Esophageal cancer (EC) is the eighth most common malignancy worldwide and ranks sixth in cancer-related mortality (1). China accounts for over half of global ESCC cases and deaths annually, underscoring its high burden (2). Despite advancements in traditional treatments such as surgery, radiotherapy, and chemotherapy, the long-term outcomes for ESCC remain suboptimal, with a 5-year overall survival rate of only 20%-30% (3). The emergence of immunotherapy has brought transformative progress to the management of EC, offering new hope for improved outcomes.

The programmed cell death protein 1 (PD-1) receptor, expressed on activated T cells, binds to its ligand programmed cell death ligand 1 (PD-L1), leading to the suppression of T cell proliferation, cytokine secretion, and antitumor immune responses (4). Anti-PD-1 monoclonal antibodies restore T cell-mediated antitumor immunity by inhibiting immune checkpoint pathways. However, due to tumor heterogeneity, PD-L1 expression alone is insufficient as a predictive biomarker for immunotherapy efficacy. For example, the ESCORT-1st study demonstrated that even patients with TPS<1% could benefit from anti-PD-1 therapy combined with chemotherapy (5). These observations underscore the need for in-depth analysis of immune cell subpopulations within the TME to better understand their roles in treatment outcomes.

The TME is a complex ecosystem composed of tumor cells, fibroblasts, immune cells, extracellular matrix, and soluble factors, all of which play critical roles in tumor progression, immune evasion, drug resistance, and metastasis. Immune cells within the TME play dual roles, either suppressing or promoting tumor progression. Among these, CD8⁺ cytotoxic T lymphocytes (CTLs) are critical effectors of antitumor immunity. These cells recognize tumor antigens presented by MHC class I molecules and secrete cytotoxic mediators, such as perforin and granzyme, to eliminate tumor cells (6). However, tumor-infiltrating T lymphocytes (TILs) often display an exhausted phenotype caused by the upregulation of inhibitory receptors such as PD-1 and TIM3 (7). Studies have shown that PD1⁺TIM3⁺ T cells demonstrate impaired proliferation and reduced interferon-γ secretion, but their functionality can be partially restored through the blockade of them (8).

Regulatory T cells (Tregs) also contribute to immune evasion by suppressing effector T cell activity (9). These cells express FOXP3, a transcription factor that regulates their development and immunosuppressive function (10). Research has shown that targeting TIM3 can enhance anti-tumor immune response by reducing Tregs in head and neck cancer (11). Understanding the roles of these molecules is essential for characterizing the immune landscape of ESCC and identifying potential therapeutic targets.

Traditional bulk sequencing methods analyze the average genomic or transcriptomic data from mixed cell populations, obscuring intercellular heterogeneity (12). In contrast, single-cell RNA sequencing (scRNA-seq) enables high-resolution profiling of individual cells, providing detailed insights into immune cell subpopulations and their functional states (13). In this study, scRNA-seq revealed the heterogeneity of the ESCC immune microenvironment and identified distinct immune cell subpopulations with variations in distribution between tumor and adjacent tissues. These findings were further validated through flow cytometry, which highlighted the immune cell subsets most closely associated with the immunotherapy efficacy.

In summary, dynamic monitoring of peripheral blood immune cells using advanced profiling techniques has the potential to enable early prediction of immunotherapy efficacy, providing insights to advance personalized treatment strategies for ESCC.

Variability in the response to anti-PD-1 therapy among ESCC patients is closely linked to immune cell heterogeneity within the TME (20). While single-cell sequencing has revealed cellular diversity and trajectories in the TME (21), the functional roles of targets such as TIM3 remain underexplored. This study examines immune marker expression in tumor-infiltrating and adjacent cells, providing new insights into ESCC immunobiology.

Consistent with an active anti-tumor immune response, we observed robust infiltration of CD8⁺ T cells in ESCC tumor tissues. However, these tumor-infiltrating CD8⁺ T cells exhibited high levels of exhaustion markers—PD-1, CD39, and TIM3, compared to adjacent tissues. Chronic antigen stimulation and suppressive cues within the TME can drive TILs toward a dysfunctional state (22, 23), characterized by sustained upregulation of inhibitory receptors such as PD-1, TIM-3, and CD39 (24). Exhausted CD8⁺ T cells produce less IL-2, IFN-γ, and TNF-α and display impaired proliferative capacity (25). This exhaustion diminishes antitumor activity and reflects localized immune dysfunction (26). In a diffuse large B-cell lymphoma model, blockade of PD-1 or TIM-3 restored cytokine production and proliferation of these exhausted CD8⁺ T cells (8). Thus, the prevalence of exhausted TILs in ESCC provides a mechanistic explanation for limited PD-1 monotherapy efficacy and suggests that reversing T cell exhaustion is key to improving outcomes.

ESCC exhibited a marked enrichment of CD8⁺ tissue-resident memory T cells (TRM), characterized by the expression of CD103, a defining integrin that mediates epithelial retention. CD103 binds to E-cadherin expressed on carcinoma cells, thereby promoting the stable localization of TRM within the TME (27). TRM cells are stationed at tumor sites for immediate effector function upon antigen re-encounter and have been associated with improved anti-tumor immunity and response to checkpoint therapy in multiple cancers (28). The abundance of CD103⁺ TRM in ESCC suggests that local immunity remains partially preserved, potentially enabling initial tumor recognition. However, this local residency was accompanied by a marked reduction in CD44⁺CD62L⁺ central memory T cells (Tcm). Tcm reside in secondary lymphoid organs, mediate long-term immune memory, and mount proliferative recall responses upon antigen re-stimulation (29). The diminished Tcm population in ESCC suggests that the T cells in tumors are skewed toward either short-lived effector or terminally exhausted states. This diminished Tcm population in tumors may impair immune system “reservoirs” that normally support durable responses and rapid recall upon tumor antigen recurrence (30).

Adding further complexity, co-stimulatory receptors CD137 (4-1BB) and CD134 (OX40) were upregulated on tumor-infiltrating CD8⁺ and CD4⁺ T cells, respectively. CD137 enhances T cell proliferation and cytokine production upon ligand engagement (31), and agonists targeting 4-1BB are being investigated to boost anti-tumor immunity (32). Notably, OX40 signaling has been shown to destabilize FOXP3⁺ Tregs and attenuate their suppressive capacity (33). Despite the upregulation of OX40, the TME in ESCC remained enriched with FOXP3⁺ Tregs, suggesting persistent immunosuppression (34). The concurrent presence of activated effector T cells and expanded Tregs in ESCC reflects a competitive immunological landscape.

TIM3, encoded by the HAVCR2 gene, is a transmembrane protein initially identified on Th1 and Tc1 cells but also expressed on DCs (35, 36). Binding of Galectin-9 to TIM3 induces the release of BAT3 from the intracellular tail of TIM3, leading to T cell apoptosis (37). TIM3 upregulation on T cells following PD-1 blockade has been observed in various cancers, correlated with tumor recurrence in preclinical models (38). Consistently, our findings show that patients with shorter PFS exhibited greater expansion of TIM3⁺CD8⁺ T cells after immunotherapy, reinforcing the role of TIM3 in immune resistance.

Taking these findings together, we propose a model of immunotherapy resistance in ESCC centered on TIM-3. In the tumor, an abundance of TIM-3⁺ exhausted CD8⁺ TILs create an immune milieu prone to tumor immune escape. Upon PD-1 blockade, patients with a highly suppressive TME mount only transient T-cell reinvigoration, after which TIM-3–mediated pathways blunt the response, leading to adaptive resistance. These patients show peripheral immune changes: increasing frequencies of circulating TIM-3⁺ CD8⁺ T cells that reflect ongoing T-cell dysfunction. Clinically, this translates into poor outcomes. This model integrates local and systemic immune signatures, revealing how their convergence dictates response to PD-1 blockade. It is consistent with prior observations in gastrointestinal cancers that “immune-hot” tumors with reinvigorated T cells respond, whereas “immune-cold/exhausted” tumors evade therapy through alternate checkpoints and suppressive cells (39).

Our results underscore that TIM-3 is not only a biomarker of T-cell exhaustion but also a potentially druggable target for overcoming resistance in ESCC. TIM-3’s role in dampening immune responses has been documented across multiple tumors, and it appears especially pertinent in the context of PD-1 blockade failure (40, 41). Encouragingly, therapeutic targeting of TIM-3 is already underway: sabatolimab, an anti–TIM-3 monoclonal antibody, received FDA Fast Track designation for advanced solid tumors (42), and early-phase trials combining TIM-3 and PD-1 blockade have shown acceptable safety and preliminary efficacy (43). Our finding that TIM-3⁺ cells associate with poorer outcomes provides a strong rationale to test such combination strategies in ESCC. We speculate that dual-checkpoint inhibition (anti–PD-1 plus anti–TIM-3) could reinvigorate exhausted TILs more completely than PD-1 blockade alone, converting partial responders into durable responders.

Despite the insights gained, several limitations should be acknowledged. First, the small, single-center cohort (N = 20) limits statistical power. Second, as an observational study, residual confounding is unavoidable. Key clinical variables (e.g. PD-L1 status) were not fully controlled. Third, although we integrated public tumor scRNA-seq data with prospective blood profiling, the absence of an external validation cohort limits the robustness of the proposed prognostic threshold, which should be considered exploratory. Finally, no functional assays were performed to establish causality. The causal role of TIM-3⁺ T cells in immune resistance remains uncertain, and mechanistic studies are required to determine whether they are active mediators or passive markers.

In summary, our study identifies TIM-3⁺ CD8⁺ T cells as key immunological features linking the TME to systemic immune changes and clinical outcomes. Patients whose tumors foster a TIM-3–high, exhausted immune contexture are more likely to exhibit peripheral T-cell dysfunction during therapy and to experience disease progression, underscoring TIM-3’s role in adaptive resistance to PD-1 inhibitors. From a translational perspective, dynamic monitoring of TIM-3⁺ T cells in blood could serve as an non-invasive indicator of immunotherapy efficacy in ESCC, aiding in treatment decisions. More importantly, our work lays a biological foundation for therapeutically targeting TIM-3 in ESCC and supports its integration into combinatorial immunotherapy to overcome resistance and improve clinical outcomes.