Using lentiviral transduction, we constructed the following cell lines to express HLA-A*11:01 or human PD-L1 from their parental cells: SW480-A1101 (KRAS G12V⁺/HLA-A*11:01⁺, human colon cancer), CFPAC-1-A1101 (KRAS G12V⁺/HLA-A*11:01⁺, human pancreatic ductal adenocarcinoma), SHP-77-A1101 (KRAS G12V⁺/HLA-A*11:01⁺, human lung carcinoma), T2-A1101 (KRAS G12V^-/HLA-A*11:01⁺, human lymphoblast), and SW480-A1101-hPD-L1. HCC-366 (KRAS G12V^-/HLA-A*11:01⁺, human lung carcinoma) and SW620 (KRAS G12V⁺/HLA-A*11:01^-, human colon adenocarcinoma) were not transduced. All cell lines (Supplementary Table 1) were grown in medium supplemented with 10% fetal bovine serum (Sigma, cat# F8687) at 37°C in a humidified 5% CO₂ incubator for less than one month, Mycoplasma negative tested with a polymerase chain reaction (PCR) kit (Beyotime, China, cat# C0301S), and authenticated with Short Tandem Repeat (STR) from supplier. Peripheral blood mononuclear cells (PBMCs) were purchased from Milestone^® Biotechnologies (China), which had been approved by the Human Subject Research Ethics Committee of the Bioland Laboratory (No. GDL-IRB2022-007).

Thirteen types of normal cells (Supplementary Table 2) were used as target cells, ten of which were engineered to express HLA-A*11:01. Cells were cultured in supplemented medium at 37°C in a humidified 5% CO₂ incubator for up to two weeks. Mycoplasma testing and STR authentication were performed, confirming all lines were mycoplasma-negative and properly identified.

On day 1, PBMCs were stained with KRAS G12V_8-16–HLA-A*11:01 tetramer and anti-CD8 antibody, and specific CD8⁺ T cells were sorted by a fluorescence-activated cell sorting (FACS) (BD Bioscience). Sorted cells were co-cultured with feeder cells in TexMACS™ Medium (Miltenyi Biotec, cat# 130-097-196) supplemented with 5% human AB serum (GeminiBio, cat# 100-512), KRAS G12V_8-16 peptide, and cytokines (IL-2 (10 IU/mL), IL-7 (10 ng/mL), IL-21 (30 ng/mL)) for 14 days. On day 15, a second sort was performed, then positive CD8⁺ T cells were diluted to single-cell level and expanded into clones with feeder-cell stimulation for 14 days. On day 29, the clones which grew well were selected to analyze the positive rate (KRAS G12V_8-16–HLA-A*11:01 tetramer) then sorted. Medium was replaced every 3–4 days. TCR genes from selected clones were identified by 5′-RACE (Clontech, cat# 634860) and sequenced (Sangon Biotech).

Mutations were introduced into the CDR1 region of the TCR α chain by overlap PCR using primers listed in Supplementary Table 3. Products were cloned into a phagemid vector, electroporated into TG1 E. coli (Lucigen, cat# 60502-1), and used to generate a TCR phage display library. Variants with enhanced affinity were isolated using the KRAS G12V_8-16–HLA-A*11:01 complex. Clones showing higher ELISA signals were sequenced, and optimized TCR binding was validated by functional cell assays.

Hybrid TCR genes containing variable regions fused to murine constant regions were cloned into a lentiviral vector. Lentiviral particles were produced by co-transfected HEK-293T cells (Cobioer) with helper vectors. On day 1, CD8⁺ and CD4⁺ T cells were isolated from PBMCs using EasySep™ selection kits (STEMCELL, cat# 17853 & 17852) and activated with Dynabeads^® Human T-Activator CD3/CD28 (Gibco, cat# 11132D). On day 2, activated T cells were transduced with lentivirus for 48 h. On day 4, transduced cells were expanded for 9 days. On day 13, TCR expression (Supplementary Table 4) was analyzed by flow cytometry (CytoFLEX S, Beckman Coulter), after which TCR-T cells were cryopreserved. Cells were maintained in ImmunoCult™-XF T Cell Expansion Medium (STEMCELL, cat# 10981) with IL-2 (100 IU/mL). Unless specified, TCR-T cells refer to TCR-expressing CD8⁺ T cells.

TCR-T cells and peptide-loaded T2-A1101 cells were co-cultured on ELISPOT plates pre-coated with anti-IFN-γ antibody (BD Biosciences, cat# 551849) for 16 h at 37°C under 5% CO₂. After cell lysis, biotinylated anti-IFN-γ antibody and streptavidin-HRP conjugate were sequentially incubated. Spots were developed with 3-amino-9-ethylcarbazole substrate (BD Biosciences, cat# 551951) and counted using an AID iSpot Reader Spectrum (AID GmbH).

Supernatants contained releasable LDH were incubated with CytoTox 96^® reagent (Promega, cat# G1780) for 30 min, followed by addition of stop solution. Absorbance was measured at 490 nm (Multiskan SkyHigh, Thermo Scientific). Cytotoxicity was calculated as per the manufacturer’s instructions.

Each amino acid in the parental peptide was substituted with alanine (alanine residues were changed to glycine; see Supplementary Table 5). TCR-T cells were co-cultured with peptide (10^-10 M)-loaded T2-A1101 cells at an effector-to-target (E:T) ratio of 5:1. Cytotoxicity measured by LDH release was used to evaluate alanine-scanning results.

Chemokines and cytokines were measured using LEGENDplex™ HU Proinflam Chemokine Panel 1 (BioLegend, cat# 741081) and LEGENDplex™ HU Th1 Panel (BioLegend, cat# 741035). Diluted supernatants and standards were incubated with capture beads for 2 h, followed by detection antibodies (1 h) and SA-PE (30 min). Fluorescence was acquired by flow cytometry, and concentrations were calculated according to the manufacturer’s protocols.

TCR-T cells pre-stained with CellTrace™ Violet (Invitrogen, cat# C34557) were co-cultured with irradiated target cells for 5 days, in the presence or absence of a 1:1 mixture of 3-hydroxyanthranilic acid (KKL Med, cat# KM16148) and kynurenine (InvivoChem, cat# V13205) (KHAA), or recombinant human TGF-β1 (rhTGF-β1; R&D Systems, cat# 7754-BH-005/CF). KHAA was dissolved in DMSO to 0.4 M stock and tested at 50–400 μM; DMSO concentration was matched in all groups. RhTGF-β1 was tested at 0.1, 1, and 10 ng/mL. ImmunoCult™-XF T Cell Expansion Medium was used throughout. Proliferation was assessed by flow-cytometric detection of diluted CellTrace™ Violet signal.

TCR-T cells, T-cell deleted PBMCs, and CellTrace™ Violet-labeled CD3⁺ T cells were placed in the upper chamber of a transwell plate (Corning, cat# 3415), with target cells in the lower chamber. After 3 days in ImmunoCult™-XF T Cell Expansion Medium, infiltration and recruitment of TCR-T cells, CD3⁺ T cells, dendritic cells, macrophages, and monocytes were analyzed by antibody staining and flow cytometry (Supplementary Table 4).

TCR-T cells were co-cultured with irradiated target cells at ratios of 1:5 or 1:10, with or without rhTGF-β1 (0.1, 1, or 10 ng/mL), for 5 days in ImmunoCult™-XF T Cell Expansion Medium. Cells were fixed, permeabilized, and stained for Foxp3 (Supplementary Table 4); Foxp3⁺ frequency was determined by flow cytometry.

All animal procedures were approved by the Experimental Animal Ethics Committee of Bioland Laboratory (No. 2022-0018). Female NOD/SCID-IL2rγ^-/^- (B-NDG) mice (Biocytogen), aged 6–9 weeks, were housed under SPF conditions with sterile food and water. Mice were subcutaneously injected on the dorsum with SW480-A1101 (2×10⁶), CFPAC-1-A1101 (1.3×10⁶), SHP-77-A1101 (3×10⁶), or SW480-A1101-hPD-L1 (2×10⁶) cells. When tumor volume reached 30–70 mm³, tumor-bearing mice (n = 20 per model) were randomized into four groups using Microsoft Excel and received a single intravenous injection of PBS, CD8⁺ T cells (3×10⁷), TCR0-T cells (3×10⁷), or TCR3-T cells (3×10⁷). Tumors were measured twice weekly with calipers; volume was calculated as (length × width²)/2 (19). Mice were anesthetized with isoflurane (REWARD, cat# R510-22) and euthanized by CO₂ inhalation when average tumor volume in a group reached 1000 mm³.

Data were analyzed and visualized using GraphPad Prism v8.0. Results are presented as mean ± SEM. Differences between groups were assessed by unpaired two-tailed Student’s t-test; P < 0.05 was considered significant. Significance levels: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns (not significant) P > 0.05. Brackets indicate compared groups. Each experiment was repeated at least three times. Independent experiments used T cells from the same healthy donor. Sample sizes are provided in figure legends.

We isolated a wild−type TCR, designated TCR0, from a healthy donor using the KRAS G12V_8-16–HLA−A11:01 complex (Supplementary Figure 1). TCR0 utilizes the variable−region genes TRAV12−2 and TRBV7−9. Hybrid genes combining the TCR0 variable region with the murine constant region were transduced into polyclonal CD8⁺ T cells, resulting in TCR0 expression on >70% of cells in the resulting TCR0−CD8⁺ T−cell population (Figure 1A). The majority of TCR0−CD8⁺ T cells bound the KRAS G12V_8-16–HLA−A11:01 tetramer but not the wild−type KRAS_8-16–HLA−A*11:01 tetramer (Figure 1A). Upon recognition of T2−A1101 cells pulsed with KRAS G12V_8-16 peptide (10^-5 M), TCR0−CD8⁺ T cells produced significant IFN−γ and lysed target cells, whereas no response was observed against cells pulsed with the wild−type KRAS_8-16 peptide (10^-5 M) (Figure 1D). These data confirm that TCR0 specifically recognizes the KRAS G12V_8-16 peptide. However, when the peptide concentration was reduced to 10^-8 M, TCR0−CD8⁺ T−cell activity became undetectable (Figure 1D). The functional avidity of TCR0 on TCR−T cells, calculated using a previously reported formula, was 2.168 × 10^-7 M (Figure 1D) (20, 21). Following bio−panning of a TCR0−CDR1α mutant phage library (Supplementary Figure 2A), a variant termed TCR3 was isolated (Supplementary Figures 2B, C). TCR3 carries five amino−acid substitutions in the CDR1α region (data not shown) and showed higher expression in both CD8⁺ and CD4⁺ T cells compared with the parental TCR0 (Figures 1B, C). TCR3−CD8⁺ T cells exhibited strong and specific lysis of T2−A1101 cells pulsed with KRAS G12V_8-16 peptide even at a concentration as low as 10^-¹¹ M (Figure 1E). The functional avidity of TCR3 was determined to be 9.211 × 10^-¹² M, representing a 23,537−fold improvement over TCR0. In contrast, TCR0−CD8⁺ T cells showed no detectable cytotoxicity under the same conditions (Figure 1E). Furthermore, both the mean fluorescence intensity and the percentage of TCR3−T cells binding to the KRAS G12V_8-16–HLA−A*11:01 tetramer were dose−dependent and significantly higher than those of TCR0−T cells across all tetramer concentrations tested (Supplementary Figure 3). The EC₅₀ values decreased by 12.7−fold and 203−fold, respectively (Figure 1F). Collectively, these results demonstrate that the engineered TCR3 retains original specificity while exhibiting substantially enhanced functional and binding avidity compared with the parental TCR0.

To assess the tumor-killing capacity of high-avidity TCR3-T cells, we compared TCR3 and TCR0 activity against KRAS G12V⁺/HLA-A*11:01⁺ tumor cells (SHP-77-A1101, CFPAC-1-A1101, and SW480-A1101), as well as control cells lacking either the KRAS mutation (HCC-366) or HLA-A*11:01 (SW620). TCR3-CD8⁺ T cells exhibited superior cytotoxicity. Compared to TCR0-CD8⁺ T cells, TCR3-CD8⁺ T cells demonstrated significantly elevated target cell lysis (LDH release, Figure 2A); apoptosis induction (active caspase-3; Figure 2B; Supplementary Figure 4); T cell activation (CD137 expression; Figure 2C; Supplementary Figure 5); IFN-γ secretion (Figure 2D; Supplementary Figure 6). Notably, both TCR3- and TCR0-CD8⁺ T cells showed minimal activity against control cells (HCC-366 and SW620), confirming antigen specificity (Figures 2A–D). Untransduced CD8⁺ T cells exhibited negligible cytotoxicity or activation. TCR3 enabled CD4⁺ T cell-mediated killing. TCR3-CD4⁺ T cells effectively lysed SW480-A1101 cells (Figure 2E) and expressed CD137 (Figure 2F, Supplementary Figure 7), while showing no response to antigen-negative controls. In contrast, TCR0-CD4⁺ T cells lacked cytotoxic or activation capacity (Figures 2E, F). These findings demonstrated that the higher avidity of TCR3-T cells conferred potent, antigen-restricted tumor targeting in vitro, with functional advantages over TCR0 in both CD8⁺ and CD4⁺ T cell subsets.

To evaluate the antitumor efficacy of TCR3-T cells in vivo, we established subcutaneous xenograft models in NOD/SCID-IL2rγ^−/− (B-NDG) mice using three A1101-positive tumor lines: intestinal cancer (SW480-A1101), lung cancer (SHP-77-A1101), and pancreatic cancer (CFPAC-1-A1101). Upon tumor formation, mice received a single intravenous infusion of TCR3-T cells, TCR0-T cells, or expanded T cells, with tumor growth monitored biweekly (Figure 3A). Without treatment, three control xenograft tumors continuously grew and reached endpoint (Figures 3B–G). Compared with the group of control, expanded T cells showed variable inhibition across models. TCR0-T cells outperformed expanded T cells in most models (e.g., 44.50% volume inhibition for SHP-77-A1101) but exhibited inconsistent efficacy (e.g., −35.03% volume inhibition for CFPAC-1-A1101). In contrast, TCR3-T cells mediated sustained tumor suppression, achieving complete regression without rhIL-2 support in all models (Figures 3B–G). Their efficacy significantly surpassed both TCR0-T and expanded T cells (P < 0.05), demonstrating that high-avidity TCR3-T cells enabled durable tumor elimination in vivo.

To evaluate whether TCR3-T cells could counteract PD-L1-mediated immunosuppression, we engineered SW480-A1101-hPD-L1 cells, which exhibited high PD-L1 expression compared to parental SW480-A1101 cells (Figure 4A). TCR3-T cells retained cytotoxicity despite high PD-L1. TCR0-T cells showed significantly reduced killing capacity against SW480-A1101-hPD-L1, while TCR3-T cells retained robust cytotoxicity with only marginal inhibition (Figures 4B, C; Supplementary Figures 8, 9). Programmed cell death-1 (PD-1) was modulated by TCR3-T cells. Upon recognizing SW480-A1101, TCR3-T cells upregulated PD-1, but this was attenuated against SW480-A1101-hPD-L1, suggesting PD-L1-induced feedback inhibition (Figure 4D, Supplementary Figure 10). In contrast, TCR0-T cells exhibited minimal PD-1 induction, further highlighting TCR3’s superior activation strength. Despite high PD-L1 expression, TCR3-T cells achieved complete elimination of SW480-A1101-hPD-L1 xenografts, whereas TCR0-T cells failed to control tumor growth (Figures 4E, F). These findings demonstrated that high-avidity TCR3-T cells not only enhanced tumor targeting but also counteracted PD-L1-mediated immunosuppression, supporting its potential as a robust immunotherapeutic strategy.

To assess the impact of indoleamine 2,3-dioxygenase (IDO)-derived metabolites on T cell proliferation, we treated TCR0- and TCR3-T cells with mixture (KHAA) of kynurenine and 3-hydroxyanthranilic acid (3-HAA) during co-culture with target cells. KHAA minimally affected TCR3-T cell proliferation. Coculturing with antigen-pulsed T2-A1101 cells (high peptide load (22)), TCR0-T cell expansion was strongly inhibited by KHAA (Figure 5A, Supplementary Figure 11A); TCR3-T cells maintained robust proliferation, with no significant difference even at 400 μM KHAA (Figure 5B, Supplementary Figure 11B). Co-culturing with target cells at physiological antigen levels (SW480-A1101 cells), TCR0-T cells failed to proliferate (Figure 5C, Supplementary Figure 12A); TCR3-T cells showed high expansion, with significant suppression only at ≥200 μM KHAA, but the max inhibition rate was only 15.40% (Figure 5D, Supplementary Figure 12B). Notably, 12.5 μM KHAA suffices to inhibit CD19-CAR T cells (23). These results demonstrated superior tolerance compared to CAR-T cells. We concluded that TCR3-T cells showed striking resilience to IDO-mediated immunosuppression, outperforming both TCR0-T cells and the CAR-T cells. This was a critical advantage for adoptive cell therapies in IDO-rich TMEs.

TGF−β1 represents a major barrier to T−cell therapies by promoting regulatory T−cell (Treg) differentiation and suppressing proliferation (24, 25). In this study, we demonstrate that TCR−T cells engineered for higher antigen avidity (TCR3) resisted TGF−β1−mediated immunosuppression and maintained effector function under tumor−like conditions. To assess TGF−β1 sensitivity, TCR0− and TCR3−T cells were co−cultured with target cells across increasing TGF−β1 concentrations. In the absence of neoantigen, both TCR types exhibited minimal baseline Foxp3 expression, which was unaffected by TGF−β1. However, upon stimulation with KRAS G12V_8-16−pulsed T2−A1101 cells, responses diverged: TCR0−CD8⁺ T cells displayed strong, dose−dependent Foxp3 induction, whereas TCR3−CD8⁺ T cells showed markedly reduced upregulation (Figures 6A–C; Supplementary Figure 13). Notably, when responding to SW480−A1101 cells, TCR3−CD8⁺ T cells maintained intermediate Foxp3 levels despite TGF−β1 exposure, while TCR0 cells remained Foxp3−negative (Figures 6E, F; Supplementary Figure 14). This antigen−specific modulation suggested that TCR3 resistance is activation−dependent. Functional assays supported these observations: TCR3−CD8⁺ T cells preserved proliferative capacity under TGF−β1, whereas TCR0-T cells were significantly inhibited (Figure 6D, Supplementary Figure 15). Together, these results identify TCR-T cell avidity optimization as a strategy to overcome TGF−β1−mediated dysfunction. The dual resistance of TCR3-T cells to both Treg conversion and proliferation arrest underscores its potential for sustained efficacy in immunosuppressive tumor microenvironments.

To assess the migratory capacity and immune-modulatory effects of TCR3-T cells, we employed a transwell co-culture system with TCR-T cells and PBMCs in the upper chamber and target cells below. TCR3-T cells demonstrated superior infiltration and recruitment of CD11c⁺ dendritic cells/macrophages toward antigen-positive SW480-A1101 cells compared to antigen-negative HCC-366 cells or TCR0-T cells (P < 0.001; Figure 7A; Supplementary Figure 16). While TCR0-T cells showed modest infiltration and recruitment relative to SW480-A1101 cells, both metrics significantly exceeded background levels observed with HCC-366 cells or untransduced CD8⁺ T cells (NC) (Figure 7A, Supplementary Figure 16). Monocyte (CD14⁺) and polyclonal CD3⁺ T cell recruitment was minimal across all groups, though SW480-A1101 cells elicited slightly higher recruitment with TCR3-T and TCR0-T cells compared to HCC-366 cells (Figure 7A, Supplementary Figure 16). No intergroup differences were observed in the upper chamber, where ~20% of monocytes and CD3⁺ T cells remained (Figure 7A, Supplementary Figure 16). These results indicated that TCR3-T cells selectively enhanced infiltration and CD11c⁺ cell recruitment upon antigen engagement.

Chemokine profiling revealed mechanistic insights. TCR3-T cells co-cultured with SW480-A1101 cells secreted significantly elevated levels of CXCL8, CCL2, CCL3, and CCL4, along with the T-cell chemoattractants CXCL9, CXCL10, and CCL17 (P < 0.01 vs. TCR0-T or HCC-366 controls; Figure 7B; Supplementary Figure 17). In contrast, CXCL11, CXCL5, CXCL1, CCL20, and CCL11 secretion was negligible and comparable to controls. TCR0-T cells and untransduced CD8⁺ T cells (NC) showed no chemokine induction. This chemokine signature aligned with the observed infiltration and recruitment phenotypes, suggesting TCR3-T cells orchestrate an antigen-dependent immune microenvironment.

To evaluate off-target reactivity, TCR3-CD8⁺ T cells were co-cultured with 13 primary human cell types, including HUVEC, WI38-VA13, HH, HRA, HSF, HEF, HISMC, HBISMC, HOF, HSkMM, HAmb, HBSMC, and HCM. LDH release assays confirmed potent killing of SW480-A1101 cells but no significant increased cytotoxicity against any primary cell type (P > 0.05 vs. untransduced CD8⁺ T cells) (Figure 8A). TCR0-T cells showed no activity against SW480-A1101 or primary cells. Activation marker CD137 further validated specificity. TCR3-T cells exhibited robust upregulation of CD137 upon SW480-A1101 recognition but not with HUVECs or other primary cells (background levels; Figure 8B, Supplementary Figure 18). TCR0-T cells showed minimal CD137 expression. Structural basis of specificity was further detected. Alanine scanning identified seven critical residues (G3, A4, V5, G6, V7, and ancho residues V2, K9) in the KRAS G12V_8–16 peptide, where substitutions reduced TCR3-T cell cytotoxicity by >50% (Figure 8C). Two additional residues (V1, G8) contributed moderately (>15% reduction, Figure 8C). This stringent dependence on the KRAS peptide-HLA interface (26) explained the absence of off-target recognition.