PD-1 and TIGIT coexpression enables selective enrichment of clonally expanded tumor-reactive CD8 T cells for melanoma TIL therapy Authors: Kathleen Ducoin, Tiffany Beauvais, James Goward, Mawuena Ahondo, Sylvia Lambot, Aurélie Mordelet, Anaïs Daminette, Morane Thabot, Jixin Deng, Song Tian, Samuel Rulli, Gaëlle Quereux, Amir Khammari, Nathalie Labarrière Published: 2026 ABSTRACT AbstractBackgroundAlthough anti-PD-1 antibodies are approved as first-line treatment for patients with metastatic melanoma (MM), many patients remain resistant. Adoptive cell therapy (ACT) using tumor-infiltrating lymphocytes (TILs) represents a promising alternative or complementary strategy, but heterogeneous clinical response rates indicate that TIL-ACT still requires optimization, particularly through enrichment of tumor-specific T lymphocytes. We recently identified a circulating CD8⁺ T-cell population, termed DPOS, defined by PD-1 and TIGIT coexpression, whose frequency correlates with anti-PD-1 efficacy in MM patients. Because these activated CD8⁺ T cells are enriched in melanoma antigen-specific lymphocytes, we hypothesized that this biomarker combination could enable the selective isolation of TILs with high therapeutic potential.MethodsDeep immunophenotyping, T-cell receptor (TCR) sequencing, in vitro functional assays, and in vivo patient-derived xenograft models were used to characterize and evaluate the antitumor reactivity of the DPOS subset. A clinically compatible workflow for selective isolation and expansion of DPOS TILs was developed using flow cytometry sorting and ex vivo expansion.ResultsDPOS TILs displayed high expression of activation, costimulatory, tissue residency, and pre-exhaustion markers, consistent with an effector-memory phenotype, while maintaining proliferative capacities similar to other CD8⁺ TILs. TCR sequencing and ELISpot assays demonstrated that the DPOS subset is enriched in clonally expanded tumor antigen-specific T cells, with limited overlap between dominant DPOS clonotypes and those from the remaining CD8⁺ TIL compartment. Co-culture assays against autologous tumor cell lines demonstrated that DPOS TILs mediate stronger tumor recognition, characterized by increased cytotoxicity, cytokine secretion, and proliferation following antigen encounter. These functional advantages translated into improved tumor control in two patient-derived xenograft models.ConclusionPD-1⁺TIGIT⁺ CD8⁺ TILs define a distinct tumor-reactive population within human melanoma lesions. Our findings support the use of PD-1/TIGIT coexpression as a clinically applicable strategy to enrich functionally tumor-reactive CD8⁺ T cells and provide a rationale for biomarker-guided optimization of TIL-based adoptive cell therapy in melanoma.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12967-026-08459-6.Keywords: Melanoma, Immunotherapy, Adoptive cell transfer, Tumor infiltrating lymphocytes, Xenograft models Background Tumor-infiltrating lymphocyte (TIL) therapy consists in the infusion of unmodified autologous, ex vivo–expanded T cells isolated from tumor lesions, independently of the immunosuppressive tumor microenvironment. TIL products are typically composed of oligoclonal effector T cells, among which only a variable fraction recognizes a heterogeneous repertoire of tumor-associated antigens [1–3]. TIL therapy has been pioneered in metastatic melanoma in the late 1980s by Rosenberg and colleagues [4] and remains one of the most effective forms of adoptive cell transfer (ACT) developed to date. The clinical landscape of melanoma was profoundly revolutionized by the approval of anti–programmed cell death protein 1 (PD-1) antibodies in 2014 [5], temporarily overshadowing TIL therapy. However, over the past three years, ACT using TILs has experienced a major resurgence. Two independent randomized clinical trials demonstrated the safety and efficacy of TIL therapy in heavily pretreated metastatic melanoma patients refractory to immune checkpoint inhibitors (ICIs) and targeted therapies [6, 7]. These results culminated in the first FDA approval of a TIL product, Lifileucel, in early 2024 for patients with advanced refractory melanoma. In parallel, encouraging results from a phase II trial [8] have led to the initiation of the TILVANCE-301 phase III study evaluating Lifileucel efficacy combined with pembrolizumab as first-line therapy compared with pembrolizumab alone (NCT05727904). Despite these advances, clinical response rates remain heterogeneous, ranging approximately from 33% to 49% in refractory patients [6, 7, 9], highlighting a major need to further optimize TIL products. Multiple studies have shown that therapeutic efficacy correlates with both the proportion of CD8⁺ T cells in infused products [2, 9–13] and the frequency of tumor-reactive CD8⁺ T lymphocytes [10, 12, 14]. Persistence, intratumoral proliferation and functional fitness of these tumor-specific CD8⁺ T cells after infusion are also critical determinants of clinical outcome [11, 12]. Together, these observations suggest that enriching TIL products in tumor-reactive, polyfunctional CD8⁺ T-cell subsets with sustained effector competences could substantially enhance therapeutic benefit. Immune checkpoint and costimulatory receptors expressed on activated T cells represent attractive candidates for such selective enrichment strategies. In this context, we recently identified a circulating population of cytotoxic follicular T-cell-like CD8⁺ T lymphocytes defined by the co-expression of PD-1 and T-cell immunoreceptor with Ig and ITIM domains (TIGIT), and enriched in CXCR5⁺ cells, compared to PD-1 or TIGIT single-positive subsets as well as double-negative cells [15]. Strikingly, only the frequency of these PD-1⁺TIGIT⁺ CD8⁺ T cells (hereafter referred to as DPOS cells) was strongly associated with clinical response to anti–PD-1 therapy in three independent cohorts of cancer patients, including two melanoma cohorts and one Merkel cell carcinoma cohort. This DPOS population exhibited hallmarks of recent activation and proliferation and was enriched in tumor-specific and newly emerging T-cell clonotypes, compared to other CD8+ T-cell subsets. These features suggested that DPOS cells may represent a highly relevant pool of tumor-reactive effector T cells with potential utility for ACT strategies. Consistent with this hypothesis, previous studies reported higher expression of PD-1 and TIGIT on CD8⁺ TILs contained in TIL products from responding metastatic melanoma patients [11, 12, 16]. However, whether the deliberate selection of PD-1⁺TIGIT⁺ CD8⁺ TILs could generate a functionally superior therapeutic product has not been directly addressed. Here, we investigate the frequency, phenotype and therapeutic potential of DPOS cells within TILs derived from human melanoma tumors. We also developed a rapid and scalable protocol allowing the selective isolation and expansion of PD-1⁺TIGIT⁺ CD8⁺ TILs in clinically compatible conditions. We demonstrate that this polyclonal DPOS TIL population is highly enriched in tumor-reactive clonotypes, displays a distinctive activated phenotypic profile characterized by TOX, TCF-1 and 4-1BB expression, and exhibits robust proliferative and cytotoxic capacities. Together, our findings identify PD-1⁺TIGIT⁺ CD8⁺ TILs as a functionally superior T-cell subset and provide a rational strategy to optimize TIL-based immunotherapy for melanoma patients. Methods Patients and specimens Twenty metastatic melanoma patients undergoing surgery for metastatic lymph node (mLN) chain removal were included in this study. Those patients were enrolled in the unit of Dermato-Cancerology at the University Hospital of Nantes (France) between 2013 and 2024. All materials were processed according to the guidelines of the French Ethics Committee for research on human tissues, and to the principles outlined in the declaration of Helsinki. Clinical informations were obtained from the RIC-Mel database, a French national registry dedicated to the follow-up of melanoma patients. The RIC-Mel database (Clinical Trials n°NCT03315468) gathers data from 49 participating centers in different French regions. It received ethics committee approval on February 9th, 2012 (no. 12.108) from the Independent Ethics Committee in Paris and received authorization from the French Data Protection Agency (CNIL, DR-2012-259, 28 May 2012). All the patients included in this study signed an inform consent. Supplementary Table S1 lists the clinicopathological features of these 20 metastatic melanoma patients. Processing of metastatic lymph nodes and cell sorting Metastatic lymph nodes were processed immediately after surgery and mechanically dissociated into 1–3 mm3 fragments in RPMI 1640 medium (Gibco) supplemented with 8% human serum (HS, PanBiotech), penicillin (Gibco, 100 U/mL) and streptomycin (Gibco, 0.1 mg/mL). Cell suspensions were filtered and subjected to double-density gradient centrifugation using Ficoll (Eurobio) to obtain fractions enriched in tumor cells and lymphoid cells, which were cryopreserved for subsequent analyses. Cells were stained in Brilliant Stain Buffer (BD Biosciences) with anti-human CD8-APC, PD-1-BV421 and TIGIT-PE-Cy7 monoclonal antibodies (mAbs). Cell sorting was performed on a BD FACSAria III cell sorter using FACSDiva software (BD Biosciences). For two patients treated with anti–PD-1 antibody in the neoadjuvant setting, PD-1 detection was performed using both anti-human PD-1 antibody and anti-human IgG4-biotin, followed by streptavidin-BV421 (BioLegend), allowing detection of total PD-1 molecules irrespective of occupancy by therapeutic antibody. Two CD8⁺ TIL populations were isolated: (i) DPOS TILs, defined as PD-1⁺TIGIT⁺ CD8⁺ T cells; (ii) nDPOS TILs, comprising PD-1⁺TIGIT⁻, PD-1⁻TIGIT⁺ and PD-1⁻TIGIT⁻ CD8⁺ T cells. This dissociation and sorting procedure is covered by a pending patent (EP 25 305 821.8) and is detailed on Fig. 1D. PD-1⁺TIGIT⁺ CD8⁺ T cells infiltrating melanoma metastatic lymph nodes are associated with overall survival and can be prospectively isolated. A, Frequencies of total CD8⁺ TILs (left) and CD8⁺ PD-1⁺TIGIT⁺ (DPOS) TILs (right) in melanoma metastatic lymph nodes (n = 20). Bars indicate median values. B, Representative contour plots showing PD-1 and TIGIT expression on CD8⁺ TILs from melanoma metastatic lymph nodes of patients P5, P6 and P14. Frequencies of DPOS TILs are indicated. C, Overall survival of patients with skin cutaneous melanoma (SKCM) according to low (blue) or high (red) expression of the PDCD1, TIGIT and CD8A gene signature (data from gepia2.cancer-pku.cn). D, Schematic representation of the developed and patented procedure for ex vivo sorting of DPOS and nDPOS TILs from melanoma metastatic lymph nodes Two populations were sorted: DPOS TILs, characterized by the coexpression of PD-1 and TIGIT among CD8+ TILs, and nDPOS TILs, the rest of CD8+ TILs containing PD-1+/TIGIT, PD-1−/TIGIT− and PD-1−/TIGIT+ cells. In vitro TIL expansion DPOS and nDPOS TILs were expanded in RPMI 1640 medium supplemented with 8% HS, penicillin, streptomycin, GlutaMAX-I (2 mM, Gibco) and recombinant human IL-2 (150 U/mL, Proleukin, Novartis). T cells were stimulated with phytohemagglutinin-L (Sigma-Aldrich) in the presence of irradiated allogeneic feeder cells (peripheral blood mononuclear cells from two healthy donors and EBV-transformed B cells), as previously described [17]. Cultures were maintained for 20 days with regular monitoring of cell density. Purity and phenotypic characterization of amplified TILs Purity and phenotype of amplified DPOS and nDPOS TIL populations were assessed by flow cytometry. Briefly, cells were incubated with Human BD Fc Block (BD Biosciences) for 10 min at room temperature to prevent non-specific binding. Viable cells were stained with fluorescently labeled antibodies targeting T-cell lineage and differentiation markers, including CD3, TCRαβ, CD4, CD8, PD-1, TIGIT, and additional phenotypic markers (CCR7, CD45RA, CD27, CD28, CD25, CD226, 4-1BB, CD39, CD103, Tim-3, Lag3) for 30 min at 4 °C except for CCR7 mAb, incubated 60 min at 37 °C to optimize staining quality. After washing, cells were acquired on a BD FACSymphony A5 flow cytometer (BD Biosciences) and analyzed using FlowJo v10.10 (purity) or OMIQ online platform (phenotypic characterization). For intranuclear staining of TCF-1 and TOX, Foxp3/transcription factors staining buffer set (eBioscience) was used, following manufacturer’s instructions. For full details of fluorochromes, suppliers, catalog numbers, and RRIDs, see Supplementary Table S2. RNA extraction Total RNA from amplified DPOS and nDPOS TIL populations for the 20 patients was extracted using sequentially RNeasy Mini Kit, RNase-Free DNase Set and RNeasy Micro Kit (all Qiagen) in triplicates, following manufacturer’s instructions. RNA quantification was determined with a NanodropND-1000 spectrophotometer (Thermo Fisher Scientific) and quality was assessed using RNA ScreenTape (Agilent Technologies) on a 4150 TapeStation system (Agilent Technologies), according to the manufacturer’s instructions. TCR sequencing For NGS based analysis of T-cell clonotypes, 100ng of RNA was used to create libraries with QIAseq Targeted RNA Panel Human TCR with a panel consisting of alpha, beta, gamma, and delta TCR. The resulting libraries were sequenced using illumina NextSeq 2000 and illumina NextSeq 6000. Bioinformatic and statistical analysis of TCR repertoires The resulting FastQ files were used for read mapping using the UMI-nea pipeline [18] which is available in the QIAGEN GeneGlobe Analyze site (https://geneglobe.qiagen.com/analyze). The clonotype calls are generated using the IMSEQ software published by researchers at the Charité Unverstitätsmedizin, Berlin. The main tasks of IMSEQ are to align the reads to model V-region and J-region sequences and extract the CDR3 region sequence [19]. To remove low-confidence TCR clonotypes, particularly those supported by few UMIs and therefore more likely to arise from uncorrected sequencing or PCR errors, we applied a knee-plot–based filtering strategy. Clonotypes generated by the GeneGlobe pipeline were ranked by UMI count for each TCR subunit. The knee point of the ranked UMI distribution was identified using the KneeLocator implementation in the Python package kneed, based on the Kneedle algorithm [20]. Clonotypes located before the knee point were retained, whereas those in the low-abundance tail beyond the knee point were excluded from downstream analyses. TCRα and TCRβ sequences were processed in R (v.4.4.1, tidyverse). Technical replicates were aggregated by median to obtain patient- and fraction-level (DPOS and nDPOS) datasets normalized to 100% total frequency. Sequencing depth, repertoire richness, and Shannon diversity index were computed for quality control. Clonal frequencies were aggregated by the median of technical replicates at the patient and fraction level (DPOS and nDPOS), then renormalized so that the sum of the proportions of each repertoire reached 100%. This consolidated dataset served as the basis for all analyses. Repertoire structure was assessed using the D50 index (minimum clonotypes covering 50% of the repertoire) and cumulative frequency of the top 20 clonotypes. Top 20 DPOS clonotypes were tracked in the nDPOS fraction to evaluate clonal sharing, visualized via alluvial plots, for each patient and globally through cumulative frequencies. The dynamics of these sequences were represented by alluvial diagrams (ggalluvial package), where each flow follows the evolution of the frequency of a clonotype between the two compartments. Clonotypes were further stratified into four abundance classes (“Top 20,” “21–50,” “51–100,” “>100”) to compare homeostasis profiles between fractions. All analyses were performed identically for TRAC and TRBC chains. IFN-γ ELISpot assays Peptide-specific IFN-γ secretion by DPOS and nDPOS TILs was measured using Human IFN-γ ELISpot Pro (Mabtech) according to the manufacturer’s instructions. TILs (3.3 × 10⁵ per well) were incubated for 24 h with 30 µM of tested peptides, anti-CD3 mAb (positive control), or medium alone (negative control) in triplicate wells. After washing, spots were developed with anti-IFN-γ-ALP and BCIP/NBT substrate and counted using an ASTOR ELISpot reader (Mabtech). Spot counts were background-subtracted, normalized per 10⁶ TILs, and responses > 30 SFU/10⁶ and above background were considered positive. Peptides (listed in Supplementary Table S3) were synthesized by Genecust at > 85% purity. Autologous tumor cell line generation Two or three fresh or cryopreserved mLN fragments were cultured per well in 24-well plates in 1.5mL of RPMI 1640 medium supplemented with 1mM L-glutamine (Gibco) and either with 8% HS or 10% FBS, as previously described [17, 21]. Alternatively, ex vivo enriched tumor cell suspensions were cultured in the same two media. Cultures were monitored weekly and expanded for downstream functional assays. TIL reactivity assays against autologous tumor cells Cryopreserved tumor-enriched cell suspensions were thawed with DNAse I (STEMCELL Technologies) according to the manufacturer’s instructions. DPOS and nDPOS TILs were cocultured with autologous tumor cells at an effector-to-target ratio of 1:2 for 3 h at 37 °C to assess degranulation using anti-CD107a (BD Biosciences). For cocultures with autologous tumor cell lines, tumor cells were optionally pre-treated with recombinant human IFN-γ (500 U/mL, Eurobio) for 48 h. Degranulation assays were performed as above, and effector cytokine production (IFN-γ, TNF-α, IL-2) was assessed following a 5-hour coculture in the presence of brefeldin A (10 µg/mL). TILs were stained with anti-CD8, fixed, and intracellular staining was performed with cytokine antibodies (Supplementary Table S2). Cells were then acquired on a BD FACSCanto II cytometer, and analyzed using FlowJo v10.10. Cytotoxicity assays Cytotoxic activity of DPOS and nDPOS TILs was assessed using the Calcein-AM release assay (BioLegend) following the manufacturer’s instructions. Autologous tumor cells were labeled with 9 µM Calcein-AM for 20 min at 37 °C and cocultured for 5 h at 37 °C with TILs at effector: target ratios of 5:1, 10:1, or 50:1 in triplicate wells. Spontaneous (SR) and maximal (MR) calcein releases were measured from tumor cells alone or treated with 5% Triton-X100, respectively. Fluorescence of supernatants was read using a FLUOstar Omega microplate reader (excitation 488 nm, emission 520 nm). Percent lysis was calculated as: (Experimental release–SR)/(MR–SR)×100. CFSE proliferative assay To evaluate DPOS proliferation abilities, CFSE assays were performed using CellTrace CFSE (invitrogen), following manufacturer’s instructions. Briefly, DPOS TILs were incubated with CFSE at 0.25µM during 20 min at 37 °C. After washes, CFSE-stained DPOS TILs were cocultured with IFN-γ-pretreated (500UI/mL, 48 h) or untreated autologous tumor cell lines for 3 or 5 days, at an effector: target ratio of 1:2. DPOS TILs were stained with anti-human CD8-PE mAb prior to acquisition on a BD FACSCanto II flow cytometer using BD FACSDiva software 8.0 and data were analyzed with FlowJo software v10.10. Mouse xenograft models Six to eight-week-old female NSG mice (Charles River laboratory) with unrestricted access to food and water were kept under specific pathogen-free conditions in the UTE animal facility (SFR Santé, Nantes, licence number: D44278). The animal experiments were performed in accordance with the recommendations established by the Ethic Committee for Animal Experiments of Nantes University (APAFIS #47663-202402201204956 v4) and the FELASA (Federation of Laboratory Animal Science Associations). Subcutaneous xenograft tumors were established by injections of respectively 5 × 106 or 2 × 106 human melanoma cells, in matrigel, from P16 and P20 patients, into the flank of NSG mice. At day five, 7 to 8 tumor-bearing mice were randomly allocated into three groups with homogeneous tumor volumes which received intravenously either: DPBS, 5 × 106 nDPOS TILs and 5 × 106 DPOS TILs in DPBS. Intravenous injections were repeated three times at days 15, 22 and 29 after engraftment. For P16 patient, an early TIL injection was performed at day 7 after engraftment. Tumor burdens were measured by an electronic caliper, and the tumor volume was calculated based on the following formula: volume = Length x Width x Height. Mice were sacrificed taking into account the appearance of necrosis in tumors, weight loss (20% of initial weight) and tumor size (> 1000 mm3), in accordance with national and international policies. To analyze individual tumor growth curves, we defined a threshold corresponding to three times the tumor volume measured at day 5, which was used as the baseline for tumor growth assessment. This threshold was used to classify mice as having controlled or non-controlled tumor growth. Based on this criterion, threshold volumes were set at 160 mm³ for P16 tumors and 100 mm³ for P20 tumors, corresponding to threefold the mean tumor volume at D5 in each cohort. Statistics Statistical analyses were performed using GraphPad Prism 10 and R softwares, using non parametric tests. All statistical details can be found in the figure legends. Comparisons between DPOS and nDPOS groups for paired phenotypic and functional variables were performed using paired Wilcoxon signed-rank tests, with confidence interval of 95%. For in vitro cytotoxicity assays, statistical analysis was performed using two-way ANOVA followed by Tukey’s multiple comparisons test to account for all pairwise comparisons. For CFSE assay, statistical analysis was performed with Friedman test, followed by Dunn’s multiple comparisons test. For in vivo tumor growth experiments, differences between groups over time were analyzed using two-way analysis of variance (two-way ANOVA), followed by Dunnett’s multiple comparisons test comparing each experimental group to the control condition (DPBS group). All tests were two-sided, and a p-value < 0.05 was considered statistically significant (*), p < 0.01 (**), p < 0.001 (***) and p < 0.0001 (****) as highly significant. For non-significant results (ns) was indicated. Results Frequency and sorting of PD-1+TIGIT+ (DPOS) CD8+ TILs from melanoma tumors We first examined the presence of PD-1⁺TIGIT⁺ CD8⁺ T cells, called DPOS TILs, within tumor lesions from 20 stage III melanoma patients. CD8⁺ TILs represented 1.5 to 36.7% of viable lymphocytes (mean of 13.8%, Fig. 1A, left panel), and among them, the DPOS subset accounted for 6.9 to 95.0% (mean of 46.0%) (Fig. 1A, right panel). Representative flow cytometry profiles illustrating the heterogeneity of DPOS frequencies across patients are shown in Fig. 1B. We explored whether baseline DPOS frequencies within melanoma lesions could be associated with clinicopathological parameters or prior responses to immune checkpoint blockade. No correlation was found between DPOS TIL frequencies and clinical characteristics, indicating the presence and infiltration of these cells in all lymph node metastases, regardless of prior treatment or mutational status. However, we considered that our cohort was not sufficiently powered nor clinically homogeneous to support robust correlation analyses. To establish the relevance of PD-1 and TIGIT as markers for the selection of therapeutically valuable TIL subsets, we further performed an in silico survival analysis using transcriptomic data from 458 melanoma patients (SKCM cohort, GEPIA2 [22]). High expression of a combined PDCD1, TIGIT and CD8A gene signature was significantly associated with improved overall survival (Fig. 1C), supporting the clinical relevance of these markers as candidates for tumor-specific TIL enrichment. To enable both functional and molecular characterization of this population, we developed a rapid and clinically compatible workflow starting from mechanically dissociated tumor fragments followed by density gradient separation. This procedure generated two complementary cellular fractions: (i) a lymphoid fraction enriched in TILs and (ii) a tumor cell-enriched fraction, preserved for downstream autologous functional assays. Lymphoid cells were immediately sorted based on CD8, PD-1 and TIGIT expression by flow cytometry, yielding two distinct populations: PD-1⁺TIGIT⁺ CD8⁺ T cells (DPOS) and a complementary nDPOS fraction comprising single positive PD-1⁺, single positive TIGIT⁺ or double-negative CD8⁺ T cells (Fig. 1D). Post-sort analyses confirmed high purity of both subsets in terms of CD8 expression and PD-1/TIGIT co-expression (Supplementary Fig. S1). This integrated strategy enabled the reproductible isolation of DPOS TILs together with autologous tumor cells in quantities compatible with downstream functional and repertoire analyses. DPOS TILs display a distinct activated and pre-exhausted phenotype We sorted between 70 and 6 × 105 DPOS TILs (mean of 1.3 × 105) and between 312 and 4.7 × 105 nDPOS TILs (mean of 1.8 × 105). To determine whether DPOS and nDPOS TILs differ in their biological properties, we compared their proliferative capacity following standard feeder-based expansion. Comparable numbers of cells were obtained from both populations (respective means of 1.7 × 109 and 7.8 × 10⁹ cells), corresponding respectively to approximately 14 and 15 cell divisions for DPOS and nDPOS subsets, after 20 days of culture (Fig. 2A and Supplementary Fig. S2A-B). Although nDPOS subset exhibits slightly better proliferation ability, these results indicate that DPOS TILs retain robust proliferative potential compatible with therapeutic manufacturing. Expanded DPOS TILs display a distinct phenotypic profile compared with nDPOS TILs. A, Absolute numbers of DPOS (left, green) and nDPOS (right, grey) TILs (n = 20) at the beginning (day 0) and at the end (day 20) of the expansion process. Statistical significance was assessed using a paired Wilcoxon test. Bars indicate median values. B, Dimensionality reduction analysis using t-SNE of all TILs (left, n = 40), DPOS TILs (middle, n = 20) and nDPOS TILs (right, n = 20). C, Expression of the indicated markers displayed as histograms or overlaid on t-SNE representations. For histograms, DPOS (n = 20) and nDPOS (n = 20) TILs are shown in green and grey, respectively. Percentages of positive cells are indicated. D, Percentage of cells positive for the indicated markers among DPOS (green) and nDPOS (grey) TILs (n = 20). Statistical significance was assessed using a paired Wilcoxon test. Bars indicate median values We first confirm the purity of amplified fractions, in terms of CD3 and TCRαβ expression (not shown). We next performed deep phenotypic profiling of expanded DPOS and nDPOS TILs using a 14-marker multiparametric flow cytometry panel. Unsupervised analysis revealed a clear segregation between the two populations (Fig. 2B), demonstrating that TIGIT and PD-1 expression identifies a biologically distinct CD8⁺ T-cell subset rather than a stochastic fraction of activated cells. Interestingly, beyond PD-1 and TIGIT, DPOS TILs exhibited significantly higher expression of co-stimulatory CD27, CD28 and 4-1BB markers, despite higher frequencies of CD39, described as an exhaustion marker, than nDPOS TILs (Fig. 2C-D). Associated with high expression of the two transcription factors TOX and TCF-1, DPOS TILs harbor a phenotype consistent with an activated yet partially differentiated “pre-exhausted” state (Fig. 2C-D). Notably, DPOS TILs were enriched in a CD103⁺ subset, suggesting preferential representation of tissue-resident TILs. Moreover, a fraction of DPOS cells coexpress CD39 and CD103 markers, associated with tumor-reactive T cells [23], while nDPOS cells barely express them. In contrast, naive T cells (CCR7⁺CD45RA⁺) were exclusively detected within the nDPOS fraction (Supplementary Fig. S3D-E). The terminally exhaustion marker Tim-3 is slightly overexpressed by nDPOS TILs and expression levels of other inhibitory/activation receptors, including LAG3, CD25 and CD226, were not significantly different between the two subsets (Supplementary Fig. S3D-E). Together, these results identify PD-1⁺TIGIT⁺ CD8⁺ TILs as a phenotypically distinct population characterized by features of activation, tumor residency and a differentiation state associated with functional persistence. DPOS TILs are enriched in clonally expanded T-cell clonotypes To investigate whether the phenotypic differences observed between DPOS and nDPOS TILs were associated with distinct T-cell receptor (TCR) repertoires, we performed deep sequencing of TCRα and TCRβ chains from both populations. Comparable numbers of sorted cells were analyzed for each subset (Fig. 3A, left panel), allowing unbiased assessment of repertoire diversity. TRBC repertoire of DPOS TILs is specifically enriched in expanded T-cell clonotypes. A, Number of cells sorted by flow cytometry (left) and number of TRBC clonotypes (right) in DPOS (green) and nDPOS (grey) TILs (n = 20). Statistical significance was assessed using paired Wilcoxon tests. B, Relationship between the number of TRBC clonotypes and the number of sorted nDPOS (grey, top) or DPOS (green, bottom) TILs (n = 20). R² values and 95% confidence intervals are indicated. C, Frequency of the top 20 TRBC clonotypes (left) and number of TRBC clonotypes required to reach 50% of the repertoire (D50 index, right) in DPOS (green) and nDPOS (grey) TILs (n = 20). Statistical significance was assessed using paired Wilcoxon tests. D, Cumulative frequency of the top 20 TRBC clonotypes from the DPOS fraction, shared or not with the corresponding nDPOS TILs, visualized by alluvial plots. Each color represents one patient (left). Representative examples showing no sharing, limited sharing, or complete sharing of clonotypes for patients P17, P3 and P1 are shown (right panels), with each color corresponding to one TRBC clonotype DPOS TILs exhibited a markedly reduced TCR diversity compared with nDPOS TILs for both TRAC and TRBC repertoires (Fig. 3A right panel and Supplementary Fig. S3A). In the nDPOS population, the number of detected clonotypes correlated with the number of input cells, consistent with a largely polyclonal repertoire. In contrast, no such correlation was observed in the DPOS fraction (Fig. 3B and Supplementary Fig. S3B), suggesting preferential amplification of selected clonotypes. This bias was further confirmed by analysis of clonotype frequency distributions, which revealed a significantly higher contribution of the top 20 clonotypes to the total repertoire in DPOS TILs compared with nDPOS TILs (Fig. 3C left panel and Supplementary Fig. S3C). Consistently, fewer clonotypes were required to account for 50% of the repertoire in DPOS TILs (D50), indicating pronounced oligoclonality (Fig. 3C, right panel and Supplementary Fig. S3C). The lower diversity of the DPOS TIL repertoire is confirmed at the global level, with a greater amplification of clonotypes (Top 1–20, 21–50, and 51–100), while clonotypic diversity is greater for nDPOS, as illustrated by the abundance of less represented clonotypes (shown in grey in Supplementary Fig. S3D). Importantly, dominant clonotypes identified within the DPOS fraction were either absent or present at very low frequency in the nDPOS compartment, demonstrating that this clonal expansion selectively occurs within the PD-1⁺TIGIT⁺ subset, as illustrated by cumulative frequencies of top 20 DPOS clonotypes in Fig. 3D. Different profiles were observed among the 20 patients with regard to the sharing of the top 20 DPOS clonotypes. For 4 patients (P6, P8, P12 and P17), none of these 20 clonotypes were detected in the nDPOS fraction (Supplementary Fig. S4). For 11 patients, some of the 20 clonotypes are found at low frequencies in the nDPOS TILs, and for the remaining 5 patients, this sharing concerns either all clonotypes found at low frequencies in nDPOS TILs, or a more significant sharing of particular clonotypes (Fig. 3D and Supplementary Fig. S4). These findings indicate that DPOS TILs are enriched in clonally expanded T cells likely shaped by antigen-driven selection, consistent with a population of activated tumor-reactive lymphocytes. DPOS TILs are enriched in tumor-antigen specific T lymphocytes To directly assess the tumor antigen specificity of DPOS and nDPOS TILs, we compared their reactivity against a panel of 49 peptides derived from 24 shared tumor antigens, including melanocytic differentiation antigens, tumor-associated antigens and unconventional antigens generated by alternative transcription or translation pathways (Supplementary Table S3). Peptides were selected according to the HLA class I molecules expressed by each patient and tested using IFN-γ ELISpot assays. Across the cohort, DPOS TILs exhibited a markedly broader and more frequent antigen recognition profile than their nDPOS counterparts (Fig. 4A, left panel). Fourteen of twenty DPOS TIL populations (70%) recognized between one and ten antigenic peptides, whereas only six of twenty nDPOS TIL populations (30%) displayed detectable reactivity, against one or two peptides (Fig. 4A, right panel). Both the number of recognized antigens and the number of peptide–HLA complexes were significantly higher in the DPOS subset (Fig. 4A), indicating an enrichment of tumor-specific T cells within this population. DPOS TILs are enriched in diverse antigen-specific T cells. A, Number of tested and recognized antigens (left) and peptides (right) by DPOS and nDPOS TIL populations (n = 20). B-E, Spot-forming units (SFU) per 10⁶ DPOS (green) or nDPOS (grey) TILs for all recognized peptides, measured by IFN-γ ELISpot assays (left). Representative ELISpot wells from one or two patients are shown (right). Positive wells are circled in green (DPOS) or grey (nDPOS). Detailed peptide-specific responses are grouped according to HLA class I restriction: (B) HLA-A*02:01, (C) HLA-A*03:01, (D) HLA-A*24:02, (E) HLA-B*35:01 and (F) for single peptide responses in other HLA-contexts. Statistical significance was assessed using paired Wilcoxon tests This pattern was consistently observed across multiple HLA contexts. Among the seven HLA-A*02:01-positive patients, all DPOS TILs recognized 11 out of the 22 tested peptides (between one and six recognized peptides per patient). In contrast only five nDPOS TIL populations responded to four peptides (one or two peptides per patients), all derived from unconventional antigens, except for Melan-A peptide (Fig. 4B; Table 1 and Supplementary Fig. S5). Similarly, one of the three HLA-A*03:01 DPOS TIL population exhibited reactivity against all four tested A*03:01-restricted peptides, and among the eight HLA-A*24:02 patients, three DPOS TIL populations recognized three of four peptides, while no detectable responses were observed in the corresponding nDPOS fractions (Fig. 4C-D; Table 1 and Supplementary Fig. S5). Table 1Recognized peptides and reactive TIL populations HLA-A*02:01 7 patients Recognized peptides (11/22) Number of reactive DPOS TIL populations (7/7) Number of reactive nDPOS TIL populations (3/7) Melan-AEAA71SSX2KAS50gp100YLE30gp100KTW20MAGEC2ALK20TYRYMD10GnTVVLP10MMP2GLP01LAGE1MLM01Linc00518PCL45AC136944.2FLP10 HLA-A*03:01 3 patients Recognized peptides (4/4) Number of positive DPOS TIL populations (1/3) Number of positive nDPOS TIL populations (0/3) TAGRLS10gp100RSY10gp100ALL10gp100ALN10 HLA-A*24:02 8 patients Recognized peptides (3/4) Number of positive DPOS TIL populations (5/8) Number of positive nDPOS TIL populations (0/8) OA1LYS30TYRAFL10gp100VYF10 HLA-B*35:01 3 patients Recognized peptides (4/6) Number of positive DPOS TIL populations (3/3) Number of positive nDPOS TIL populations (1/3) NYESO-1MPF20gp100VLP20Melan-AEAA11MAGE-A3EVD10 HLA-B*07:02 4 patients Recognized peptide (1/2) Number of positive DPOS TIL populations (2/4) Number of positive nDPOS TIL populations (0/4) gp100SSP20 HLA-B*38:01 1 patient Recognized peptide (1/1) Number of positive DPOS TIL populations (1/1) Number of positive nDPOS TIL populations (0/1) TYRLHH10 HLA-B*44:03 1 patient Recognized peptide (1/2) Number of positive DPOS TIL populations (1/1) Number of positive nDPOS TIL populations (0/1) MAGE-A3MEV10 HLA-C*08:02 3 patients Recognized peptide (1/2) Number of positive DPOS TIL populations (1/3) Number of positive nDPOS TIL populations (0/3) gp100SND10 Recognized peptides and reactive TIL populations HLA-A*02:01 7 patients Recognized peptides (11/22) Number of reactive DPOS TIL populations (7/7) Number of reactive nDPOS TIL populations (3/7) HLA-A*03:01 3 patients Recognized peptides (4/4) Number of positive DPOS TIL populations (1/3) Number of positive nDPOS TIL populations (0/3) HLA-A*24:02 8 patients Recognized peptides (3/4) Number of positive DPOS TIL populations (5/8) Number of positive nDPOS TIL populations (0/8) HLA-B*35:01 3 patients Recognized peptides (4/6) Number of positive DPOS TIL populations (3/3) Number of positive nDPOS TIL populations (1/3) HLA-B*07:02 4 patients Recognized peptide (1/2) Number of positive DPOS TIL populations (2/4) Number of positive nDPOS TIL populations (0/4) HLA-B*38:01 1 patient Recognized peptide (1/1) Number of positive DPOS TIL populations (1/1) Number of positive nDPOS TIL populations (0/1) HLA-B*44:03 1 patient Recognized peptide (1/2) Number of positive DPOS TIL populations (1/1) Number of positive nDPOS TIL populations (0/1) HLA-C*08:02 3 patients Recognized peptide (1/2) Number of positive DPOS TIL populations (1/3) Number of positive nDPOS TIL populations (0/3) A comparable enrichment was observed in HLA-B and –C-restricted contexts. All three HLA-B*35:01 DPOS TIL populations recognized between one and four peptides, whereas no reactivity was detected in nDPOS fractions, except one specific Melan-AEAA recognition (Fig. 4E and Supplementary Fig. S5). Additional peptide-specific responses were also detected exclusively within the DPOS compartment in B*07:02, B*38:02, B*44:02, B*44:03 and C*08:02 contexts. Together, these data demonstrate that PD-1⁺TIGIT⁺ CD8⁺ TILs are preferentially enriched in tumor antigen–specific lymphocytes across diverse HLA backgrounds and antigen classes. This functional enrichment is fully consistent with the oligoclonal structure observed in DPOS TCR repertoires and supports the notion that clonal expansion within this subset is driven by antigen-dependent selection of tumor-reactive clonotypes. DPOS TILs display superior antitumor activity in vitro To directly assess the tumor reactivity of DPOS and nDPOS TILs, we co-cultured each subset with autologous tumor cell suspensions isolated from the same lesions, using the tumor-enriched fraction generated during the sorting workflow (Fig. 1D). In line with their enhanced peptide-specific reactivity, DPOS TILs exhibited significantly stronger responses against autologous tumor cells, as measured by degranulation capacity (CD107a expression) (Fig. 5A). For 13/20 patients, the proportion of degranulating DPOS TILs ranged from 1% to 34.1%, whereas reactive cells within nDPOS fraction did not exceed 7.6% (Fig. 5A-B), demonstrating a marked enrichment of tumor-reactive T cells within the PD-1⁺TIGIT⁺ subset. DPOS TILs recognize and kill autologous tumor cells. A, Frequency of CD107a⁺ cells among DPOS (green) and nDPOS (grey) TILs (n = 20) after 3-hour coculture with ex vivo tumor-enriched autologous cell suspensions. Statistical significance was assessed using paired Wilcoxon test. Bars represent means. B, Representative dot plots showing CD107a expression by DPOS and nDPOS TILs from patients P7 and P8, unstimulated or after coculture. Frequencies of positive cells are indicated. C, Frequencies of CD107a⁺, IFN-γ⁺, TNF-α⁺, and IL-2⁺ cells among DPOS (green) and nDPOS (grey) TILs (n = 9) after coculture with autologous tumor cell lines pretreated with recombinant human IFN-γ (500U/mL) for 48 h prior to a 5-hour coculture. Statistical significance was assessed using paired Wilcoxon tests. D, Representative histograms showing expression of CD107a, IFN-γ, TNF-α, and IL-2 by DPOS (green) and nDPOS (grey) TILs from patient P8 after coculture with untreated (dashed lines) or IFN-γ-pretreated (bold lines) autologous tumor cell lines. TIL-only controls are shown in white. E, Percentage of autologous tumor cell lysis by DPOS (green) and nDPOS (grey) TILs from patients P16 (top) and P20 (bottom) after 5-hour coculture. Untreated tumor cell lines are shown as dashed lines and IFN-γ-pretreated lines as solid lines. SEM is indicated. Effector-to-target (E: T) ratios are shown. Statistical significance was assessed using two-way ANOVA followed by Tukey’s multiple comparisons test to account for all pairwise comparisons. F, Expansion index of DPOS TILs from patients P8 and P16 after 3-day (left) or 5-day (right) coculture with autologous tumor cell lines, untreated or pretreated with IFN-γ (500U/mL, 48 h), measured by CFSE dilution. Statistical significance was assessed using Friedman test, followed by Dunn’s multiple comparisons test. G, Representative CFSE histograms at days 3 (left) and 5 (right) for P8 DPOS TILs following coculture with untreated (dashed lines) or IFN-γ-pretreated (bold lines) autologous tumor cell lines. TIL-only controls are shown in white. H, Expression of HLA-ABC (blue), PD-L1 (purple) and CD155 (pink) by P8 tumor cell lines, untreated (dashed lines) or IFN-γ pretreated (solid lines). Controls are indicated in black These results were confirmed using autologous tumor cell lines established from nine patients. Upon co-culture with IFN-γ-treated cell lines (to enhance peptide presentation), 7/9 DPOS TILs displayed robust functional activation, characterized by CD107a expression ranging from 1.2 to 70.2%, whereas CD107a degranulating T cells did not exceed 2.2% in nDPOS fractions. The polyfunctionality of DPOS TILs was further documented by the production of IFN-γ, TNF-α and IL-2, whereas nDPOS TILs showed minimal or no detectable responses at all (Fig. 5C-D). We next directly evaluated the cytotoxic capacity of each subset against autologous tumor cell lines. DPOS TILs efficiently lysed tumor cells in vitro, an effect that was further increased following IFN-γ pretreatment of target cells, whereas nDPOS TILs displayed little or no cytotoxic activity (Fig. 5E). Given that DPOS TILs express the inhibitory receptors PD-1 and TIGIT, we finally examined their proliferative capacity upon tumor recognition. DPOS TILs underwent significant proliferation following co-culture with autologous tumor cell lines, irrespective of IFN-γ pretreatment (Fig. 5F-G), indicating preserved proliferative potential upon antigenic stimulation, despite the expression of PD-L1 and CD155 by IFN-γ treated cells (Fig. 5H). Together, these functional assays demonstrate that PD-1⁺TIGIT⁺ CD8⁺ TILs are highly enriched in clonotypes capable of recognizing and killing autologous tumor cells and retain strong effector and proliferative capacities, consistent with a functionally competent, tumor-reactive T-cell state. DPOS TILs display superior antitumor activity in preclinical models As a proof of principle, we evaluated the antitumor efficacy of DPOS and nDPOS TILs derived from patients P16 and P20 in autologous patient-derived xenograft models established in immunodeficient mice. Autologous tumor cells were engrafted subcutaneously, followed by intravenous infusion of either DPOS TILs, nDPOS TILs or vehicle control (DPBS) starting 7 (P16) or 15 days (P20) after tumor implantation and repeated weekly for a total of four or three injections (Fig. 6A). DPOS T cells control autologous tumor growth in two preclinical models. A, Experimental design of P16 and P20-derived tumor cell line xenografts and TIL injections. B, Median tumor volumes of P16 (left) and P20 (right)-derived melanoma tumors in NSG mice receiving either DPBS (black), autologous nDPOS TILs (grey), or DPOS TILs (green). Statistical significance was assessed using two-way ANOVA with Dunnett’s multiple comparisons, with mice receiving DPBS, as control group. C, Individual tumor volumes (Top: P16; Bottom: P20) were calculated as: volume = (Length × Width × Height) and measured from Day 0 to Day 42. Tumors with volumes < initial volume x 3 at day 42, were considered as controlled In both models, DPOS TILs significantly delayed tumor growth compared with DPBS-treated controls, as illustrated by median tumor growth (p = 0.0002 for P16 and p < 0.0001 for P20) in contrast to nDPOS TILs (p > 0.05) (Fig. 6B), despite variability in tumor engraftment, especially for P16 patient. Individual curves for both models are illustrated by Fig. 6C. Using a predefined threshold corresponding to three times the tumor volume at day 5 for each model, mice were classified as having controlled or non-controlled tumor growth. In contrast to mice receiving either DPBS or nDPOS TILs, more than half of mice treated with DPOS TILs controlled tumor growth, in both models (P16: 4/7; P20: 5/8). These results demonstrate that selective enrichment of PD-1⁺TIGIT⁺ TILs can translate into enhanced tumor control in vivo. Collectively, these results demonstrate that PD-1⁺TIGIT⁺ CD8⁺ TILs define a functionally distinct T-cell subset within human melanoma tumors, characterized by an activated pre-exhausted phenotype, clonal enrichment, broad tumor antigen specificity and superior cytotoxic and proliferative capacities. Importantly, selective enrichment of this subset translates into enhanced antitumor activity in two patient-derived xenograft models, providing an integrated functional validation of this rational TIL selection strategy. Discussion TIL therapy has demonstrated durable clinical benefit in patients with advanced melanoma, leading to the recent regulatory approval of Lifileucel for immune checkpoint blockade–refractory disease [6, 7, 24, 25]. However, current TIL products are heterogeneous and contain a substantial proportion of bystander, non-tumor-reactive lymphocytes or exhausted phenotype [26]. Retrospective analyses consistently indicate that clinical benefit correlates with the infusion of tumor-reactive T cells [12, 27–29], highlighting the need for prospective strategies to enrich therapeutic products in functionally relevant lymphocytes. Because high tumor mutational burden and neoantigen load are associated with response to TIL therapy [30], several strategies have attempted to enrich TIL products in neoantigen-specific T cells through ex vivo stimulation with predicted mutated epitopes [31–34]. Yet, their clinical applicability remains limited by the low proportion of naturally processed and presented neoepitopes derived from mutations in melanoma [35]. These limitations have prompted the development of antigen-agnostic selection strategies based on phenotypic markers associated with tumor reactivity, including PD-1, CD39, 4-1BB, CD103 or TIGIT [23, 36–39] (and ongoing clinical trial NCT05902520). In this context, we previously reported that circulating CD8⁺ T cells co-expressing PD-1 and TIGIT predicted clinical response to anti–PD-1 therapy in melanoma and Merkel cell carcinoma [15]. These cells were enriched in emerging clonotypes, tumor antigen-specific T cells and displayed features of a pre-exhausted differentiation state. Building on this previous work, we now demonstrate that this phenotype defines a tumor-reactive T-cell state within melanoma lesions that can be prospectively selected to optimize TIL-based ACT. Deep phenotyping performed at the end of the expansion procedure revealed that DPOS TILs combine markers of activation (4-1BB, CD27, CD28, CD226), tissue residency (CD103) and pre-exhaustion (TOX, TCF-1), consistent with a poised effector population capable of sustained tumor control rather than terminal exhaustion. Amplified DPOS TILs were enriched in CD8 T cells co-expressing CD39 and CD103, which were associated with tumor reactivity [23], and expressed additional markers linked to enhanced tumor reactivity or clinical efficacy, including 4-1BB [39], CD27, CD28 and CD39 [24, 40, 41]. Together, these features support the concept that DPOS TILs display several phenotypic characteristics associated with therapeutically relevant TIL products. Importantly, despite the differentiation pressure associated with prolonged ex vivo culture, DPOS TILs retained enrichment in activation- and tumor-reactivity-associated markers, while preserving proliferative capacity and therapeutically relevant expansion yields. These observations suggest that the PD-1⁺TIGIT⁺ signature is not restricted to a transient ex vivo activation state but identifies a relatively stable tumor-reactive compartment maintained throughout clinically relevant expansion conditions. Our TCR repertoire analysis confirms that PD-1+TIGIT+ TILs represent a clonally expanded, antigen-experienced compartment. The marked reduction in TCR diversity and the dominant contribution of top 20 clonotypes in the DPOS subset indicate strong selective pressures consistent with antigen-driven expansion within the tumor microenvironment. In contrast, the largely polyclonal nDPOS repertoire suggests enrichment in bystander or weakly stimulated T cells, confirmed by the detection of naive T cells in this subset. The preferential amplification of dominant clonotypes within PD-1⁺TIGIT⁺ TILs, together with their absence or very low representation in the nDPOS fraction, indicates that clonal expansion is tightly related to this differentiation state. Importantly, the very limited overlap between dominant DPOS clonotypes and the nDPOS compartment further supports the notion that clonal amplification preferentially occurs within the PD-1⁺TIGIT⁺ subset. Inter-patient heterogeneity in clonotype sharing likely reflects differences in tumor antigenic landscapes, local microenvironmental cues, and dynamic transitions between differentiation states. These data establish that PD-1+TIGIT+ TILs represent a selectively expanded, antigen-experienced pool, providing a mechanistic basis for their enhanced tumor reactivity in downstream assays. Indeed, by assessing reactivity against 49 peptides derived from 24 shared tumor antigens, we found that peptide-specific responses were overwhelmingly enriched within the PD-1⁺TIGIT⁺ subset, with the exception of Melan-AEAA-specific responses, which were also detected in only two nDPOS fractions. Interestingly, some T-cell responses directed against unconventional antigens, either from lncRNAs [42] or alternative antigen presentation pathways [43], were preferentially detected within the nDPOS compartment. One possible explanation is that these antigens may be expressed at lower levels or presented under more restricted conditions within tumor cells, resulting in weaker or intermittent TCR stimulation in vivo. Such activation conditions could limit the acquisition of the PD-1⁺TIGIT⁺ phenotype associated with sustained tumor recognition and effector differentiation. Together with repertoire analyses, these observations further suggest that PD-1/TIGIT coexpression may preferentially identify CD8⁺ T cells exposed to strong and persistent antigenic stimulation within the tumor microenvironment, leading to clonal expansion. Importantly, the peptide panel interrogates only a fraction of the antigenic repertoire potentially expressed by individual tumors and does not encompass patient-specific mutated neoepitopes. Thus, peptide-based assays provide a partial and indirect measure of tumor recognition. Functional evaluation using autologous tumor cells offers an integrative and physiologically relevant readout of antitumor activity, capturing responses to the full spectrum of naturally processed and presented antigens. Using enriched tumor cell suspensions and derived tumor cell lines, we show that effective tumor recognition and cytotoxicity were largely restricted to the PD-1⁺TIGIT⁺ TILs, despite occasional peptide-specific responses in nDPOS TILs. Importantly, DPOS TILs also retain proliferative capacity following antigen recognition, consistent with TCF-1 and TOX expression, even in the presence of PD-1/PD-L1 and TIGIT/CD155 inhibitory axes. Finally, in two patient-derived xenograft models, selective enrichment of PD-1⁺TIGIT⁺ TILs led to superior tumor growth control compared with nDPOS TILs, providing in vivo proof of principle for the translational potential of this approach. Although these experiments were performed in immunodeficient mice and therefore do not fully recapitulate the complexity of the human tumor microenvironment, the observed tumor control nevertheless supports the intrinsic antitumor potential of DPOS TIL products. These results highlight the translational relevance of our ex vivo observations while underscoring the need for further studies to robustly validate the in vivo therapeutic potential of DPOS TILs. Several studies have already proposed phenotypic markers to enrich tumor-reactive CD8⁺ TILs, including CD39, CD103, and 4-1BB, some of which are currently being evaluated in biomarker-guided TIL manufacturing strategies. In our study, PD-1⁺TIGIT⁺ CD8⁺ TILs were enriched in cells expressing these previously described tumor-reactive and tissue-residency-associated markers, including CD39⁺CD103⁺ and CD137⁺ subsets, supporting the concept that the DPOS compartment identifies a biologically relevant tumor-reactive state sharing multiple features associated with effective TIL products. Importantly, despite partial phenotypic overlap between DPOS and nDPOS compartments, functional tumor reactivity was predominantly detected within the DPOS fraction in our assays. Beyond tumor recognition itself, DPOS TIL products also fulfilled several characteristics currently associated with clinically effective TIL products, including enrichment in CD8⁺ T cells, preservation of a diverse yet clonally enriched TCR repertoire, maintenance of proliferative capacity after tumor recognition, and expansion yields compatible with therapeutic applications. Indeed, despite a slightly lower expansion rate of DPOS TILs compared to nDPOS, the absolute cell numbers of DPOS TILs obtained at the end of the expansion process (median > 1 billion of cells) if fully compatible with their use as a cell therapy product. Together, these observations suggest that PD-1/TIGIT coexpression may represent a practical phenotypic convergence signature enabling prospective enrichment of functionally tumor-reactive CD8⁺ T cells while preserving key attributes required for clinically relevant TIL manufacturing. Supplementary Table S4 provides an overview of the main phenotypic, functional, and manufacturing-related features associated with effective TIL products and highlights how the DPOS subset aligns with several previously described tumor-reactive TIL signatures and characteristics linked to clinical efficacy. Recent work also demonstrated that heterotypic CD8⁺ T-cell clusters physically interacting with tumor cells or antigen-presenting cells are highly enriched in tumor-reactive lymphocytes (overexpressing PDCD1 and additional exhaustion markers) and can be therapeutically exploited [44]. Our findings complement and extend this concept by identifying a simple PD-1⁺TIGIT⁺ phenotypic signature that enables prospective enrichment of clonally expanded tumor-reactive CD8⁺ T cells using a clinically compatible flow cytometry workflow. In this context, the PD-1⁺TIGIT⁺ selection strategy, based on three highly expressed markers, can be adapted to a closed, GMP-compliant cell sorting system (such as Tyto cell sorter, Miltenyi Biotec), to ensure compatibility with clinical manufacturing requirements. In conclusion, by bypassing the technical constraints associated with cluster preservation, this strategy offers a practical and scalable approach to optimize TIL-based immunotherapy for melanoma patients. Beyond prospective enrichment of tumor-reactive CD8⁺ T cells, DPOS-based selection strategies could also provide a relevant platform for future combinatorial approaches integrating immune checkpoint blockade, including anti-PD-1 or anti-TIGIT therapies, as well as next-generation genetic engineering strategies aimed at further enhancing TIL persistence and antitumor activity. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We thank the core facilities from UMS BioCore: Cytometry and cell sorting “Cytocell” and the animal facility “UTE” for expert technical assistance. We also thank the Unit of cell and Gene Therapy from Nantes hospital, for providing patient samples. Abbreviations Adoptive cell transfer Double positive Immune checkpoint inhibitors Monoclonal antibody Metastatic lymph node Maximum release Non-double positive Programmed cell death protein 1 Spontaneous release T-cell receptor T-cell immunoreceptor with Ig and ITIM domains Tumor infiltrating lymphocytes Author contributions Conceptualization: KD and NL. Supervision: KD and NL. Investigation: KD, TB, JG, SL, AM, AD and ST. Validation: KD and NL. Project administration: NL. Resources: MT, GQ and AK. Methodology: KD and SL. Software: MA, JD and SR. Data curation: MA and JD. Writing original draft: KD and NL. Writing, review and editing: KD, TB, JG, MA, AK and NL. Funding acquisition: KD, AK and NL. Funding This work was supported by the “Ligue nationale contre le cancer”: Team certification EL2022.LNCC/NaL, and the “Innovative cell therapy – 2024” program. This work was carried out within the framework of Labex ImmuNE, a program supported by NExT (ANR-16-IDEX-0007) and the French government as part of the « France 2030 » investment plan managed by the National Research Agency (ANR). Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate All information for human data and mice experiments are available in the Methods section. Consent for publication Not applicable. Competing interests The authors declare that they have no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Kathleen Ducoin, Email: Kathleen.ducoin@univ-nantes.fr. Nathalie Labarrière, Email: Nathalie.labarriere@univ-nantes.fr. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.