Background: HCC is a leading cause of cancer-related morbidity and mortality worldwide, of which glypican 3 is a highly specific marker. GPC3-directed CAR-T had shown promising safety but limited efficacy in the treatment of HCC. We developed an armored GPC3-directed CAR-T G3-CAR-ori2 by inserting of a novel and proprietary Ori2 element following the 4-1BB and CD3ζ domains in a second-generation CAR-T. Pre-clinical studies showed a significantly higher memory stem cell ratio and dramatically improved proliferation and persistence, compared with traditional CAR-T, thus offering prolonged efficacy in vitro and in vivo and potentiality leading to improved activity in the clinical setting (ChiCTR1900028121). Methods: This is an open-label, dose-escalation study of G3-CAR-ori2 cell HCC patients in two centers. Eligible patients were aged 18-70 years with histologically confirmed GPC3+ HCC, Child-Pugh score A or B, ECOG≤1, relapsed or refractory to standard therapies. Patients were pre-conditioned with fludarabine (25̃30 mg/m2) and cyclophosphamide (200̃300 mg/m2) daily for 3 days. G3-CAR-Ori2 was administered as a single infusion via intrahepatic or intravenous route with a total dose of 0.9 to 3x10e8 CAR-T. The objective was to assess the safety, preliminary efficacy, persistence and cytokine profiling of G3-CAR-ori2. Adverse events were graded using the Common Terminology Criteria for Adverse Events (version 5.0). Tumor responses were evaluated per RECIST (version 1.1). CAR-T cell expansion and persistence were measured by qPCR and flow cytometry. Results: As of Jan 21, 2021, 10 patients had received single infusion, in which 6 received G3-CAR-ori2 via intravenous route and 4 via intrahepatic route. 7 patients received the highest dose level of 3x10e8. 9 patients reached at least 1 month of follow-up and tolerated the treatment well with no dose-limiting toxicity. All patients experienced transient grade 4 decrease in lymphocyte count resulted from the lymphodepletion regimen. Cytokine release syndrome (CRS) was observed in 8 patients, in which 6 at grade 1 or 2 and 2 at grade 4 notably infused with 3 x10e8 both via intravenous route and reversed within 7 days by administering high-dose steroids and tocilizumab. Other grade 4 hematologic toxicities include thrombocytopenia (2/9) and neutropenia (1/9). No neurotoxicity was observed. Two subjects were not evaluable due to early withdrawal from the trial. Among the 7 evaluable subjects, the best responses achieved are 3 PR, 2 SD, 2 PD. The duration of remission of one patient with PR is more than 4 months, follow up is ongoing. CAR-T gene detected by q-PCR provide preliminary indication that G3-CAR-ori2 is able to expand and persist well in the clinical setting. Conclusions: These initial data provide evidence that G3-CAR-ori2 is safe and holds promising antitumor potential, and supports its continuing development in the treatment of r/r GPC3+HCC. Clinical trial information: ChiCTR1900028121.