Background: Adoptive cell therapy using tumor-infiltrating lymphocytes (TILs) is effective against solid tumors with convincing evidence. But T-cells infused were exhausted when they encountered PD-1-mediated immunosuppression in the microenvironment. Here, we generated a modified autologous TILs secreting antibody targeting PD-1 (PD1-TIL), and conducted a first-in-human, two arms, phase 1 study (NCT03347097) in recurrent glioblastoma who had a poor prognosis with estimated 6-9 months of median overall survival (OS). Methods: Patients were included to be treated with PD1-TIL or TIL cell products intravenously every one month following surgery. The primary endpoint was safety and feasibility. The secondary endpoints were disease control rate (DCR), OS, immunologic responses and biomarker analysis. The clinical response to cell therapy was assessed by MRI according to iRANO criteria. Here, patients with clinical benefit (CR, PR, and SD) were defined as responder. Results: A total of 21 consecutive patients (median age 45 years, range 20–64) received infusions of cell products (7 patients for PD1-TIL with 1 excluded, 14 patients for TIL with 2 excluded). 52% (11/21) and 29% (6/21) of 21 patients had methylated MGMT promoter and IDH1/2 mutant respectively. No significant differences in baseline characteristics between groups were found. Both of PD1-TIL and TIL infusions were well tolerated with no unexpected high-grade adverse events. Of 7 evaluable patients in PD1-TIL group, 1 had PR for more than 26 months, 2 had SD lasting for 7.7 months and 24 months, and 4 had PD after infusion. The DCR, 1-year and 2-years survival rates in PD1-TIL group were 43% (3/7), 86% (6/7) and 29% (2/7) respectively. In TIL group, 3 patients had SD and the others had PD. The DCR, 1-year and 2-years survival rates in TIL group were half of that in PD1-TIL group. Survival analysis showed that patients treated by PD1-TIL had extended OS than those received TIL (16.1 vs. 11.2 months, P = 0.047). Among 21 patients, clinical responders had significantly improved OS (30.9 vs. 10.7 months, P = 0.001) than non-responders. We also found a high capacity for engraftment and persistence of infused TILs in peripheral blood which could be detectable at 6 months after treatment. Existence of neoantigen-reactive T cell clonotypes in cell products may be important for responding because missing of missense mutations predicted to generate a neoantigen after cell therapy was a remarkable signature in responders. Responsive tumors were also associated with tumor microenvironment profiles. Conclusions: This first-in-human study establish the preliminary feasibility, safety and encouraging efficacy of PD1-antibody-secreting TILs, and may constitute a new treatment strategy by T-cell genetic engineering in glioblastoma. Clinical trial information: NCT03347097.