Human derived dimerization tag enhances tumor killing potency of a T-cell engaging bispecific antibody

Abstract

Bispecific antibodies (BsAbs) have proven highly efficient T cell recruiters for cancer immunotherapy by virtue of one tumor antigen-reactive single chain variable fragment (scFv) and another that binds CD3. In order to enhance the antitumor potency of these tandem scFv BsAbs (tsc-BsAbs), we exploited the dimerization domain of the human transcription factor HNF1α to enhance the avidity of a tsc-BsAb to the tumor antigen disialoganglioside GD2 while maintaining functional monovalency to CD3 to limit potential toxicity. The dimeric tsc-BsAb showed increased avidity to GD2, enhanced T cell mediated killing of neuroblastoma and melanoma cell lines in vitro (32-37 fold), exhibited a near 4-fold improvement in serum half-life, and enhanced tumor ablation in mouse xenograft models. We propose that the use of this HNF1α-derived dimerization tag may be a novel and effective strategy to increase the potency of T-cell engaging antibodies for clinical cancer immunotherapy.

Keywords: BiTE; BiTE, bispecific T-cell engager; Bispecific antibody; BsAb, bispecific antibody; GD2; HDD, HNF1α dimerization domain; T-cell; ganglioside; immunotherapy; melanoma; neuroblastoma; tsc-BsAb, tandem single-chain bispecific antibody.

Figure 1.

Engineering bispecific antibodies for cancer immunotherapy. Bispecific antibody formats that were compared in this investigation include: (A) GD2xCD3 tandem single chain variable fragment (scFv) bispecific antibody (BsAb) (tsc-BsAb) monomer, (B) GD2xCD3-HDD dimer generated by the addition of the hepatocyte nuclear factor 1 α (HNF1α) dimerization domain (HDD) to C-terminus which forms an anti-parallel 4 helix bundle, (C) GD2xCD3-dHLX dimer using a synthetic helix-loop-helix domain (dHLX), (D) GD2xCD3-Fc dimer using a human IgG1 hinge and CH2 and CH3 domains. Other possible bispecific antibody formats considered: (E) diabody (non-covalently associated dimers in which each chain comprises 2 domains consisting of VH and VL domains from 2 different antibodies), (F) tandem diabody (TandAb), (G) quadromas (IgG like BsAbs formed by hybrid hybridomas or by engineering Fc heterodimers), and (H) Dock-and-Lock BsAbs (covalent trimer of 2 antitumor Fab fragments and one anti-CD3 scFv).

Figure 2.

Comparison of bispecific antibody GD2 tumor antigen-binding kinetics. Bispecific antibody GD2 binding kinetics by surface plasmon resonance of (A) GD2xCD3 BsAb and (B) GD2xCD3-HDD. Traces are shown at the following BsAb concentrations: 62.5, 125, 250, 500, 1000, and 2000 nM. Analyses of the association and dissociation rates are shown in Table 1. (C) Binding to tumor antigen GD2 by ELISA for 4 BsAb constructs. Data are shown as the mean ± SD. The resultant EC₅₀ of binding are shown in Table 2.

Figure 3.

Bispecific antibody binding to human T cell CD3 and cytokine release. T cells purified from human peripheral blood mononuclear cells were incubated with the indicated bispecific antibody (BsAb) and characterized for CD3 binding (cytofluorimetric analysis) and cytokine release (ELISA). (A) Binding of GD2xCD3, GD2xCD3-HDD, GD2xCD3-Fc, and GD2xCD3-dHLX BsAbs to CD3 on the surface of T cells by flow cytometry. (B) Cytokine release from T cells induced by GD2xCD3 and GD2xCD3 BsAb when compared with a huOKT3 IgG. (C) Cytokine release from T cells co-cultured with neuroblastoma IMR-32 cells in the presence of the indicated BsAb. Data points are shown as mean ± SD. Cytokine values and statistical analyses are shown in Table S1 and Table S2.

Figure 4.

Bispecific antibody stimulated T cell-mediated cancer cell cytolysis. T cells purified from human peripheral blood mononuclear cells were incubated with the indicated bispecific antibody (BsAb) in the presence of cancer cells. T cell-mediated killing of (A) M14 melanoma, (B) LAN-1 neuroblastoma, and (C) IMR-32 neuroblastoma cells lines was analyzed by chromium⁵¹-release assay. Data are shown as the mean ± SD. The EC50, maximal killing, and statistical analyses are shown in Table 4.

Figure 5.

Bispecific antibody treatment affect on tumor growth in vivo. Animal xenograft study of neuroblastoma IMR-32 and melanoma M14 tumors mixed with human donor derived peripheral blood mononuclear cells (PBMCs) subcutaneously implanted into immunocompromised double knockout (DKO; BALB-Rag2^−/−IL-2R-γc-KO) mice. Recipient mice were treated with BsAbs for 2 weeks and then monitored for tumor growth and survival. (A) Tumor size for mice implanted with IMR-32 cells. (B) Survival of mice implanted with IMR-32 cells. (C) Tumor size for mice implanted with M14 cells. (D) Survival of mice implanted with M14 cells. Data points are shown as the mean ± SE. Mice were sacrificed for humane reasons if tumor volume exceeded 2 cm³.

Figure 6.

Pharmacokinetics of the indicated bispecific antibody (BsAb) Serum pharmacokinetics of the indicated bispecific antibody (BsAb) in immunocompromised double knockout (DKO; BALB-Rag2^−/−IL-2R-γc-KO) mice injected with either GD2xCD3 or GD2xCD3-HDD BsAb. Mice were given 50 μg bolus injection of BsAb, and then tail vein blood was sampled over the course of 6 h for protein clearance, using an anti-5F11 idiotype antibody for detection and sandwich ELISA. Parameters from non-compartmental pharmacokinetic analysis are shown in Table 6.