Abstract Background: CAR T-cells targeting solid tumor antigens have not yet demonstrated similar success as seen for CD19, BCMA and CD22 in B-cell malignancies. Truly tumor-specific antigens are rare, yet pediatric solid tumors manifest stalled fetal developmental programs and often overexpress oncofetal antigens, normally restricted to prenatal tissues. Antigen density is a major factor determining CAR potency and whether such nonmutant antigens can be effectively and safely targeted remains unknown. Here, we engineered CAR T-cells targeting GPC2 to recognize clinically relevant antigen thresholds and assess safety. Methods: Expression of GPC2 was assessed during organ development (Cardoso-Moreira et al. 2019), by IHC in prenatal, infant and pediatric brain and in ssRNAseq datasets of fetal (La Manno et. al 2016) and adult brain (Allen Brain Atlas). To measure the molecules/cell of GPC2 and other immunotherapy-relevant targets on bone-marrow infiltrating neuroblastoma (NB) cells of high-risk patients, we developed a Flow Cytometric quantification assay and engineered GPC2-CAR T-cells to recognize relevant antigen thresholds. To assess binding specificity towards GPC2, a membrane proteome array encompassing >5,300 human proteins was utilized. Taking advantage of the murine cross-reactivity of our lead GPC2-CAR, toxicity was assessed in a relevant xenograft model. Results: During organ development, we found GPC2 expression restricted to the developing brain and a gradual decrease until silenced after birth. IHC of prenatal brain revealed an inverse correlation between GPC2 and gestational age (r = -0.775, p = 0.008) and very low staining in infant and pediatric brain (H-scores <100). Analysis of fetal midbrain ssRNAseq data spatializes GPC2 to the neuronal progenitor compartment, a correlation absent in adult brain. We measured a mean of ~4782 GPC2 molecules/cell on patient's metastatic NB cells, a significantly lower number than on NB cell lines (p=0.0496). We found KYMRIAH-based CAR T-cell constructs unable to control tumor due to subthreshold antigen density. Sequential modifications to the CAR architecture, such as CD28-derived hinge-transmembrane and signaling domains domains resulted in a potent lead GPC2-CAR capable of impressive tumor regression in three xenograft models of NB (metastatic, orthotopic and patient-derived xenograft). Potency of our lead GPC2-CAR could be further enhanced by overexpression of c-Jun, mediating long-term complete responses. We found no clinically relevant off-target binding to other membrane proteins and no clinical or histological evidence of toxicity. Conclusion: Our results demonstrate safety and efficacy of a lead GPC2-CAR and illuminate oncofetal antigens as a promising class of targets for CAR T cell therapy of solid tumors. This work establishes a strong basis for testing GPC2 CAR T-cells in early phase clinical trials. Citation Format: Sabine Heitzeneder, Kristopher R. Bosse, Zhongyu Zhu, Dontcho Jelev, Shaurya Dhingra, Robbie Majzner, Elena Sotillo-Pineiro, Samantha Buongervino, Peng Xu, Jing Huang, Alberto Delaidelli, Martin Hasselblatt, Kevin Parker, Hima Anbunathan, Anya Alag, Jennifer Hwang, Min Huang, Dorota D. Klysz, Johanna L. Theruvath, Jose G. Vilches-Moure, Ansuman T. Satpathy, Poul H. Sorensen, Dimiter S. Dimitrov, John M. Maris, Crystal L. Mackall. Potent activity of CAR T cells targeting the oncofetal protein GPC2 engineered to recognize low antigen density in neuroblastoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 1548.