way should minimize the steric hindrance that could obstruct simultaneous binding of both target antigens that might result from an N-terminal fusion to the light and/or the heavy chain. To date, we have synthesized several versions of the construct with various IgG and scFv domains, and all molecules bind simultaneously to their respective targets and retain parental affinities within 2-fold. No linker-length optimization is required for expression and retention of scFv and IgG binding. The bispecific construct also exhibits IgG-like stability, blood clearance and in vivo tumor targeting. The bsAb construct appears to work generally to pair any stable and functionally expressing IgG and scFv into a bispecific format, while retaining IgG-like properties. However, it should be noted that all of the bsAb constructs tested in this study have IgG domains that bind to cell surface proteins and scFv domains that bind to haptens. While we believe that this bsAb construct will also work when the scFv specificity is a protein target due to flexibility in the Gly4Ser-based linker and in the hinge region of the IgG, this has yet to be tested.
Coloma and Morrison (1997) also used an scFv for introducing additional specificity to an IgG, by attaching it to the C-terminus of the heavy chain of an IgG3. They report excellent results obtaining fully assembled monomeric functional protein from transfectoma supernatants. However, the IgG-scFv fusion results in notably faster clearance in an in vivo mouse model compared with the parent IgG. This may be due to a decrease in FcRn binding possibly from steric hindrance of the attached scFv, or perhaps aggregation or instability driven by the scFv moiety.
We are interested in bsAbs for pretargeted radioimmunotherapy applications, in which the bispecific is administered first and allowed to localize to tumor cells before the addition of a second, radioactive molecule that only binds to the bispecific and otherwise clears rapidly from the body (Goodwin et al., 1988; Le Doussal et al., 1989; Chang et al., 2002; Boerman et al., 2003; Gruaz-Guyon et al., 2005). The first bsAbs for pretargeting used streptavidin/antibody conjugates and have shown promise at the proof-of-concept stage (Axworthy et al., 2000). However, recent pretargeted efforts have moved away from using streptavidin due to its immunogenicity, high kidney uptake (Wilbur et al., 2004), and issues with endogenous biotin (Hamblett et al., 2002). Several bispecific formats that do not utilize streptavidin have been used for pretargeted applications including chemically conjugated Fab and (Fab)2 formats (Kraeber-Bodere et al., 1999; Schuhmacher et al., 2001; Griffiths et al., 2004), hybrid hybridoma bsAbs (van Schaijk et al., 2003; van Schaijk et al., 2005), recombinant diabodies and scFv fusions (DeNardo et al., 2001; Rossi et al., 2005), and a dock and lock tri-Fab (Sharkey et al., 2008).
It is desirable to preserve the Fc region in a bsAb for pretargeting applications, which will result in prolonged plasma retention due to FcRn binding and hence increased tumor penetration (Olafsen et al., 2006; Thurber et al., 2008). Full IgG molecules have demonstrated significantly higher tumor accumulation compared with minibodies, diabodies and scFvs (Schneider et al., 2009). The addition of the scFv to the C-terminus of the IgG light chain does not impact blood clearance, indicating that the scFv does not affect FcRn binding to the bispecific construct. The Fc-binding domain may also retain FcγR binding, for ADCC and additional therapeutic benefit.
The long-circulating half-life of the bsAb will result in increased tumor uptake as discussed above but may also result in significant residual antibody retention in the blood at the time of hapten dosing. Thus, a clearing agent to quickly clear antibody from the blood prior to hapten dosing may help to accelerate hapten clearance from the body. Rapid hapten clearance is necessary for high tumor to background ratios for imaging and low off-target radiation for therapy.
The Sm3e/C825 bsAb retains approximately 90% of its cmyc tag at 37°C in serum after 7 days, indicating little if any protease cleavage of the Gly4Ser linker. However, the binding activity of the C825 scFv decreases to about 60% after 7 days, with a rapid decrease during the first 24 h followed by a plateau. The decreased binding activity does not appear to be pH mediated as the pH of the serum solution remains neutral for the length of the experiment (data not shown). Nor does it appear to be simply due to thermal instability, because 90% of binding is retained after similar incubation at 37°C in PBSA. Thus, the loss of activity may be due to serum protein binding or protein- or enzyme-assisted denaturation.
One aspect of the bsAb format is an intermolecular disulfide bond between the VH and VL domains of the scFv. The open conformation of an scFv can be prone to aggregation (Reiter et al., 1994; Worn and Pluckthun, 2001). Disulfide stabilization of the scFv should prevent the scFv from assuming an open conformation and hence reduce the risk of aggregation. In addition, as one would expect, the stability of the scFv in the bispecific format is limited by the stability of the parent scFv. Thus, it is important to select highly stable scFvs.
It is interesting to note that attaching an scFv, or even an 18 amino acid flexible peptide, to the C-terminus of the light chain appears to disrupt formation of the disulfide bond between the light chain and the heavy chain of the human IgG1. This disulfide bond naturally exists at the C-terminal cysteine of the Cκ domain in IgGs. It does not appear to be necessary for function or stability of the bsAb, as all bsAbs tested retain parental affinities and exhibit excellent serum stability, even after protein A purification with acidic pH elution. We nevertheless added a disulfide bond between the VH and VL domains of the A33 IgG with peptide to determine if an interdomain disulfide bond can be introduced in this region to reform a covalent linkage between the LC and HC. Indeed, both disulfide stabilized versions of the A33 IgG with peptide exhibited significantly reduced dissociation after boiling, confirming that a covalent bond between the LC and HC can be reformed. While in this study, a covalent linkage between the heavy and light chains does not appear necessary for bsAb function or stability, it is possible that other IgGs may be less stable and require interdomain disulfide stabilization for stable bsAb construction.
While we have designed this bispecific format to be used for pretargeting approaches, this platform may be beneficial for other applications requiring bsAbs. We demonstrate that the light chain of an IgG can be extended with an scFv without affecting IgG function and stability. Other proteins or peptides, such as affibodies (Nygren, 2008), single variable domains (Harmsen and De Haard, 2007) and peptide toxins could be attached to IgGs in this site specific manner, to yield homogenous IgG fusion products for targeted delivery. This platform could be used in a straightforward fashion to modify current FDA-approved antibodies to add additional functionalities. Production and purification should scale directly with current antibody manufacturing methods. As a robust modular platform, this bsAb format obviates the need for extensive optimization of each new combination of binding domains and retains IgG-like properties both in vitro and in vivo.
This work was supported by National Institutes of Health [grant number RO1-CA-101830] and a National Science Foundation Graduate Research Fellowship [KDO].
We thank Steven Sazinsky and Michael Schmidt for their helpful discussions and Vladamir Voynov for his assistance with HEK293 cell culture.
Edited By Dennis Burton