- Top of page
- Materials and methods
A recombinant bispecific single-chain fragment variable antibody (bsscFv), directed against the B-cell antigen CD19 and the low affinity Fc-receptor FcγRIII (CD16), was designed for use in the treatment of patients with leukaemias and lymphomas. The Fc-portions of whole antibodies were deliberately eliminated in this construct to avoid undesired effector functions. A stabilised bsscFv, ds[CD19 × CD16], was generated, in which disulphide bonds bridging the respective variable light (VL) and variable heavy (VH) chains were introduced into both component single-chain (sc)Fvs. After production in 293T cells and chromatographic purification, ds[CD19 × CD16] specifically and simultaneously bound both antigens. The serum stability of ds[CD19 × CD16] was increased more than threefold when compared with the unstabilised counterpart, while other biological properties were not affected by these mutations. In antibody-dependent cellular cytotoxicity experiments, ds[CD19 × CD16] mediated specific lysis of both CD19-positive malignant human B-lymphoid cell lines and primary tumour cells from patients with B-cell chronic lymphocytic leukaemia or B-cell acute lymphoblastic leukaemia. Natural killer cells, mononuclear cells (MNCs) from healthy donors and, in some instances, MNCs isolated from patients after allogeneic stem cell transplantation, were used as effectors. Thus, ds[CD19 × CD16] holds promise for the treatment of CD19+ B-lineage malignancies.
Relapse of leukaemias and lymphomas, caused by minimal residual disease (MRD) cells that were not eradicated by previous chemo- and radiotherapy, still remains a major problem after transplantation of haematopoietic stem cells (Handgretinger et al, 2003). The first months after allogeneic stem cell transplantation offer an advantageous window of time for the elimination of persisting MRD cells. Early reconstituted donor-derived effector cells, such as natural killer (NK) cells can be used to redirect cellular cytotoxicity against leukaemic blasts and to increase graft-versus-leukaemia effects, without the induction of graft-versus-host disease. Indeed, in two studies with completely T cell-depleted grafts, the relapse rates were not clearly increased, despite delayed T cell regeneration. This observation may be ascribed to the rapid reconstitution of NK cells in those patients (Eyrich et al, 2001; Lang et al, 2003; Lang et al, in press). Antibody-based therapies offer advantages, particularly antigen specificity and the recruitment of immune effector cells. A chimaeric antibody targeting CD19 was recently shown to mediate specific lysis of primary acute lymphoblastic leukaemia (ALL) blasts with donor-derived effector cells obtained from paediatric leukaemia patients after transplantation of purified allogeneic stem cells (Lang et al, 2004).
However, some limitations restricting the therapeutic efficacy of conventional monoclonal antibodies are known. Penetration of the tumour is limited by the size of the whole antibody (approximately 150 kDa). Interactions of the Fc domain with Fc receptors on non-cytotoxic cells, e.g. platelets or B cells, or non-activating Fc receptors, such as FcγRIIIb on granulocytes, may also reduce their therapeutic effects (Peipp & Valerius, 2002). Interaction with inhibitory Fc receptor isoforms, such as FcγRIIb on monocytes/macrophages, may further decrease their cytotoxic activity (Clynes et al, 2000). In addition, the glycosylation pattern of the IgG1 Fc-region at amino acid Asn297 influences binding to Fc receptors and the induction of antibody-dependent cellular cytotoxicity (ADCC) (Shields et al, 2002). Finally, Fc receptor polymorphisms may critically determine the clinical response to antibody therapy. This effect was demonstrated for the bi-allelic polymorphism of FcγRIIIA (Val 158 versus Phe 158) in clinical applications of the CD20 antibody Rituximab (Cartron et al, 2002; Weng & Levy, 2003).
By contrast, bispecific antibodies (bsAbs) have the potential to overcome at least some of the limitations associated with conventional antibodies (Peipp & Valerius, 2002). bsAbs combine two antigen-binding sites, one directed against a tumour-associated antigen, the other against a trigger molecule on effector cells. Thereby, bsAbs very efficiently recruit cytotoxic effector cells, such as NK cells, T-cells, monocytes/macrophages or granulocytes to the tumour cells and mediate elimination of the tumour cells via ADCC or phagocytosis. For the induction of cellular cytotoxicity, activation of effector cells is a critical requirement, which is achieved by 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 (Peipp & Valerius, 2002). Initially, bsAbs were generated by the hybrid-hybridoma technique, but subsequently different types of genetically-engineered bispecific antibody-derivatives were designed, e.g. diabodies, mini-antibodies, single chain diabodies, and bispecific single chain fragment variable antibodies (bsscFvs) (Peipp & Valerius, 2002). Single-chain diabodies and bsscFvs have the particular advantage of being single-chain polypeptides and, therefore, are easier to produce in a homogeneous and defined final state.
One of the most interesting targets for antibody therapy on malignant human B cells is CD19 (Grossbard et al, 1992). This surface antigen is expressed on nearly all developmental stages of the B cell lineage. More importantly, CD19 is not shed from the cell surface, lost from tumour cells, or expressed on haematopoietic stem cells, T cells, or other non-lymphoid cells. Thus, CD19 is a particularly attractive target antigen for antibody therapy. So far, CD19 antibodies have been investigated in various formats for therapeutic studies in cell culture and in vivo (Hekman et al, 1991; Pietersz et al, 1995; Lang et al, 2004). The first CD19-directed conventional bsAbs used CD3 as the trigger molecule for the recruitment of T cells as effectors (Haagen et al, 1992; Weiner & De Gast, 1995; Kipriyanov et al, 1998; Loffler et al, 2000). Although cell culture data demonstrated significant lytic activity for these [CD19 × CD3] bsAbs, the bsAb-mediated cross-linking of CD3 led to non-specific T-cell activation, causing bsAb-associated toxicity in vivo (Segal et al, 1999). So far, a [CD19 × CD16]-directed recombinant diabody has been reported to trigger NK cell-mediated tumour cell lysis in vitro and in a mouse model (Kipriyanov et al, 2002).
CD16 is the low affinity receptor for IgG (FcγRIII), which is constitutively expressed as a transmembrane isoform on the surface of NK cells and macrophages (CD16a), and as a glycosyl phosphatidyl inositol (GPI)-anchored molecule on the surface of neutrophils (CD16b) (Ravetch & Kinet, 1991; van de Winkel & Anderson, 1991). For intracellular signalling, CD16a requires association with the FcRγ chain containing the immunoreceptor tyrosine-based activation motif (ITAM), which triggers downstream signalling. Studies with conventionally coupled bsAbs redirecting NK cells via CD16 demonstrated potent cytolysis of cultured malignant cells and in animal models (Garcia de Palazzo et al, 1992; Hombach et al, 1993; Kipriyanov et al, 2002). Therefore, clinical trials with CD16-directed bsAbs were initiated (Weiner et al, 1995; Hartmann et al, 1997). However, immunogenicity of hybrid-hybridoma antibodies, as well as undesired side effects caused by the presence of Fc-domains, and difficulties in producing sufficient amounts of clinical-grade material limited these clinical trials.
Lysis of CD19+ MRD cells by donor-derived NK cells, mediated by a chimaeric CD19 antibody, was previously reported (Lang et al, 2004). However, this chimaeric antibody is still afflicted with the limitations of conventional antibodies. To overcome some of these limitations, a recombinant bsscF was used instead of a whole antibody to eliminate the Fc-region, which was recognised as one of the causes of the problems in the past, and which is not strictly required for the present purpose. The instability of scFvs has been one of the main arguments precluding their use in therapeutic constructs to date. Therefore, a key innovation was utilised in the present study to enhance the serum stability of this recombinant bsscFv: disulphide bridges were introduced to stabilise both scFv components of our construct. The combination of these two advances promises to be useful, and in the preclinical studies reported here, ds[CD19 × CD16] triggered potent ADCC of lymphoma cell lines and primary human B-cell chronic lymphocytic leukaemia (B-CLL) and B-lineage ALL cells.
- Top of page
- Materials and methods
The main results of this report are: (a) a recombinant bispecific scFv molecule in the tandem format has an increased serum stability after disulphide-stabilisation of both of its scFv components; (b) the disulphide-stabilised bsscFv was more efficient in mediating specific lysis of primary ALL-blasts by donor-derived MNCs after transplantation than the chimaeric IgG1 antibody CD19 4G7chim.
One of the perceived shortcomings of scFv and bsscFv-proteins for clinical applications is their instability in human serum. BsscFvs usually have a serum stability of only a few hours to a few days, as demonstrated also by the non-stabilised controls in this study (Fig 3). To overcome this limitation, disulphide-stabilised Fvs have been generated. It had previously been shown that disulphide-stabilisation significantly increased the serum stability of individual scFv-molecules (Brinkmann et al, 1993). However, to our knowledge, there are no prior reports demonstrating that the simultaneous stabilisation of two different scFv components in a tandem format bsscFv increases the serum stability of the entire molecule.
In ADCC experiments with malignant cells obtained either from the bone marrow (BM) or peripheral blood (PBL) of six different adult B-CLL patients, MNCs from unrelated healthy human donors were used as effectors (Fig 6). Although no deliberate effort was made to ensure a mismatch of their MHC class I molecules, lysis was significantly enhanced by addition of the bsscFv molecule in all six cases. Furthermore, the lytic effect was consistently higher with the bsscFv molecule than with the chimaeric CD19 antibody, regardless of whether the malignant cells were taken from the bone marrow or peripheral blood (Fig 6). In four of six cases (patients 1, 2, 3 and 5) the cytotoxic effects obtained with the bsscFv were more than twice as large as those obtained with the chimaeric CD19 antibody. It is presently unknown whether this advantage will also hold true in human patients, but we anticipate certain advantages over the chimaeric CD19 antibody, including improved tissue penetration due to its smaller size.
Normally, lytic ability of NK cells is impaired by high expression levels or expression of matched MHC class I molecules on target cells. Apparently, this inhibitory effect, which is caused by killer inhibitory receptors (KIR) on the surface of NK cells, is overcome in ADCC situations. Here, the lysis-promoting effect produced by the therapeutic antibody apparently compensates the killer inhibitory effects (Lang et al, 2004). Rather large variations in the extent of lysis observed for individual pairs of tumour and effector cells (Fig 6) may be partly due to the variable extent of MHC class I mismatch between the target and effector cells. Other potential causes for this observation may include variations in the expression levels of activating NK cell receptors, such as NKG2D, DAP10 or DAP12.
The ADCC experiments with primary leukaemic ALL blasts from paediatric patients with the bsscFv in comparison with the chimaeric antibody produced an unexpected observation. When unstimulated enriched NK cells from healthy unrelated donors were used as effectors (Fig 7A), the chimaeric CD19 antibody produced somewhat higher lysis than the bsscFv. However, when unstimulated donor-derived MNCs from transplanted patients were used, then the bsscFv produced a greater extent of lysis (Fig 7B). In this situation, MNCs were used and no attempt was made to enrich NK cells, because the amount of blood cells available from transplanted paediatric patients was too limited to permit enrichment of NK cells. An explanation for this observation might be that transplanted patients regularly received standard human IgG infusions (200 mg/kg every 3 weeks) in order to prevent infectious complications. Such infusions are likely to block human Fc receptors as a side effect, and therefore may reduce free binding sites for the chimaeric antibody on the effector cells. By contrast, it is conceivable that the function of the bsscFv is not inhibited under these conditions, because the CD16-specific reading head binds FcγRIII at an epitope outside its Fc-binding site. This experimental setting using donor-derived MNCs comes closest to the in vivo situation for which our construct was primarily designed, namely the treatment of MRD cells in a post-transplantation setting. In this situation, the bsscFv format had superior properties over the chimaeric antibody, which remain to be confirmed by clinical investigations.
CD19 has long been recognised as a potentially very useful target antigen for the therapy of B-lymphoid malignancies, because of its restriction to the B cell lineage. It is expressed on most B-lineage ALLs, including infant pro-B ALLs, which usually lack CD20, and therefore appears to be particularly attractive for the treatment of CD20-negative paediatric leukaemias. Consequently, CD19-directed antibodies have been investigated for therapeutic use against human B-lymphoid malignancies, but until now, therapeutic IgG antibodies have not produced clinical benefits comparable with those of CD20 antibodies (Hekman et al, 1991). In addition, conventional bsAbs targeting CD19 were generated for the recruitment of T cells via CD3. These bsAbs were effective in vitro (Bohlen et al, 1993a,b; Haagen et al, 1994; Csoka et al, 1996) and in animal models (Demanet et al, 1992; Bohlen et al, 1997; Daniel et al, 1998), but not in first clinical trials (De Gast et al, 1995; Haagen, 1995).
More recently, recombinant bsscFvs comprising only antibody variable domains have been constructed. These molecules are expected to be less immunogenic than complete antibodies and can be produced at relatively high yields in a more defined final state. The currently most advanced recombinant protein in this format is a [CD19 × CD3] bsscFv (Loffler et al, 2000). For the particular purpose of our work, the elimination of MRD cells in paediatric ALL patients after transplantation of T cell-depleted grafts, T cells clearly are not the ideal class of effector cells, because of their delayed reconstitution (Eyrich et al, 2001). NK cells and granulocytes show much faster reconstitution, and, therefore, CD16 was the more promising choice of a trigger molecule on the available population of effector cells: the NK cells. It is also important to note that there are few MRD cells in the first few months after transplantation, and high effector-to-target cell ratios can be achieved.
CD16 has been appreciated by other groups as a potent trigger molecule on the surface of NK cells (Gessner et al, 1998). CD16 antibodies, such as 3G8 used in this study, bind FcγRIII outside of the Fc-binding pocket, activate NK cells, and induce cytotoxic responses. Some properties of CD16 may also limit its use as a trigger molecule. Among these is the inability of CD16 antibodies to discriminate between the CD16a isoform present on NK cells and macrophages, which is capable of triggering a cytolytic response, and the GPI-linked CD16b isoform present on neutrophilic granulocytes, which is unable to mediate ADCC. In addition, soluble CD16 is present in considerable concentrations in human plasma (Koene et al, 1996) and may compete with FcγRIII on the surface of NK cells. However, CD16-directed bsAbs have been successfully used in clinical trials, although none have advanced past stage II and none have yet been approved for clinical application. Interestingly, the cytotoxicity of these bsAbs was not inhibited by the presence of CD16-positive PMNs in ADCC assays, a still unexplained observation (Weiner et al, 1996). In addition, the in vivo cytotoxicity of CD16-directed bsAbs was not compromised by competition with CD16b on neutrophils. This effect was also observed in preclinical studies and phase I/II clinical trials of patients with refractory Hodgkin's disease treated with a [CD30 × CD16] bsAb (Hartmann et al, 2001). These reported properties of CD16-directed bsAbs provided the basis for our current study and it is anticipated that recombinant bsscFv constructs directed against CD16 as the trigger molecule may be therapeutically useful. Based on the results of this study, we concluded that the format of the tandem bsscFv may have distinct advantages over conventional bsAbs used to date. In fact, this particular format allows investigators to fully exploit the perceived benefits of CD19 as a target molecule, which had remained below expectations when other formats of CD19-directed antibody-derived therapeutics were used. The data presented here provide a clear impetus for further in vivo evaluation of [CD19 × CD16] bsscFvs.