GD2/B7-H3 bispecific antibodies for next-generation neuroblastoma treatment

376 Objectives: Targeting disialoganglioside (GD2) with monoclonal antibodies (mAb) is included as standard-of-care in high-risk pediatric-neuroblastoma. Despite high-expression in tumor cells, GD2 is also present in peripheral nerves, leading to substantial, at times dose-limiting, neuropathic-pain in many treated patients. Another molecular target B7-H3 (CD276), is also overexpressed in many cancers including neuroblastoma, but not on nerves. Our goal is to generate anti-GD2/B7-H3 bispecific antibodies (bs-mAb) that conserve high tumor avidity with significantly reduced off-target GD2-binding. Methods: Three bispecific bs-mAb, I7.33-GD2.2, I7.34-GD2.6, and I7.36-GD2.6 were engineered to present Fab fragments with moderate affinity for GD2 or B7-H3 alone, but strong binding when combined. The bs-mAbs were conjugated to deferoxamine and radiolabeled with 89Zr for in vitro and in vivo studies. The affinities of the bs-mAbs for GD2+/B7-H3+ melanoma (MEL) cells (B78-H3), GD2+/B7-H3‒ MEL (B78), or GD2‒/B7-H3+ MEL (B16-H3), were determined via an in vitro radioactive competitive binding assay and compared to that of monospecific anti-GD2 (dinutuximab) and anti-B7-H3 (I7.36) mAbs. To assess in vivo tumor avidity, C57BL/6 mice (n=3) bearing subcutaneous B78-H3 MELs were intravenously administered 89Zr-I7.33-GD2.2, 89Zr-I7.34-GD2.6,89Zr-I7.36-GD2.6, or 89Zr-dinutuxumab; longitudinal PET scans were acquired at 3, 24, 48, and 72 h post-injection. PET imaging of an isotype control and blocking studies were performed to confirm the specificity of the bs-mAbs in vivo. Ex vivo biodistribution was carried out following the last PET imaging timepoint. PET imaging quantification and biodistribution results were reported as percent injected activity per gram (%IA/g) of tissue. Results: Competitive binding assays in B78 (GD2+/B7-H3‒) or B16-H3 (GD2‒/B7-H3+) cells revealed a markedly reduced binding affinity (IC50 >333 nM) for all bs-mAbs compared to the monospecific dinutuxumab (IC50=49.8±12.4 nM) and I7.36 (IC50=13.0±4.5 nM) antibodies. Conversely, the bs-mAbs I7.33-GD2.2 (IC50=16.1±4.7 nM) and I7.34-GD2.6 (IC50=12.2±3.8 nM) demonstrated higher affinity for GD2+/B7-H3+ B78-H3 cells than dinutuximab (IC50=33.9±15.4 nM). In vivo PET imaging in mice bearing B78-H3 tumors confirmed elevated and persistent tumor uptake of 89Zr-I7.33-GD2.2, 89Zr-I7.34-GD2.6, and 89Zr-I7.36-GD2.6 which peaked at 17.0±1.9, 19.3±0.7 and 18.3±1.5 %IA/g, respectively, 48-h post-injection of the radiotracers. Co-injection of non-radioactive I7.33-GD2.2 (10 mg/kg) resulted in a 46% reduction (9.1±1.7 %IA/g; p<0.0001) in tumor uptake of 89Zr-I7.33-GD2.2, and together with a low tumor uptake of the 89Zr-isotype control (6.9±1.9 %IA/g) corroborated the in vivo specificity of the bs-mAb. Consistent with its lower in vitro affinity for B78-H3 cells, 89Zr-dinutuxumab had significantly lower uptake (12.8±3.8 %IA/g; p<0.01) in B78-H3 tumors compared to all three bs-mAb tested. Conclusions: We successfully generated bs-mAbs with high tumor avidity to GD2+/B7-H3+ cells and limited off-target binding to GD2+/B7-H3‒ cells. These data suggest that future testing may show improved treatment efficacy in treating patients with GD2+/B7-H3+ tumors (likeneuroblastoma) with reduced neuropathic toxicity.