Phase 2 basket trial of precision TIL therapy for immunotherapy-resistant advanced cancers.

2505 Background: Adoptive cell therapy (ACT) using unselected autologous tumor-infiltrating lymphocytes (TIL) has demonstrated clinical activity in metastatic cutaneous melanoma, an immunotherapy-sensitive cancer with high tumor mutational burden (TMB). However, we recently demonstrated in metastatic uveal melanoma, a prototypic low-TMB, immunotherapy-resistant cancer, that efficacy of TIL ACT requires infusion of TIL with measurable anti-tumor reactivity. Thus, to extend TIL ACT to additional immunotherapy-resistant cancers, we developed TILScore , a pan-cancer in situ transcriptomic biomarker designed to preoperatively identify metastases enriched with clinically active TIL. Here we report initial results from a pilot phase 2 TIL therapy trial prospectively evaluating TILScore as a precision biomarker for immunotherapy-resistant cancers. Methods: This single-center phase 2 trial (NCT03935893) enrolled patients with treatment-refractory, locally advanced, recurrent, or metastatic cancers into ten histology-defined cohorts. After a feasibility run-in, metastases with core biopsy TILScore > 0.248 were selected for TIL harvest and manufacturing. Patients received non-myeloablative lymphodepletion (NMA-LD) with cyclophosphamide and fludarabine, followed by infusion of TIL and high-dose interleukin-2. The primary endpoint was cohort-specific objective response rate (ORR) by RECIST v1.1 using a Bayesian basket design targeting ORR > 20% as evidence of promising activity. Results: As of 1/17/2026, 19 patients received TIL therapy. Tumor types included pancreatic adenocarcinoma (PDAC, n = 7), leiomyosarcoma (n = 3), peritoneal mesothelioma (PeM, n = 2), and single cases each of non–small cell lung cancer, cervical squamous cell carcinoma (SCC), nasopharyngeal SCC, and melanoma subtypes (cutaneous, acral, mucosal, unknown origin). All had progression after frontline therapy, with a median of 5 prior metastatic treatments; 58% had received checkpoint blockade. Liver metastases were the most common TIL source (37%), followed by lung and soft tissue (32% each). Infusion products contained a median of 5.36E10 TIL with an even mix of CD4 and CD8 cells. Adverse events were consistent with NMA-LD and interleukin-2; no grade 5 events occurred. Among 18 evaluable patients, ORR was 33% (6/18) with cohort specific responses in PDAC (50%, 3/6: 1 CR, 2 PR), PeM (100%, 2/2: 1 CR, 1 PR), and melanoma of unknown origin (100%, 1/1: 1 CR). TILScore predicted clinical response with an area under the receiving operating characteristic curve of 0.806 (P = 0.041). Conclusions: TILScore -guided selection of metastases for TIL manufacturing is feasible and enables consistent generation of clinically active TIL in patients with immunotherapy-resistant cancers. Promising clinical activity in PDAC and PeM supports evaluation of this biomarker-driven TIL ACT approach in larger, histology-specific cohorts. Clinical trial information: NCT03935893 .