Heterodimeric bispecific antibody-derivatives against CD19 and CD16 induce effective antibody-dependent cellular cytotoxicity against B-lymphoid tumor cells

Abstract

Bispecific scFv antibody-derivatives (bsscFvs) recruiting natural killer (NK) cells for the lysis of malignant cells have therapeutic potential. However, a bsscFv specific for the B-lymphoid tumor antigen CD19 and the trigger molecule CD16 on NK cells had similar affinities for both antigens (42 and 58 nM, respectively) and was not optimal for cytotoxicity. Therefore, a bispecific tribody (bsTb) was constructed with two binding sites for CD19 and one for CD16. This bsTb contained a CD19-specific Fab fragment carrying a CD16-specific scFv fused to its light chain and a CD19-specific scFv fused to its heavy chain. The bsTb was compared with a bispecific bibody (bsBb) lacking the CD19-specific scFv. The bsTb had 3-fold greater avidity for CD19 than the bsBb (8 and 24 nM, respectively), while both had equal affinity for CD16 (56 nM). Both molecules mediated antibody-dependent cellular cytotoxicity (ADCC) of leukemia-derived SEM cells and primary cells from leukemia patients. The bsTb showed half-maximum effective concentrations (EC₅₀) of 55 pM and promoted equal lysis as the bsBb and the bsscFv at 6- and 12-fold lower concentrations, respectively. Among these three molecules the bsTb showed the most promising in vitro properties which are anticipated to be displayed also in vivo.

Introduction

The cell surface antigen CD19 is a 95 kDa transmembrane glycoprotein belonging to the immunoglobulin superfamily [1], [2]. It is exclusively expressed on B-lymphoid cells in nearly all developmental stages from the early pro-B cell on, but no longer on plasma cells. The protein is abundantly expressed on the majority of B-lineage acute lymphoblastic leukemias (B-ALL) and different types of non-Hodgkin-lymphomas (NHL) [3], [4], and its presence on leukemia repopulating cells (leukemia stem cells) of certain types of B-ALL has recently been reported [5], [6], [7]. The antigen is neither shed from the surface nor lost from the tumor cells upon antibody treatment and is weakly internalized after binding of antibodies and antibody fragments [8], [9], [10]. Thus, CD19 is an attractive target for the treatment of different types of B-cell malignancies, and possibly certain chronic inflammatory diseases and autoimmune disorders with antibody-derived agents [9], [11].

Although unconjugated monoclonal antibodies (mAbs) against CD19 showed encouraging results in pre-clinical tests [11], [12], [13], they displayed only modest therapeutic effects in clinical trials [14]. Therefore, different antibody-based strategies were followed to target CD19 more efficiently, including both immunotoxins and antibodies with improved effector cell-mediated functions [15], [16]. The recruitment of immune effector cells and the induction of antibody-dependent cellular cytotoxicity (ADCC) by engagement of Fc receptors (FcRs) are key events in vitro and in vivo, as demonstrated for the CD20-specific antibody Rituximab and others [17], [18], [19], [20]. Both functions are mediated by the antibody Fc-portion, and therefore, it was anticipated that optimization of the Fc-portion of CD19 antibodies may improve their therapeutic performance. Indeed, the pre-clinical efficacy of full-length CD19 antibodies was increased by strengthening the interaction between their Fc-portion and FcRs either through glyco-engineering or through amino acid substitutions achieved by mutagenesis [10], [21].

A strong improvement of effector functions was also obtained through the development of bispecific antibodies [22]. These molecules combine one binding site for a target antigen on tumor cells with a second site for an activating trigger molecule on effector cells, such as CD3 or CD28 on T-cells and CD16 (FcγRIII) on NK cells and macrophages [22], [23], [24]. In some cases, bispecific antibodies demonstrated more favorable cytotoxic properties in vitro than the parental mAbs and offered the opportunity to selectively address a desirable effector cell population [22]. In addition, the choice of antigen-combining domains from particular mAbs specific for non Fc-binding epitopes on trigger molecules allowed to avoid competition with circulating endogenous immunoglobulin (Ig) G, a known competitor of mAbs [25].

Bispecific recombinant antibody-derivatives targeting CD19 have been designed in a variety of formats [26]. Most of them used single-chain variable fragments (scFvs) as binding domains and avoided the unspecific uptake by non-activating FcRs on various cell types due to the lack of Fc-portions. Among them are tandem bispecific scFvs (bsscFvs), diabodies and tandem diabodies addressing CD16 on NK cells or CD3 on T cells [26], [27], [28]. The most advanced bispecific molecule is a tandem bsscFv with specificities for CD19 and CD3 (Blinatumomab™), which produced promising results in a phase I clinical study with NHL patients and established clinical potency for bispecific antibody-derivatives [29], [30]. However, in some clinical settings, such as the treatment of ALL patients after stem cell transplantation, NK cells may represent a more favorable population of effector cells than T cells [31]. Therefore it is still important to establish potent strategies activating NK-cells for efficient tumor cell lysis.

Although tandem bsscFvs and some of the alternative formats mentioned above offer distinct advantages, they are associated with certain limitations that might affect their therapeutic efficacy. One critical parameter for the potency of bispecific proteins is the strength of binding to the tumor cells, which is determined by the intrinsic affinity of the individual combining site and the overall functional avidity of the molecules. In one approach, an important improvement was achieved through in vitro affinity maturation of the tumor cell-directed scFv-component, which resulted in an increased cytotoxic potential in vitro [32]. In another approach, the ADCC potential of a bispecific antibody-derivative has been enhanced by including a second binding site for the tumor antigen: a heterodimeric bispecific minibody containing two scFvs for the tumor antigen Her2/neu and one scFv for CD16 was more effective in mediating ADCC than a control protein with monovalent binding for both antigens [33].

Fab fragments, assembling a binding site from an antibody light chain (L) and the Fd-fragment of the heavy chain (H), have been used as hetero-dimerization scaffolds to engineer recombinant Fab–scFv fusion proteins [34], [35], [36]. Such molecules were constructed by genetic fusion of an scFv with a different specificity to the C-terminus of either the L- or the Fd-chain of a Fab fragment. Co-expression of the extended chain with its corresponding heterologous counterpart lead to an efficient production and efficient hetero-dimerization in eukaryotic cells mediated by the complementarity of the contact zones of the L- and Fd-components and a disulfide bridge formed between them [34], [35]. Through this assembly, the binding-site of the Fab fragment was reconstituted and bispecific proteins were formed, so-called bibodies (Fig. 1A). Both chains of Fab fragments have further been extended by fusion with scFvs to build so-called tribodies (Fig. 1A). Tribodies and bibodies have intermediate molecular masses of approximately 110 and 75 kDa, respectively, and were initially designed to achieve longer plasma retention times in vivo than the smaller bsscFv fragments [34]. BsscFvs have molecular masses of only 55–65 kDa, close to the renal exclusion limit, and molecules with masses below this limit are rapidly cleared from the blood-stream by renal filtration [37]. Finally, the use of Fab fragments as binding domains in bispecific proteins is attractive, because Fab fragments in some cases had a higher affinity and a greater stability than the corresponding scFvs derived from the same hybridoma antibody [38], [39].

Our group has previously generated a disulfide-stabilized (ds) tandem bsscFv with monovalent binding for the tumor antigen CD19 and the trigger molecule CD16, ds[19 × 16] [27]. This molecule displayed strong cytotoxicity for CD19-positive human leukemic cells in ADCC reactions with mononuclear cells (MNCs) in vitro. However, monovalent targeting of tumor cells and almost similar affinities for the target antigen and the trigger molecule were considered limiting. To overcome this limitation we have designed a CD19 × CD16 tribody with one Fab and one scFv binding site for CD19 and an scFv binding site for CD16 and, for comparison, a CD19 × CD16 bibody lacking the CD19-specific scFv (Fig. 1). The tribody had the predicted increased overall avidity for CD19 over the bibody and the bsscFv, accompanied by a corresponding gain in functional ADCC activity. These results provide strong motivation to also test the in vivo properties of this molecule in future experiments.