Georg H. Fey, Chair of Genetics, University of Erlangen-Nuremberg, Staudtstrasse 5, D-91058 Erlangen, Germany. E-mail: email@example.com
Bispecific antibodies offer the possibility of improving effector-cell recruitment for antibody therapy. For this purpose, a recombinant bispecific single-chain Fv antibody (bsscFv), directed against FcγRIII (CD16) and human leucocyte antigen (HLA) class II, was constructed and tested in functional assays. RNA from the hybridomas 3G8 and F3.3, reacting with CD16 and HLA class II, respectively, was used to generate phage display libraries. From these libraries, reactive phages were isolated and the bsscFv was constructed by connecting both single-chain Fv components through a 20 amino acid flexible linker. After expression in SF21 insect cells and chromatographic purification, the bsscFv bound specifically and simultaneously to both antigens. The affinities of the anti-CD16 and the anti-HLA class II scFv components of the bsscFv were 8·6 × 10−8 mol/l and 13·7 × 10−8 mol/l, respectively, which was approximately sevenfold lower than the F(ab) fragments of the parental antibodies. In antibody-dependent cellular cytotoxicity experiments with human mononuclear cells as effectors, the bsscFv-mediated specific lysis of both HLA class II-positive, malignant human B-lymphoid cell lines and primary cells from patients with chronic B-cell lymphocytic leukaemia. Optimal lysis was obtained at bsscFv concentrations of approximately 400 ng/ml, similar to the concentration required for maximum lysis by the corresponding chemically linked bispecific antibody. Thus, this recombinant bsscFv-antibody is an efficient molecule for effector-cell mediated lysis of malignant human B-lymphoid cells.
As a result of the antigen-specificity of antibodies, they present distinct advantages over other agents in the therapy of patients with malignancies. In recent years, a number of monoclonal antibodies (MoAbs) have been used successfully in the treatment of certain patients with tumours. For example, the CD20 antibody Rituximab, the Herceptin antibody directed against the proto-oncoprotein HER-2/neu, and the CD52 antibody CAMPATH-1H demonstrated significant anti-tumour activity in patients (White et al, 2001). However, therapy with conventional MoAbs also presents certain disadvantages. Tumour penetration is often limited by the size of the whole antibody (approximately 150 kDa). Moreover, interactions of the Fc domain of therapeutic antibodies with Fc receptors on non-cytotoxic cells, such as platelets or B cells, or non-activating Fc receptors (e.g. FcγRIIIb on granulocytes), may limit their cytotoxic effects (Ravetch & Kinet, 1991; van de Winkel & Anderson, 1991; Gessner et al, 1998). In addition, the interaction of the antibodies with inhibitory Fc receptor isoforms, such as FcγRIIb on monocytes/macrophages, may further diminish their therapeutic activity (Clynes et al, 2000). Finally, Fc receptor polymorphisms may critically determine the clinical response to antibody therapy, as it was demonstrated for the bi-allelic polymorphism of FcγRIIIA (Val 158 versus Phe 158) in clinical applications of Rituximab (Cartron et al, 2002; Weng & Levy, 2002).
Some limitations of conventional antibodies – commonly of human IgG1 isotype – can be overcome by the use of bispecific antibodies (bsAbs) (Fanger et al, 1989; Segal et al, 1999; van Spriel et al, 2000; van Ojik & Valerius, 2001). Bispecific antibodies have two binding specificities, one typically directed against a lytic trigger molecule on immune effector cells, and another against a tumour-associated antigen. Thereby, bsAbs offer the potential to recruit specific effector cell populations, such as natural killer (NK) cells, T cells, monocytes/macrophages or granulocytes according to the choice of the investigator, and to enhance the direct killing of tumour cells via antibody-dependent cellular cytotoxicity (ADCC) or phagocytosis. First generation bsAbs were produced by fusing two hybridomas to generate quadromas (Milstein & Cuello, 1983), or by chemically cross-linking two F(ab) fragments (Glennie et al, 1987). Both techniques confirmed the potential of lysing cancer cells via the recruitment of immune effector cells in vitro and in animal studies. However, the required production and purification procedures are expensive and do not enable further modifications to be added easily to the resulting molecule. These disadvantages may be reduced at least in part by generating recombinant bsAbs (Pluckthun & Pack, 1997; Carter, 2001). Over the past few years, different types of genetically engineered bsAb have been described, including diabodies (Holliger et al, 1993), single-chain diabodies (Alt et al, 1999), mini-antibodies (Muller et al, 1998) or bispecific single-chain Fv (bsscFv) antibodies, also called tandem scFvs (Mack et al, 1995). Compared with other formats, bsscFvs have the potential advantage of being single-chain polypeptides and are thus easier to produce in a homogeneous and defined final state.
Key requirements for target antigens in antibody therapy are high expression levels and tumour cell restriction. A number of surface antigens on malignant B cells possess these properties, but only a few of these antigens trigger potent effector cell-mediated lysis in vitro and in vivo (Tutt et al, 1998; Wurflein et al, 1998). Recent studies suggested that intracellular domains of target antigens significantly contribute to the outcome of ADCC reactions (Tiroch et al, 2002). Human leucocyte antigen (HLA) class II represents a potentionally very suitable target on the surface of malignant B-lymphoid cells, because anti-HLA class II antibodies effectively trigger ADCC, complement-dependent cytotoxicity (CDC), growth inhibition and the induction of apoptosis. Furthermore, anti-HLA class II antibodies demonstrated tumoricidal activity in xenografted models of non-Hodgkin's lymphoma (Longo, 2002). Consequently, HLA class II antibodies, including Lym-1 and Hu1D10, have been used in clinical trials and induced clinical responses, at least in individual patients (Dechant et al, 2003).
For the induction of cellular cytotoxicity, trigger molecules on the surface of immune effector cells must be activated. Tumour cell lysis typically occurs upon antibody binding to cytotoxic trigger molecules, such as CD3 on T cells, CD16 on NK cells, CD64 on activated neutrophils and monocytes/macrophages, or CD89 on neutrophils (Segal et al, 1999; van Spriel et al, 2000; van Ojik & Valerius, 2001). CD16 is the low affinity receptor for IgG (FcγRIII), which is constitutively expressed as a transmembrane isoform on NK cells and macrophages (CD16a), and as a glycosylphosphatidylinositol (GPI)-linked molecule on neutrophils (CD16b) (Ravetch & Kinet, 1991; van de Winkel & Anderson, 1991). For signal transduction, CD16a requires the association with the immunoreceptor tyrosine-based activation motif (ITAM)-containing Fc receptorγ chain. Because of its restricted expression on potential effector cells, CD16 represents an interesting target for the recruitment and induction of cytolysis. Studies with CD16-directed conventionally coupled bsAb demonstrated efficient lysis of malignant cells in vitro and in animal models (Garcia de Palazzo et al, 1992; Hombach et al, 1993; Kipriyanov et al, 2002), and clinical trials were initiated (Weiner et al, 1995; Hartmann et al, 1997). However, these trials were hampered by limited amounts of available bsAb, by the induction of human anti-mouse antibodies (HAMA), and by the presence of an Fc-portion in these hybrid-hybridoma antibodies, which may have contributed to their toxicity. To overcome these obstacles, we have generated a recombinant bsscFv-antibody, consisting only of variable regions and lacking Fc portions. This construct was produced in insect cells and effectively triggered ADCC of lymphoma cell lines and primary leukaemia samples. Thus, this bsscFv offers a basis for the improvement of CD16-directed bsAbs.
Materials and methods
Monoclonal antibodies and bispecific antibodies
The negative control hybridoma antibody 3·6·2 (mIgG2a) and the hybridoma cell line 3G8 (FcγRIII, CD16; mIgG1) (Fleit et al, 1982) were from the American Type Cell Culture Collection (ATCC, Manassas, VA, USA). The 3G8 F(ab)2 fragment was provided by Medarex (Annandale, NJ, USA). The HLA class II hybridoma F3.3 (mIgG1) (Elsasser et al, 1996) and the chemically cross-linked [3G8 × F3.3] bsAb were provided by Dr M. Glennie (Tenovus Research Laboratory, Southampton, UK). Chemical cross-linking of F(ab) fragments was performed as described (Glennie et al, 1987).
The MoAbs used for detection of recombinant proteins were Penta-His (Qiagen, Hilden, Germany), horseradish peroxidase (HRP)-coupled sheep anti-mouse IgG (Dianova, Hamburg, Germany) and phycoerythrin (PE)-coupled goat anti-mouse IgG (DAKO Diagnostica GmbH, Hamburg, Germany).
To obtain F(ab) fragments of the F3.3 antibody, 30 μg of the antibody were digested with 1 μg papain (Sigma, Taufkirchen, Germany) in a 200 μl volume containing 50 mmol/l l-cysteine, 1 mmol/l EDTA, and 20 μl sodium acetate, pH5·5. After incubation at 37°C for 16 h, the reaction was stopped by addition of iodoacetamide to a final concentration of 75 mmol/l and incubation at 30°C for 30 min. To remove non-digested antibodies and Fc fragments, 10 μg of protein A-agarose was added and incubated at 4°C for 4 h. After centrifugation, the supernatant was collected and dialysed against phosphate-buffered saline (PBS).
Culture of eukaryotic cells
Chinese hamster ovary (CHO) cells, stably transfected with a human CD16A cDNA expression construct, were provided by Dr Jan van de Winkel (University Medical Centre, Utrecht, The Netherlands). Leukaemia-derived SEM cells [t4;11 acute lymphoblastic leukaemia (ALL)] (Greil et al, 1994), Raji (Burkitt's lymphoma), ARH-77 (mature B cells) (both from ATCC), RS4;11 (t4;11 ALL), NALM-6 [common (c)ALL)] (both from the German Collection of Microorganisms and Cell Lines, DSMZ, Braunschweig, Germany), and the two hybridomas 3G8 and F3.3 were cultured in Roswell Park memorial Institute (RPMI) 1640-Glutamax-I medium (Invitrogen, Karlsruhe, Germany), containing 10% fetal calf serum (FCS), 100 U/ml penicillin and 100 μg/ml streptomycin (RF10+ -medium). Murine fibroblast L66 cells, stably transfected with a human HLA DR cDNA expression construct, were provided by Dr George L. Weiner (University of Iowa, Iowa City, USA) and maintained in RPMI 1640-Glutamax-I medium, supplemented with 10% FCS, 100 U/ml penicillin, 100 μg/ml streptomycin, and 400 μg/ml geneticin (Invitrogen) to select for the continued presence of the transgenic cDNA transgenic construct. Human 293T cells (ATCC) were cultured in Dulbecco's modified essential medium-Glutamax-I medium supplemented with 10% FCS, 100 U/ml penicillin and 100 μg/ml streptomycin. SF21 insect cells (Invitrogen) were cultured in SF-900 II medium (Invitrogen) containing 50 U/ml penicillin and 50 μg/ml streptomycin at 27°C in a humidified incubator.
Bacterial strains and plasmids
Escherichia coli XL-1-Blue (Stratagene, Amsterdam, The Netherlands) was used for the amplification of plasmids and cloning, and E. coli TG1 [from Dr G. Winter, Medical Research Council (MRC), Cambridge, UK) was used for the screening of antibody libraries. Libraries were generated in the phagemid vector pAK100, and pAK400 (both from Dr A. Plückthun, University of Zurich, Switzerland) was used for expression of soluble scFvs (Krebber et al, 1997) in E. coli HB2151 (from Dr G. Winter, MRC, Cambridge, UK). The vectors pSecTag2HygroC and pFASTBAC1 (both Invitrogen) were used for expression in eukaryotic cells.
Cloning of variable immunoglobulin domains, scFvs, bsscFvs and plasmids
The cloning of variable light (VL) and variable heavy-chain (VH) domains from the hybridomas F3.3 and 3G8 was performed following published procedures (Krebber et al, 1997; Peipp et al, 2002). To produce anti-HLA class II-red fluorescent protein (RFP) and anti-CD16-green fluorescent protein (GFP) fusions, the vector pAK400 containing the scFv coding for either anti-HLA class II or anti-CD16 was digested with SfiI. The resulting insert was cloned into the vector pSecTag2HygroC-RFP (unpublished observations) or pSecTag2HygroC-GFP (unpublished observations) linearized with SfiI. This resulted in the vectors pSecTag2HygroC-HLA class II-RFP and pSecTag2HygroC-CD16-GFP.
The secretion leader sequence from the melittin gene of the honeybee was amplified by polymerase chain reaction (PCR) with the primers 5′-BamHI, 5′-GCGGATCCATGAAGTTCTTAGTAAACGTAGCATTA G-3′; and 3′-SfiI-EcoRI,5′-GCGAATTCGGCCGGCTGGGCCGCTGCGTATATGTAGCTTATGTATACT-3′ (Bradl & Jack, 2001). The resulting PCR fragments were digested with BamHI and EcoRI, and cloned into the plasmid pFASTBAC1 linearized with BamHI and EcoRI, thereby creating the vector pFASTBAC1-Mel. Sequences for a (Gly4Ser)4 flexible linker and for the anti-CD16 scFv were cloned as an EcoRI and PvuII fragment into pFASTBAC1-Mel, linearized with EcoRI and StuI, thus generating the vector pFASTBAC1-Mel-3G8. To introduce the anti-HLA class II scFv, the vector pAK400-F3.3 was digested with SfiI. The resulting scFv insert was cloned into the vector pFASTBAC1-Mel-3G8 linearized with SfiI. This produced the expression vector pFASTBAC1-Mel-F3.3 × 3G8. Sequence analysis of the inserts was performed using dideoxynucleotide sequencing (Sambrook & Russel, 2001) on an Applied Biosystems automated DNA sequencer (ABI Prism 310 Genetic Analyzer; Perkin Elmer, Ueberlingen, Germany). The previously unpublished sequence of 3G8 antibody was determined and deposited (GENBANK accession codes AY173024 and AY173025).
To generate a chimaeric protein consisting of the extracellular domain of FcγRIII fused to the GFP, the extracellular domain of FcγRIII (CD16ex) was amplified by PCR using the primers 5′-SfiI; 5′-ATGCTAGGCCCAGCCGGCCATGCGGACTGAAGATCTCCCAAAG and 3′-SfiI; 5′-AATCGAGGCCCCCGAGGCCCTACCTTGAGTGATGGTGATGTTCAC-3′. The resulting PCR fragment was digested by SfiI and inserted to vector pSecTag2HygroC-GFP linearized with SfiI. This step completed the construction of the vector pSecTag2HygroC-CD16ex-GFP, verified by sequence analysis.
Bacterial expression and purification of soluble scFv fragments
For soluble expression of scFv fragments, plasmids were propagated in E. coli HB2151. Overnight cultures were diluted in SB medium supplemented with 1% glucose and 30 μg/ml chloramphenicol to an OD600 of ≤0·1. Cultures were then grown at 37°C to an OD600 of 0·7–0·8. Expression was induced by the addition of isopropyl β-D-thiogalactoside (IPTG) to a final concentration of 1 mmol/l, and by lowering the temperature to 30°C. After 4 h, bacteria were collected by centrifugation. Periplasmic extracts were prepared as previously described (Kipriyanov & Little, 1997), and dialysed against a 2000-fold excess of buffer containing 50 mmol/l NaH2PO4, 300 mmol/l NaCl, and 10 mmol/l imidazole, pH 8·0 at 4°C. Purification of the 6 × His-tagged soluble scFvs was performed by affinity chromatography with magnetic agarose beads, substituted with nickel-nitrilotriacetic acid (Ni-NTA) (Qiagen). Finally, the purified scFvs were dialysed against PBS.
Expression and purification of bsscFv and GFP fusion proteins
To generate the recombinant bacmid-construct, 1 μg of the expression vector pFASTBAC1-Mel-F3.3 × 3G8 was transformed into the bacterial strain DH10Bac (Invitrogen). Preparation of recombinant bacmid DNA and transfection of SF21 cells was performed following the manufacturer's guidelines (Invitrogen). After 10 d, culture supernatants containing recombinant baculovirus were collected and titrated by endpoint dilution.
To express the bsscFv, SF21 cells were infected with the recombinant baculovirus with a multiplicity of infection (MOI) of 5. After 7 d, supernatant was collected and dialysed at 4°C against a 2000-fold excess of a buffer containing 50 mmol/l NaH2PO4, 300 mmol/l NaCl, 10 mmol/l imidazole, pH 8·0. Purification of the 6 × His-tagged bsscFv was achieved by affinity chromatography with Ni-NTA agarose beads (Qiagen) and a final dialysis against PBS.
The GFP fusion proteins were expressed in 293T cells. For this purpose, 10 μg of the expression vector were transfected using the calcium phosphate procedure including 5 mmol/l chloroquine (Sambrook & Russel, 2001). After 12 h, the transfection medium was replaced by fresh culture medium. Supernatants were collected every day for 5 d, combined, dialysed and purified as described above. Concentrations of the final purified proteins were determined by colorimetric assay using a Bradford Reagent (Sigma).
Sodium dodecyl sulphate polyacrylamide gel electrophoresis and Western blot analysis
Reducing sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) was carried out according to standard procedures (Laemmli, 1970). Gels were stained with Coomassie brilliant blue R250 (Sigma). ScFvs and bsscFvs were detected with a Penta-His antibody (Qiagen). Western Blots were developed with secondary antibodies (sheep anti-mouse IgG coupled to HRP; Dianova), and developed using enhanced chemiluminescence reagents (Amersham Pharmacia Biotech, Freiburg, Germany).
Isolation of mononuclear cells, neutrophil effector cells (polymorphonuclear cells), and CLL tumour cells
After receiving informed consent, 20 ml of peripheral blood was obtained from healthy volunteers, and both mononuclear cells (MNCs) and neutrophil effector cells were isolated as described (Elsasser et al, 1996). The purity of polymorphonuclear cells (PMNs) and MNCs was assessed by cytospin preparations and exceeded 95%. Viability of cells was >95%, as tested by trypan blue exclusion. Tumour cells from patients with CD5/CD19 positive CLL were isolated from citrate-anti-coagulated peripheral blood by centrifugation over Ficoll.
A total of 5 × 105 cells were washed with PBS containing 0·1% bovine serum albumin (BSA) and 7 mmol/l Na-azide (PBA). Cells were then incubated with 50 μl of purified scFv fragments or bsscFv (1 μg/ml) for 30 min on ice. Purified non-relevant scFv fragments were used as controls. Cells were washed with PBA and incubated with 20 μl of Penta-His antibody (Qiagen) at 200 ng/ml for 30 min on ice. Unbound antibodies were then removed by washing with PBA, and 50 μl of 1 μg/ml PE-conjugated goat anti-mouse IgG (DAKO) were added to the cells. After incubation for 30 min on ice, cells were washed with PBA and fixed in 1% paraformaldehyde in PBS. For simultaneous binding and competition experiments, 50 ng bsscFv was added to 5 × 105 washed cells and incubated for 30 min on ice. After washing with PBA, 50 μl of CD16ex-GFP at 1 μg/ml was added to the cells and incubated for 30 min on ice. For competition experiments, washed cells were further incubated with a 100-fold molar excess of the parental antibodies 3G8 or F3.3 for 30 min at room temperature. Isolated CLL cells were incubated for 30 min with MoAb at 4°C. Cells were washed twice in PBS supplemented with 1% BSA, and fluorescein isothiocyanate-labelled F(ab) fragments of goat anti-mouse IgG were added as secondary reagent for 30 min at 4°C. After a final wash in PBA, cells were resuspended in 250 μl PBA and analysed by flow cytometry. Flow cytometry analysis was performed on a FACSCalibur instrument using CellQuest software (Becton Dickinson, Mansfield, MA, USA). For each sample, 1 × 104 events were collected and analyses of whole cells were performed using appropriate scatter gates to exclude cellular debris and aggregates.
Determination of affinity constants for antibodies and scFv components by flow cytometry
Determination of the affinity constants (KD) of the F3.3 antibody and the F3.3-scFv component of the bsscFv was performed by flow cytometry using published procedures (Benedict et al, 1997). Experiments for the determination of antibody KD values were repeated three to four times and mean values are reported. Values and graphical analyses were generated using GraphPad Prism Software (GraphPad Software Inc., San Diego, CA, USA).
Surface plasmon resonance
Soluble extracellular domain of human FcγRIII (sFcγRIII) – a gift from Dr Peter Sondermann (Max Planck Institute, Martinsried, Germany) – was immobilized on a CM5 sensor chip by standard amine coupling following the manufacturer's instructions (Biacore AB, Freiburg, Germany). The amount of deposited sFcγRIII was monitored by surface plasmon resonance until it corresponded to approximately 1000 resonance units. For binding analyses, purified bsscFv or F(ab)2 fragments of 3G8 were allowed to interact in the fluid phase with immobilized sFcγRIII at room temperature for 2 min at a flow rate of 10 μl/min. Varying concentrations of the bsscFv or the F(ab)2 fragment were analysed using the same chip after regeneration for 2 min with 2·5 mol/l NaCl, pH 3 at a flow rate of 10 μl/min. Binding constants were calculated using the BIAevaluation 3.0 software (Biacore).
ADCC and CDC assays
The ADCC assays were performed as described (Elsasser et al, 1996). For analysis of the effects induced by Fc-receptor blockade, antibody 3G8 F(ab)2 (FcγRIII,CD16) was added at 10 μg/ml. Relative inhibition was calculated as follows: %inhibition = (%lysis without blocking antibody −%lysis with blocking antibody)/(%lysis without blocking antibody) × 100.
Group data are reported as mean ± SEM. Differences between groups were analysed by unpaired (or, when appropriate, paired) Student's t-test. Levels of significance are indicated. ADCC data were statistically analysed for normal distribution by Kolmogorov–Smirnov and Shapiro–Wilk tests.
Experiments reported here were approved by the Ethics Committee of the University of Erlangen-Nuremberg, in accordance with the Declaration of Helsinki.
Construction of the recombinant (HLA class II × CD16) bsscFv
To target NK cells against malignant B-lymphoid cells, a recombinant bsscFv was constructed. Two scFv fragments, one directed against HLA class II on B cells, and the other specific for human CD16, were subcloned from the corresponding hybridomas F3.3 and 3G8, using phage display technology. The resulting recombinant scFvs retained specific binding in immunofluorescence experiments for both HLA-DR and FcγRIIIA on HLA-DR transfected L66 cells and CD16A-transfected CHO cells, respectively (Fig 1A and F). No binding to the corresponding non-transfected cells was observed (Fig 1B and G). The specificity of binding was further assessed by competition experiments. For this purpose, the scFvs were fused to GFP or RFP. The scFv-GFP or scFv-RFP proteins were allowed to bind first to antigen-positive cells (Fig 1C and H). Subsequent addition of the corresponding parental antibody in excess concentrations reduced the fluorescence signal to background values (Fig 1D and I), while addition of excess amounts of a non-specific antibody produced no decrease in fluorescence (Fig 1E and J). Thus, the recombinant scFvs retained the antigen specificities of the parental antibodies. These scFv-fragments were then used to generate the recombinant bsscFv (HLA class II × CD16). For this purpose, the respective scFvs were connected through a 20 amino acid flexible linker consisting of four tandem repeats of a (Gly4Ser) building block. For effective secretion from insect cells, a melittin leader peptide was fused to the N-terminus, and c-myc and hexa-histidine tags were fused to the C-terminus to facilitate purification and detection (Fig 2).
Expression and purification of the recombinant bsscFv
Recombinant bsscFvs were expressed in SF21 insect cells using baculoviral vectors. The secreted bsscFvs were collected from the culture supernatants and purified by affinity chromatography with Ni-NTA agarose beads. The apparent molecular mass, as determined by SDS-PAGE (Mr = 56–60 kDa), was in close agreement with the molecular mass calculated from its sequence (Mr = 57,996 kDa; Fig 3A). The purified protein was intact, and no antigenically reactive degradation products were observed. The purity of the enriched protein was estimated from Coomassie blue stained gels (Fig 3B). Yields ranged from 400–500 μg of purified bsscFv per litre of culture medium.
The recombinant bsscFv retained the ability to bind to each of the two antigens HLA-DR and CD16, as evidenced by its ability to specifically bind to cells transfected with the respective antigens (Fig 4A). However, these results did not exclude a mixture of partially misfolded bsAb (binding either to HLA class II or CD16) and, therefore, additional experiments were performed to demonstrate the simultaneous binding of both bsscFv components to their antigens. HLA class II-positive SEM cells were incubated with bsscFv and stained with a recombinant fusion protein, consisting of the extracellular domain of CD16, and GFP (CD16ex-GFP). In flow cytometry experiments, a fluorescent GFP signal on SEM cells was obtained with CD16ex-GFP, but not with a control GFP-fusion protein (Fig 4B,a). Thus, both scFv-binding sites of one single bsscFv-molecule were capable of binding simultaneously to their respective antigens. This conclusion was further supported by competition experiments. When each of the parental antibodies, directed either against HLA class II (Fig 4B,b) or CD16 (Fig 4B,c), was separately added in excess, the fluorescence signal decreased to negative control levels. This did not occur with a non-relevant antibody added in excess (Fig 4B, d). Therefore, both binding sites of the bsscFv were antigen-specific.
Both scFvs retained binding affinity after incorporation into the recombinant bsscFv
To assess whether incorporation of the scFvs into the recombinant bsscFv resulted in a change of their affinities, binding of the scFv components was measured and compared with the affinity of the parental MoAbs. For measurement of the affinity against FcγRIII, the recombinant soluble extracellular domain of CD16 (CD16ex) was coupled to a Biacore chip, and different concentrations of the bsscFv or the F(ab)2 fragment of the parental 3G8 antibody were used for surface plasmon resonance measurements (Fig 5A, B). To reduce background binding of the 3G8 antibody to FcγRIII, F(ab)2 fragments of this antibody were used. The resulting affinity constants were: KD = 8·61 × 10−8 mol/l for the anti-CD16 scFv-binding site in the bsscFv, and 5·73 × 10−9 mol/l for one binding site of the F(ab)2-fragment of the 3G8 antibody, calculated with a bivalent-binding model (Table I). Thus, the parental antibody had a high affinity in the nanomolar range, which was reduced 6·6-fold in the recombinant bsscFv. As the binding site of F3.3 on the HLA class II heterodimer is not defined, affinity measurement by Biacore appeared impractical, and an alternative immunofluorescence-based method was employed (Benedict et al, 1997). For this purpose, different concentrations of F3.3 F(ab) fragments or the bsscFv were incubated with HLA class II-positive SEM cells. KD was determined as the antibody concentration, at which half-maximal binding (half-maximal fluorescence intensity associated with the cells) was reached. Thus, the KD value for a F(ab) fragment of the F3.3 antibody was 2·01 ± 0·20 × 10−8 mol/l, and 13·7 ± 3·4 × 10−8 mol/l for the scFv in the bsscFv (Fig 6). Again, the affinity of the parental antibody was high, and was reduced 6·8-fold in the bsscFv. Thus, both scFvs lost affinity compared with their parental antibodies, but binding was in the range of that published for other scFvs in recombinant bsAbs (Kipriyanov et al, 1998; McCall et al, 1999). Therefore, it was promising to evaluate the functional activities of this bsscFv in ADCC reactions.
Table I. Measurement of association and dissociation rate constants (ka, kd) and affinities (KD = ka/kd) for the recombinant bsscFv (HLA class II × CD16) and F(ab)2 fragment of the parental 3G8 antibody by surface plasmon resonance.
KD values were calculated either by the *1:1 Langmuir equation or †bivalent model.
CD16 scFv in bsscFv
1·22 × 105
1·05 × 10−2
8·61 × 10−8*
2·39 × 105
1·37 × 10−3
5·73 × 10−9†
The recombinant bsscFv mediates effector cell lysis (ADCC) of malignant B-lymphoid cells
The recombinant bsscFv and the corresponding chemically linked (3G8 × F3.3) bsAb were compared for their ability to induce effector cell-mediated lysis of four different HLA class II-positive malignant human B-lymphoid cell lines. In 51Cr-release assays with isolated MNCs, specific lysis of all four lines was observed with optimum bsscFv concentrations in the range of 1–10 nmol/l. However, both agents were less effective in the killing of the immature B-cell lines Nalm-6 and RS4;11 compared with the more mature B-lymphoid cells Raji and ARH-77 (Fig 7). The recombinant bsscFv and the chemically linked bsAb displayed similar activity against three of the lines (Nalm-6, Raji, ARH-77). The significantly lower activity of the bsscFv against RS4;11 may be related to the reduced affinity of the bsscFv compared with conventional F(ab) fragments (see Fig 6), which becomes relevant at lower HLA class II density, e.g. on RS4;11. To exclude general activation of effector cells by binding of bsAbs to CD16, blocking experiments were performed. In the presence of bsscFv (HLA class II × CD16), MNC-mediated ADCC of ARH-77 cells was significantly (P < 0·05) inhibited by adding F(ab)2 fragments of FcγRIII antibody 3G8 (93·1% ± 6.0%, n = 3, data not shown). Furthermore no MNC-mediated lysis was observed, when a control bsscFv that also binds to CD16 was used. Thus, unspecific activation of FcγRIII by bsscFvs was not found.
To identify the effector mechanism responsible for this lysis, peripheral blood from healthy donors was fractionated into MNCs, PMNs and plasma. These fractions were then used in 51Cr-release assays with ARH-77 cells as targets. As expected, only the MNC fraction, containing the NK cells, produced significant lysis in combination with the bsscFv, whereas PMNs were inactive (Fig 8). Importantly, no complement-mediated killing in the plasma fraction was observed. Thus, cells responsible for lysis were enriched in the MNC fraction, suggesting that CD16a-positive NK cells are the predominant effectors.
The recombinant bsscFv was able not only to induce lysis of established tumour-derived cell lines, but also of primary tumour cells from three different B-CLL patients (Fig 9). All three patient samples were positive for HLA class II expression in immunofluorescence assays, and both the recombinant and the chemically linked bsAb triggered comparable lysis of all three samples. Thus, overall, the recombinant (HLA class II × CD16) bsscFv induced lysis comparable with the chemically linked (3G8 × F3.3) bsAb, not only of established tumour-derived B-lymphoid cell lines, but also of primary HLA class II-positive CLL cells.
The main result of this study was that a recombinant bsscFv (HLA class II × CD16) triggered effective lysis of human leukaemia and lymphoma cells. Notably, the recombinant bsscFv and the corresponding chemically linked [F(ab) × F(ab)] bsAb were able to mediate effective lysis not only of malignant human B-lymphoid cell lines, but also of primary leukaemia samples. Both molecules demonstrated overall similar killing levels with maximum lysis at concentrations of 1–10 nmol/l. Only RS4;11 cells were lysed less efficiently by the recombinant bsscFv than by the chemically linked bsAb. One possible explanation may be that RS4;11 are the only cells in this panel that were derived from a leukaemia with a characteristic t(4;11) translocation. Conceivably, this translocation may render the cells particularly resistant to lysis (Kersey et al, 1998; Pui et al, 2002). In this case, the approximately sevenfold lower affinity of both scFvs in the bispecific construct compared with the parental F(ab) fragments may become relevant. This loss of affinity had no apparent influence in the killing of the other target cell lines and primary patient samples. This is a somewhat unexpected result, because the literature suggested that the affinity for the target antigen is the dominant parameter that determines ADCC by a bsscFv (McCall et al, 2001). So far, the influence of the affinity for the trigger component of bsAbs has only been investigated for a bispecific antibody targeting CD3, but not CD16. In this case, a bsAb with weaker binding to the CD3 trigger molecule produced more effective T-cell activation and cytotoxicity than a bsAb with stronger CD3 binding (Bortoletto et al, 2002). The data presented above indicate that while the affinities may be important, they may vary over a significant range without a major effect on the efficacy of the bsscFv. Therefore, for this particular pair of antigens, other parameters may be dominant in determining the efficacy of the bsscFv, e.g. the antigens against which bsscFvs are directed, the density of the epitopes on effector and target cells, and the selection of trigger molecules. For example, we have encountered several cases in which bsAbs did not induce efficient ADCC, although both binding sites were capable of binding their cognate antigens (Stadick et al, 2002; Stockmeyer et al, 2002). Apparently, the success of bsAbs strongly depends on the particular pair of target and trigger antigens.
CD16 was reconfirmed by the present study as a suitable trigger molecule. It was known from a substantial body of published work by other authors that CD16a is a potent trigger molecule on NK cells (Gessner et al, 1998). Furthermore, CD16 antibodies, such as 3G8 used here, were demonstrated to bind to FcγRIII outside of the IgG binding site, to activate NK cells and to initiate cytolytic responses. Equally well known were certain drawbacks that may argue against CD16 as an effector molecule for bsAb. These notably include the fact that most CD16 antibodies cannot discriminate between the CD16a transmembrane isoform present on NK cells and macrophages, and the GPI-linked CD16b variant on neutrophils. Only the former is capable of triggering cytolytic responses. Furthermore, soluble CD16 is present in considerable amounts in human plasma (Koene et al, 1996). Therefore, the cytolytic activity of bsAbs directed against CD16 may be reduced in vivo by virtue of their interaction with CD16b on neutrophils and with soluble CD16 in human plasma, as long as the bsAb is not present in large molar excess. In spite of these conceivable limitations, results of other authors demonstrated that bsAbs that recruited NK cells via CD16 had biological activity in vitro and, importantly, also in tumour patients (Weiner et al, 1995; Hartmann et al, 1997).
It was also known that HLA class II represents a suitable target antigen on malignant B-lymphoid cells to mediate ADCC (Vaickus et al, 1990; Ottonello et al, 1999), and that chemically linked bsAbs directed against HLA class II and Fc receptors were potent inducers of specific cell lysis in cell culture assays. Interestingly, HLA class II proved significantly more effective than other B-cell-related antigens in triggering ADCC (Elsasser et al, 1996; Wurflein et al, 1998). However, it is also known that HLA class II is expressed on a broad variety of cells, including normal B cells, dendritic cells, monocytes, macrophages, myeloid and erythroid precursors and some epithelial and endothelial cells. Therefore normal HLA class II expressing cells are probably also affected when the bsscFv (HLA class II × CD16) is used in vivo. Therefore, it was reasonable to anticipate that in spite of the potency of this antigen in in vitro assays, HLA class II may be of limited use in vivo because of its broad expression profile. Importantly, variant forms of the HLA-DR antigen have been identified (1D10, Lym-1 and Lym-2), which have a narrower expression pattern and, thus, are more promising targets for immunotherapeutic agents. Primate experiments with one of these antibodies, a humanized IgG1 antibody directed against a polymorphic epitope on the HLA-DR β-chain (Hu1D10, Remitogen: Protein Design Labs, Fremont, CA, USA), demonstrated side effects, which were reduced by slower infusion rates (Klingbeil & Hsu, 1999). Consequently, Hu1D10 is currently under evaluation in phase I and II clinical trials, and encouraging results in individual lymphoma patients have been reported (Link et al, 2001).
In addition to ADCC and CDC, HLA-DR antibodies can effectively trigger apoptosis in B-lymphoma cells (Truman et al, 1994; Nagy et al, 2002), which is not dependent on the Fc region of the antibody. A HLA-DR antibody of human IgG4 isotype has recently been reported to produce effective tumour cell lysis in vitro and in mice xenografted with human non-Hodgkin lymphoma cells. In primates, this antibody was well tolerated and induced no relevant haematological toxicity (Nagy et al, 2002). Several lines of evidence suggest that systemic complement activation contributes to the toxicity of antibody therapy (van der Kolk et al, 2001), but appears dispensable for antibody efficacy (Weng & Levy, 2001). Thus, it may be anticipated that HLA class II-directed antibody constructs, which effectively trigger ADCC or apoptosis, but not complement activation, may be of therapeutic value and may be well tolerated by patients. Therefore, in spite of reasonable reservations against HLA class II as target antigen, selected HLA class II-directed antibody constructs may prove useful for clinical applications. Antibody constructs fulfilling these criteria are complete antibodies of selected isotypes – such as human IgG4 or human IgA (Dechant et al, 2002) – or bsAbs. HLA class II-directed bsAbs, however, should not contain Fcγ regions, because CD16-directed hybrid-hybridoma bsAbs against other target antigens have already demonstrated toxicity in clinical trials, which was considered to be mediated, at least in part, by the Fc region of the bsAb (Weiner et al, 1995; Hartmann et al, 1997). The construct presented here binds monovalently to both antigens and thereby avoids non-specific immune activation. It also does not trigger complement-mediated lysis.
Finally, insect cells provide a suitable expression system for recombinant bsscFvs. So far, only one other published report regarding the production of bsscFvs in insect cells has come to our attention (Yoshida et al, 2003). The yields reported above were acceptable in our hands (500 μg of purified protein per litre of culture supernatant), and the recombinant bsscFv (HLA class II × CD16) was functionally active in ADCC assays. It produced comparable lysis of human B-lymphoid cells as the corresponding chemically linked counterpart, although the bsscFv had reduced affinities. The recombinant protein can be functionally produced in insect cells, while chemically linked bsAb have to be constructed from two antibodies.
In summary, we observed that the recombinant bsscFv was a potent molecule for the recruitment of effector cells to malignant B-lymphoid cells and for the induction of target-cell lysis. In its present form, this molecule may perhaps not be ideal for clinical applications, because of the broad distribution of the HLA class II antigen. However, this molecule established the principle that a bsscFv against HLA class II and CD16 is functional, and therefore provides the basis for further refinements of this concept. If, for example, the more restricted HLA class II variants Lym-1 or Lym-2 were substituted as target antigens, then indeed a clinically useful molecule may result.
We thank Dr A. Plückthun for the scFv vectors; Dr P. Sondermann for the extracellular domain of human CD16; Dr G. Winter for E. coli TG1; Dr M. Glennie for providing the F3.3 hybridoma and chemically crosslinked bsAbs; Dr J. G. van de Winkel and Dr G. Weiner for CD16-transfected CHO, and HLA-DR transfected L66 cells, respectively. Th. Lange is gratefully acknowledged for administrative assistance. This research was supported by research grant DFG RE 1276/2-1 to R. Repp, T. Valerius and G.H. Fey from the Deutsche Forschungsgemeinschaft (DFG) and by research grant No. 2001.048.1 from the Wilhelm Sander Foundation, Neustadt, Germany to SJZ.