HCAR2 Orchestrates an Immunosuppressive Niche and Determines Checkpoint Inhibitor Responsiveness in Esophageal Squamous Cell Carcinoma Authors: Zekun Li, Guangcong Shen, Xiaoqing Ma, Chenyang Meng, Diliyaer Abudukeremu, Rui Zhao, Zuoyu Chen, Qiang Song, Xiaofan Guo, Tianxing Zhou, Chao Xu, Bo Ni, Yueying Shan, Boyang Fu, Yiheng Chen, Zhansheng Jiang, Yongjie Xie, Kaiyuan Wang, Liangliang Wu, Xiuchao Wang Published: 2025 ABSTRACT AbstractBackgroundEsophageal squamous cell carcinoma (ESCC) is a highly aggressive malignancy with poor prognosis, where immune checkpoint inhibitors (ICIs), such as anti-PD-1/PD-L1 antibodies, show limited efficacy with heterogeneous patient responses. The tumor microenvironment (TME) and tumor subclonal heterogeneity significantly influence immune evasion and therapeutic resistance. This study aimed to identify distinct tumor subpopulations in ESCC, characterize their roles in immunosuppression, and determine their association with ICI responsiveness.MethodsWe analyzed single-cell multi-omics data from tumor specimens and peripheral blood mononuclear cells (PBMCs) of 12 ESCC patients receiving neoadjuvant therapy. Using bioinformatics tools (Seurat, CellChat), we identified tumor subpopulations, inferred cell-cell communication, and validated functional mechanisms via in vitro assays (cell culture, flow cytometry, western blotting), patient-derived organoid (PDO) co-culture systems, and humanized mouse models. Clinical correlations were assessed using bulk transcriptome data, survival analysis, and multi-cohort validation (GEO datasets, IMvigor210 cohort).ResultsA distinct HCAR2+ epithelial subclone was identified, characterized by high PD-L1 expression, immunosuppressive cell interactions (e.g., PD-L1/PD-1 axis with exhausted NK-like T cells), and enrichment in pathways driving immune evasion (STAT3, IL-6, PI3K-AKT-mTOR). HCAR2+ cells upregulated PD-L1 via STAT3 activation, impairing CD8+ T cell cytotoxicity and antigen presentation. Clinically, HCAR2+ infiltration correlated with poor prognosis, increased copy number variations (CNVs), high tumor mutational burden (TMB), and enhanced sensitivity to neoadjuvant anti-PD-1 therapy in vivo. Multi-cohort analysis confirmed HCAR2 expression as a predictive biomarker for favorable ICI response, associated with reduced immune exclusion and TIDE scores.ConclusionsHCAR2+ tumor subclones orchestrate an immunosuppressive TME via PD-L1-mediated CD8+ T cell exhaustion while paradoxically conferring sensitivity to PD-L1 blockade. HCAR2 expression serves as a dual biomarker for poor prognosis and ICI responsiveness, offering a novel molecular target for stratifying ESCC patients who may benefit from neoadjuvant immune checkpoint therapy. Targeting the HCAR2-STAT3-PD-L1 axis could enhance immunotherapy efficacy in this aggressive cancer.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12575-025-00305-1.Keywords: Esophageal squamous cell carcinoma, HCAR2, Immune microenvironment, PD-L1, Immunotherapy Introduction Esophageal squamous cell carcinoma (ESCC) is one of the most lethal malignancies worldwide, with approximately 70% of patients diagnosed at locally advanced or metastatic stages [1, 2]. In recent years, immune checkpoint blockade (ICB) therapy, particularly PD-1/PD-L1 inhibitors, has shown significant efficacy in various solid tumors. However, the response rate in ESCC remains below 20% [3, 4]. The China Multicenter Study (ESCORT-1st) demonstrated that camrelizumab combined with chemotherapy reduced the risk of death by 30%, but no significant survival benefit difference was observed between PD-L1-positive and PD-L1-negative patients [5]. This suggests that the predictive value of PD-L1 as a biomarker in ESCC is limited, highlighting the need to elucidate the tumor microenvironment (TME) in ESCC, especially tumor cell-immune cell interactions, to identify patients more likely to respond to anti-PD-1/PD-L1 therapy. Aberrant interactions between tumor cells and immune cells are key drivers of immune evasion. In ESCC, tumor cells evade immune surveillance through various mechanisms, overexpression of the immune checkpoint molecule PD-L1 being a critical contributor [6]. The binding of PD-L1 to PD-1 on T cells inhibits their proliferation, activation, and cytotoxic function, thereby promoting tumor immune evasion [7]. As a a central mediator of tumor cell-mediated T-cell suppression, PD-L1 expression is regulated by epigenetic modifications, transcription factors, and inflammatory signals. Studies have confirmed that aberrant PD-L1 methylation in ESCC tumor cells leads to dysregulated expression and contributes to ICB resistance [8, 9]. The complexity of PD-L1 regulatory networks underscores the limitations of single-target therapeutic strategies, necessitating the identification of upstream master regulators. Hydroxycarboxylic acid receptor 2 (HCAR2), a G protein-coupled receptor primarily regulates metabolic reprogramming and inflammatory responses [10]. Recent studies have revealed abnormal HCAR2 expression in multiple cancers, with emerging evidence linking it to tumor invasion, metastasis, and immune evasion in breast, colorectal, and hepatocellular carcinomas [11–13]. However, the functional role of HCAR2 in ESCC and its interplay with the immune microenvironment remain unexplored. Our study demonstrates that HCAR2+ epithelial cells drive CD8+ T cells exhaustion through activation of the STAT3/PD-L1 axis, thereby establishing an immunosuppressive microenvironment that facilitates malignant progression. HCAR2 may serve as a predictive biomarker for immunotherapy response, offering novel molecular subtyping criteria and combination therapy strategies to enhance the identification of patients responsive to anti-PD-1/PD-L1 therapy. Materials and Methods Single-cell RNA Sequencing in Author’s Center The single-cell transcriptomic profiles utilized for intercellular communication mapping were retrieved from the ESCC repository within our institutional Genome Sequence Archive (GSA). This database comprises paired tumor specimens and PBMCs from 12 ESCC cases who underwent neoadjuvant therapy, systematically capturing the immunocellular architecture of neoplastic microenvironments and systemic circulation. Raw sequencing files from 12 ESCC cases collected at Tianjin Medical University Cancer Center have been archived in GSA-Human under accession HRA003312. Differential gene expression profiling was conducted using the Seurat toolkit, enabling identification of biologically relevant transcriptional trajectories and cellular subpopulations. Non-linear manifold learning via Uniform Manifold Approximation and Projection (UMAP) facilitated low-dimensional representation of high-parametric scRNA-seq data, permitting spatial visualization of phenotypic heterogeneity. For decoding cell-cell signaling networks, the CellChat framework was implemented to model ligand-receptor interactions, pathway activation patterns, and transcriptional regulatory circuits across discrete cellular states. Preprocessing pipelines incorporated quality-controlled read alignment (Cell Ranger), batch effect mitigation (Harmony algorithm), and multimodal data integration. Parameter optimization for Harmony-based integration focused on preserving biological variance while neutralizing technical artifacts, complemented by canonical correlation analysis for cross-modal harmonization. Statistical inference adhered to established workflows for scRNA-seq analytics, with differential expression thresholds set at |log2FC| >0.25 and adjusted p < 0.05. This analytical continuum encompassed: (1) Dimensionality reduction-driven cluster annotation, (2) Pseudotemporal reconstruction of differentiation cascades, (3) Quantitative assessment of intercellular crosstalk networks. Data Analysis Public datasets used in this study were obtained from TCGA (The Cancer Genome Atlas) dataset, which was downloaded from the cBioPortal database. Copy number variations (CNVs) were analyzed using GISTIC2 (Genomic Identification of Significant Targets in Cancer 2). The maftools package was employed to quantify mutation frequencies of target genes and generate visualizations of mutation profiles. Three complementary enrichment analyses were performed to investigate the enrichment patterns of differentially expressed genes (DEGs) across biological pathways: Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, Gene Ontology (GO) enrichment, and Gene Set Enrichment Analysis (GSEA). Immune-related scores and tumor microenvironment (TME) scores for each sample were calculated using the TIMER (Tumor Immune Estimation Resource) and ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumors using Expression data) algorithms. Pathway gene sets for GSEA, applicable to both single-cell and bulk sequencing datasets, were retrieved from the Hallmark Gene Set Collection in the Molecular Signatures Database (MSigDB). Data analyses were performed using R software version 4.4.3. Cell Culture and Transfection The Human ESCC cell lines (KYSE450 and KYSE150) were obtained from the American Type Culture Collection (Manassas, VA). Mycoplasma contamination was excluded in all cell lines. The cells were cultured in Dulbecco’s Modified Eagle Medium (C11995500BT, Gibco/Life Technologies) or RPMI-1640 (C11875500BT; Gibco) basic medium with 10% fetal bovine serum (10099141 C, Gibco) at 37 °C in a humidified atmosphere of 5% CO2. The KYSE450 cell line carries the amplified oncogenes c-myc (2×) and c-erb B (2×), while the KYSE150 cell line carries the amplified oncogenes c-erb B (8×) and cyclin D1 (4×). Both cell lines are capable of inducing tumor formation in humanized mice. Human HCAR2 cDNA (NM_177551.4) was cloned into a pLV expression vector (pLV-HCAR2). The empty pLV vector was used as a control. Lentiviral transductions were performed according to standard procedures. For the stable knockdown cell lines, shRNA sequences were designed using the Sigma-Aldrich shRNA designer (Supplementary Table 1). Three recommended sequences targeting the gene of interest were synthesized and cloned into the pLV-RNAi-Puro vector (Biosettia). Among the three shRNA sequences, the most effective one was selected for subsequent experiments. Flow Cytometry The harvested tumors from humanized mice were processed into single-cell suspensions using collagenase (1 mg/mL), hyaluronidase (2.5 U/mL), and DNase (0.1 mg/mL). The infiltration of CD8+ T cells (CD3+ and CD8+) in tumors, phenotype of CD8+ T cells (IFN-γ+, PD-1+, and TIM-3+), and apoptosis of CD8+ T cells (Annexin V-PE) were analyzed with the primary antibodies in Supplementary Table 2. For intracellular IFN-γ staining, the cells were stimulated with PMA (5 ng/mL) and ionomycin (500 ng/mL; Sigma-Aldrich) in the presence of GolgiStop for 4 h, followed by surface and intracellular staining. The data were analyzed using CytoExpert version 2.6 and FlowJo version 10.0. Western Blotting The cells were washed twice with ice-cold PBS and lysed in RIPA buffer with a proteinase inhibitor and a phosphatase inhibitor cocktail to extract total protein. A total of 20 mg of protein sample was separated on 10% polyacrylamide sodium dodecyl sulfate gels and transferred to polyvinylidene difluoride membranes. After the blocking procedure, membranes were incubated with primary antibodies, and visualized in Imager (Bio-Rad, Hercules) using the ECL system (Thermo Fisher Scientific). The target protein was detected by western blotting with the primary antibodies in Supplementary Table 3. Establishment and Culture of Patient-Derived Organoids (PDOs) Fresh ESCC specimens were washed three times with cold PBS containing 10% penicillin and streptomycin. The specimens were cut into small pieces (< 1 mm3) with sterile blades, and then digested with a mixture of 5 mg/mL collagenase Ⅱ (C2-22-1G; Sigma-Aldrich) and 5 mg/mL collagenase Ⅳ (C4-22-1G; Sigma-Aldrich) for 0.5 h at 37℃ under severe vibration. Advanced DMEM/F12 (12634010, Gibco) containing 0.5% BSA was added to terminate digestion. Tumor cells were harvested, and centrifuged at 400 g for 5 min. Subsequently, tumor cells were embedded in Matrigel (356231, Corning Life Sciences, Corning, NY, USA) and grown in ESCC Organoid Complete Feeding Medium. Depending on growth, the organoid medium was changed approximately every 3 days, and organoids were passaged approximately every 14 to 21 days. All the patients provided written consent for the use of their specimens and disease information for future investigations according to the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital, China, and in accordance with the recognized ethical guidelines of Helsinki (Approved No.: EK20240211). In Vivo CD8+ T Cells Depletion For CD8+ T cells depletion, administer 500 µg of purified anti-human CD8α (clone OKT-8, BioXcell) diluted in 200 µL sterile PBS via intraperitoneal injection per mouse, followed by maintenance doses of 250 µg every 3–4 days to sustain depletion. Control groups receive equivalent volumes of isotype IgG (clone LTF-2, BioXcell). Validate depletion efficiency 72 h post-initial injection by collecting peripheral blood and analyzing CD3+ CD8+ cells populations via flow cytometry. After successful depletion (CD8+ T cells < 1% among CD3+ lymphocytes), subsequent subcutaneous PDOs inoculation was performed. Construction of CD34+ Humanized PDOX Model Human umbilical cord blood-derived CD34+ HSCs (STEMCELL, Cat. No.: 70008) were intravenously transplanted into NSG mice (SPF Biotechnology Co., Ltd. TianJin, China) that had undergone sub-lethal irradiation and bone marrow clearance. Eight to sixteen weeks later, PDOX models of human ESCC were transplanted into the above-mentioned humanized NSG mice. The Animal Ethical and Welfare Committee (AEWC) of Tianjin Medical University Cancer Institute and Hospital approved all animal experiments (Approved No.: 2024001). Statistical Analysis Statistical analyses were performed with SPSS version 18.0 and GraphPad Prism version 9.5.0. For data following a normal distribution, Student’s t test or ANOVA for unpaired data was used to compare mean values. For data with a non-normal distribution, the Wilcoxon rank-sum test was applied to compare differences between groups. Kaplan-Meier survival curves were made according to these cut points; the log-rank test was used to obtain a P-value for the significance of Kaplan-Meier curves’ divergence. *, **, *** and **** indicated p < 0.05, 0.01, 0.001 and 0.0001. ns, non-significant. Results Single-cell Sequencing of ESCC Reveals the Association Between High-copy Number Variation (CNVhigh) Malignant Epithelial Subpopulations and Immunotherapy Response The efficacy of neoadjuvant chemoradiotherapy in ESCC has been reported to correlate with cytotoxic immune cell activity in the tumor microenvironment [14]. To identify ESCC patients likely to respond to anti-PD-1/PD-L1 therapy, single-cell RNA sequencing was performed on on ESCC tissues, and data were analyzed using the Seurat pipeline. After quality control and dimensionality reduction, cell clusters were annotated (Fig. 1A), with marker gene expression profiles provided in Figure S1. Epithelial cells were further subdivided into seven subpopulations (Fig. 1B). Mutational signature profiles were systematically analyzed through multidimensional bioinformatic approaches, and a somatic mutational landscape was reconstructed using integrated visualization tools (Figure S2A). Malignancy of these subpopulations was assessed via copy number variation (CNV) analysis (Fig. 1C), with CNV scores quantified in Fig. 1D. Class1, Class2, and Class7 were classified as malignant epithelial subpopulations. Subclustering of high-CNV malignant epithelial cells identified seven distinct clusters (CNVhigh_Epi1-7) (Fig. 1E). Cell-cell communication analysis revealed that CNVhigh_Epi5 cells engaged with PDCD1+ TIGIT+ NK-like T cells via the CD274 (PD-L1)-PDCD1 (PD-1) ligand-receptor axis (Fig. 1F), exhibiting the highest interaction strength among all subpopulations (Fig. 1G). These findings prompted further investigation into the unique role of CNVhigh_Epi5 cells. In summary, CNVhigh_Epi5 tumor cells in ESCC may mediate immunotherapy resistance via the PD-L1/PD-1 signaling pathway, prompting subsequent exploration of molecular targets specific to this subpopulation. Single-cell analysis of ESCC reveals distinct epithelial subpopulations and their interactions with T cells. A UMAP plot showing the distribution of cell clusters. B UMAP plot showing the distribution of epithelial subpopulations. C Heatmap showing inferCNV analysis of epithelial subpopulations. D Violin plot showing CNV scores of epithelial subpopulations. E UMAP plot showing the distribution of malignant epithelial subpopulations. F Bubble plot showing ligand-receptor interactions between CNVhigh_Epi5 cells and T cell subpopulations. G Network plot showing the interaction intensity of the CD274-PDCD1 ligand-receptor pair between malignant epithelial cells and T cells HCAR2+ epithelial Cells Drive Immunosuppression Via the PD-L1/PD-1 Signaling Axis To identify potential drivers of elevated PD-L1 expression in CNVhigh_epi5 cells, differential gene expression analysis was performed across the seven malignant epithelial subpopulations. Tumor-derived PD-L1 expression in CNVhigh_Epi5 cells showed the strongest positive correlation with HCAR2 (Fig. 2A-B and Table 1.). Cell-cell communication analysis revealed that CNVhigh_Epi5 cells functioned as signaling senders, transmitting PD-L1 signals to PDCD1+ TIGIT+ NK-like T cells (Fig. 2C). Based on HCAR2 overexpression, we designated CNVhigh_Epi5 cells as HCAR2+ epithelial cells. Pathway enrichment analysis demonstrated significant upregulation of the PD-L1/PD-1, IL-6, and PI3K-AKT-mTOR pathways in HCAR2+ epithelial cells (Fig. 2D-E). UCell pathway activity scoring confirmed that these pathways were most highly enriched in HCAR2+ epithelial cells compared to other malignant subpopulations (Fig. 2F-H). Differential expression profiling of HCAR2+ epithelial cells (Fig. 2I) and cross-validation through GO, KEGG, and WikiPathway enrichment analyses (Fig. 2J) further supported their functional involvement in immune checkpoint regulation. Collectively, these findings indicate that HCAR2+ epithelial cells mediate immunosuppression by activating key immune checkpoint-related pathways. HCAR2+ epithelial cells exhibit unique gene expression and pathway enrichment profiles. A - B UMAP plots showing co-expression of HCAR2 and PD-L1 in malignant epithelial subpopulations. C Heatmap showing PD-L1 signaling communication between malignant epithelial cells and NK-like T cells. D Heatmap showing pathway enrichment in malignant epithelial subpopulations. E Bubble plot showing key pathway enrichment in HCAR2+ epithelial cells. F-H Raincloud plots showing UCell scores for the PD-L1/PD-1, IL-6, and PI3K-AKT-mTOR signaling pathways in HCAR2+ epithelial cells. I Volcano plot showing differentially-expressed genes in HCAR2+ epithelial cells. J Heatmap showing differential gene expression and pathway enrichment in malignant epithelial subpopulations Table 1.Genes Positively Correlated with PD-L1 Expression in the CNVhigh_epi5 ClusterGene SymbolGene IDPCCHCAR2ENSG00000182782 .20.64ITPRIPENSG00000148841.150.63PKP1ENSG00000081277.110.63TRIM29ENSG00000137699.160.62RARGENSG00000172819.160.62DENND2CENSG00000175984.140.62CERS4ENSG00000090661.110.61GJB5ENSG00000189280.30.61GNA15ENSG00000060558.30.6ZNF385AENSG00000161642.170.6EXOC6BENSG00000144036.140.6PRRG4ENSG00000135378.30.6IRF6ENSG00000117595.100.6SERTAD2ENSG00000179833.40.59PVRL4ENSG00000143217.80.59HAS3ENSG00000103044.100.59SGK1ENSG00000118515.110.59CERS3ENSG00000154227.130.59ZNF750ENSG00000141579.60.59PAK6ENSG00000137843.110.58GPR87ENSG00000138271.50.58N4BP1ENSG00000102921.70.57FAM160A1ENSG00000164142.150.57PTPN13ENSG00000163629.120.57DUOXA1ENSG00000140254.120.57NFKBIAENSG00000100906.100.57JMJD7ENSG00000243789.100.57TBC1D10AENSG00000099992.150.57PVRL1ENSG00000110400.100.57AC055736.1ENSG00000280827.10.57TMEM79ENSG00000163472.180.57GM2AENSG00000196743.80.57STARD5ENSG00000172345.130.56HSPB1ENSG00000106211.80.56TNFAIP8ENSG00000145779.70.56FAM83CENSG00000125998.70.56PGLYRP3ENSG00000159527.30.56ANXA8ENSG00000265190.60.55MAST4ENSG00000069020.180.55BNC1ENSG00000169594.120.55HSPB1P2ENSG00000230216.10.55SH3PXD2A-AS1ENSG00000280693.20.55TBC1D2ENSG00000095383.190.55RAB10ENSG00000084733.100.55SLC9A9ENSG00000181804.140.55BDKRB1ENSG00000100739.100.55MCCENSG00000171444.170.55SOWAHCENSG00000198142.40.55KLC3ENSG00000104892.160.55PERPENSG00000112378.110.54GRHL3ENSG00000158055.150.54LINC01010ENSG00000236700.50.54DUSP7ENSG00000164086.90.54DSG3ENSG00000134757.40.54BICD2ENSG00000185963.130.54SH3BP1ENSG00000100092.200.54BAIAP2ENSG00000175866.150.54NOD2ENSG00000167207.110.54RNF19BENSG00000116514.160.54SUSD4ENSG00000143502.140.54ARL4DENSG00000175906.40.54PLCH2ENSG00000149527.170.54ARHGAP23ENSG00000275832.40.54CD9ENSG00000010278.110.54WDR47ENSG00000085433.150.54FGFBP1ENSG00000137440.40.54IPPKENSG00000127080.90.53DGKAENSG00000065357.190.53CAPN1ENSG00000014216.150.53KAZNENSG00000189337.150.53HK1ENSG00000156515.210.53FOXN1ENSG00000109101.70.53FAM110AENSG00000125898.120.53LTB4R2ENSG00000213906.90.53CAPNS2ENSG00000256812.10.53ALDH3B2ENSG00000132746.140.53GJB6ENSG00000121742.150.53SLC12A6ENSG00000140199.110.53NIPAL4ENSG00000172548.140.53IFFO2ENSG00000169991.100.53YBX3ENSG00000060138.120.53FRMD6ENSG00000139926.150.53BTBD11ENSG00000151136.140.53MREGENSG00000118242.150.52TMEM154ENSG00000170006.110.52KRT5ENSG00000186081.110.52SEMA4AENSG00000196189.120.52RP11-116G8.5ENSG00000256462.10.52FEM1BENSG00000169018.50.52P2RY1ENSG00000169860.60.52S1PR5ENSG00000180739.130.52SDC1ENSG00000115884.100.52NCK1ENSG00000158092.60.52KRT6AENSG00000205420.100.51S100A11ENSG00000163191.50.51FGFR3ENSG00000068078.170.51SMAD3ENSG00000166949.150.51DLX3ENSG00000064195.70.51FAT2ENSG00000086570.120.51BICC1ENSG00000122870.20.51 Genes Positively Correlated with PD-L1 Expression in the CNVhigh_epi5 Cluster HCAR2+ Epithelial Cells Infiltration Level Influences the Genomic CNV Landscape Emerging evidence suggests that tumor-immune cell interactions within the microenvironment can shape genomic instability, a hallmark of cancer progression [15]. Given that HCAR2+ epithelial cells may contribute to immune suppression by upregulating immune checkpoint-related pathways, we hypothesized that HCAR2+ epithelial cells might further drive oncogenesis by altering the genomic landscape of tumors through structural variations. To test this, we analyzed mutations in ESCC samples stratified by high versus low infiltration of HCAR2+ epithelial cells. Oncoplots showed no significant differences in the mutation frequencies of the top 15 genes between the two groups (Figure S3A-B). However, CNV analysis revealed that samples with high HCAR2+ epithelial cells infiltration exhibited a higher burden of CNVs in genomic regions associated with oncogenic pathways (Figure S3C-D). These results suggest that HCAR2-mediated cellular interactions may promote genomic instability through CNV-driven mechanisms rather than through the accumulation of somatic mutations. Terminal Differentiation of HCAR2+ Epithelial Cells Correlates with Immunosuppressive Signatures Pseudotime trajectory analysis of malignant epithelial cells was performed to reconstruct differentiation dynamics, with arrows denoting developmental progression. The trajectory originated from CNVhigh_Epi3 and CNVhigh_Epi7 clusters and terminated in HCAR2+ epithelial cells, indicating their identity as the terminal differentiation state (Fig. 3A-C). Dynamic expression profiling of HCAR2 and CD274 (PD-L1) across the trajectory revealed coordinated upregulation, with both genes peaking in HCAR2+ epithelial cells at the terminal stage (Fig. 3D). This suggests that the HCAR2+ subpopulation may drive immunosuppressive microenvironment formation. Pseudotime trajectory analysis reveals the developmental dynamics of malignant epithelial cells. A UMAP plot showing the developmental trajectory of malignant epithelial cells. B - C UMAP plots showing the developmental paths of (B) Lineage1 and (C) Lineage2. D Scatter plot showing dynamic expression of HCAR2 and CD274 along differentiation trajectories. E Heatmap showing differential pathway enrichment along the differentiation trajectory of malignant epithelial cells Branch-specific differential enrichment analysis further demonstrated that HCAR2+ epithelial cells exhibited a significant negative correlation with immune activation markers. Throughout tumor evolution, these cells progressively lost immune cell cytotoxicity-related signals and antigen presentation pathways (Fig. 3E). Collectively, these findings establish HCAR2+ epithelial cells as terminally differentiated immunosuppressive effectors. High HCAR2 Expression in Epithelial Cells Predicts Poor Prognosis and is Associated with Immunosuppressive Pathways in ESCC To investigate the functional properties of HCAR2+ epithelial cells in ESCC, we compared HCAR2 expression levels between tumor tissues and matched adjacent normal tissues using TCGA bulk transcriptome data from the cBioPortal database. HCAR2 was significantly upregulated in ESCC tissues compared to adjacent normal epithelium (Fig. 4A). Single-cell deconvolution analysis of the same dataset stratified ESCC samples based on HCAR2+ epithelial cell infiltration levels. Kaplan-Meier survival analysis revealed that patients with high HCAR2+ epithelial cell infiltration exhibited significantly shorter overall survival (OS) and disease-specific survival (DSS) (Fig. 4B-D). Integrative analysis of differential gene expression and pathway enrichment (GSEA, KEGG, GO) revealed that high HCAR2+ epithelial cells infiltration correlated with activation of the PD-L1/PD-1 immune checkpoint signaling pathway (Fig. 4E-H). Collectively, our data further validate the conclusions obtained at the single-cell level. Infiltration of HCAR2+ epithelial cells correlates with poor prognosis and immune-related pathways. A Differential HCAR2 expression in ESCC and adjacent tissues from cBioPortal database. B-C Disease-specific survival and Overall survival of groups with high and low HCAR2+ epithelial cells infiltration from the cBioPortal database. D Overall survival prognosis stratified by high and low HCAR2 mRNA expression levels using data from the cBioPortal database. E Heatmap showing differential gene expression in groups with high and low HCAR2+ epithelial cells infiltration. F Bar plot showing GO pathway enrichment of differentially-upregulated genes. G GSEA plot showing key pathway enrichment scores of HCAR2+ epithelial cells-specific genes. H Bar plot showing KEGG pathway enrichment of differentially-upregulated genes HCAR2 Modulates PD-L1 Expression to Promote Immune Effector Cell Exhaustion Building on these observations, we hypothesized that HCAR2 promotes tumor CD274 (PD-L1) expression, thereby driving immune effector cell exhaustion. To investigate this mechanism, HCAR2-overexpresion (OE) and knockdown (KD) cell lines were generated in both KYSE450 and KYSE150. Western blot analysis revealed that PD-L1 expression was significantly upregulated in HCAR2-OE cells and downregulated in HCAR2-KD cells compared to their respective controls (Fig. 5A-B). Flow cytometry further validated HCAR2-dependent regulation of PD-L1 on the cell surface (Fig. 5C-D). To explore the underlying mechanism, RNA sequencing was performed on HCAR2-OE (n = 3) and Vector (n = 3) KYSE450 cells. KEGG enrichment analysis of differentially expressed genes (DEGs) highlighted significant activation of the JAK-STAT signaling pathway in HCAR2-OE cells (Fig. 5E). Consistent with prior studies implicating STAT3 phosphorylation (Tyr705) as a key mediator of PD-L1 expression downstream of JAK-STAT signaling [16], Western blot analysis demonstrated that HCAR2 overexpression induced STAT3 phosphorylation at Tyr705. Critically, treatment of both Vector and HCAR2-OE cells with the STAT3 inhibitor cryptotanshinone abolished HCAR2-mediated PD-L1 regulation (Fig. 5F-G). These findings establish a mechanistic link between HCAR2 and PD-L1 expression via STAT3 activation. HCAR2 regulates PD-L1 expression through the STAT3 pathway.A-B Western blot showing HCAR2 and PD-L1 expression in KYSE450 (A) and KYSE150 (B) cells after HCAR2 overexpression and knockdown. C-D Flow cytometry showing PD-L1 expression in KYSE450 (C) and KYSE150 (D) cells after HCAR2 overexpression and knockdown. E KEGG pathway enrichment of differentially-expressed genes. F Western blot showing PD-L1 expression in HCAR2-overexpressing KYSE450 cells treated with a STAT3 inhibitor. G Flow cytometry showing PD-L1 expression in HCAR2-overexpressing KYSE450 cells treated with a STAT3 inhibitor. Data in C-D, G were presented as mean ± SD and analyzed using two-tailed, unpaired Student’s t test. *, **, *** and **** indicated p < 0.05, 0.01, 0.001 and 0.0001 HCAR2+ epithelial Cells Drive an Immunosuppressive Tumor Microenvironment and CD8+ T Cell Exhaustion To explore HCAR2-driven remodeling of the tumor immune microenvironment, we identified genes highly expressed in HCAR2⁺ epithelial cells, which were then compiled into a gene set. Based on enrichment scores of the gene set, we stratified 82 ESCC samples from TCGA into high-HCAR2 (n = 41) and low-HCAR2 (n = 41) groups, and analyzed immune-related genes and chemokine expression profiles. Heatmap analysis demonstrated significantly elevated expression of PD-L1 and other immunosuppressive markers in high-HCAR2 group compared to low-HCAR2 group (Figure S4A). We employed the TIMER (Tumor Immune Estimation Resource) algorithm to analyze immune cell infiltration between the two groups. The results demonstrated that compared with the high-HCAR2 group, the low-HCAR2 group had increased infiltration of macrophages and neutrophils, whereas the infiltration of CD8⁺ T cells was significantly decreased (Figure S4B). Subsequently, we conducted an in-depth analysis of functional markers on CD8+ T cells. Compared with the low-HCAR2 group, the high-HCAR2 group exhibited a higher proportion of PD-1+ CD8+ T cells and TIM-3+ CD8+ T cells, accompanied by a lower proportion of CD8+ T cells producing TNF-α, IFN-γ, and Granzymes (Figures S4C-G). Taken together, these results demonstrate that the HCAR2+ epithelial cell subpopulation exerts immunosuppressive effects during ESCC progression and harbors molecular and microenvironmental features associated with immune checkpoint inhibitor responsiveness. Organoid and Humanized Mouse Models Demonstrate HCAR2-mediated CD8+ T Cell Suppression Via PD-L1 Upregulation To further validate the immunomodulatory role of HCAR2 in vivo, HCAR2-KD patient-derived organoids (PDOs) were generated from ESCC tissues. HCAR2-KD significantly reduced PD-L1 expression in PDOs compared to Scramble (Fig. 6A). A humanized mouse model was established by subcutaneously transplanting these PDOs into NSG mice with immune systems reconstituted through engraftment of human CD34+ hematopoietic stem cells (Fig. 6B). Longitudinal monitoring revealed markedly slower tumor growth in the HCAR2-KD group, as evidenced by growth curves (Fig. 6C) and terminal tumor weight measurements at day 35 (Fig. 6D). Flow cytometric analysis of tumor-infiltrating CD8+ T cells demonstrated that HCAR2-KD tumors exhibited increased CD8+ T cells infiltration, reduced apoptosis, decreased expression of exhaustion markers and enhanced effector function compared to Scramble (Figs. 6E-H). To confirm CD8+ T cells dependency, we administered anti-human CD8α depleting antibodies were administered intraperitoneally prior to PDOs transplantation. Complete depletion of CD8+ T cells abolished the tumor-suppressive effects of HCAR2-KD (Figs. 6I-J), directly linking HCAR2-driven immunosuppression to functional CD8+ T cells activity. Collectively, these findings establish that HCAR2+ epithelial cells promote immune evasion in ESCC via PD-L1-dependent suppression and exhaustion of CD8+ T cells. HCAR2 knockdown enhances CD8+ T cell infiltration and function in vivo. A Flow cytometry showing PD-L1 expression in HCAR2-KD and Scramble PDOs. B Schematic illustration of the humanized mouse model with subcutaneous PDOs transplantation. C Tumor growth curves of indicated groups are shown. D Tumor images and weights at day 35. E Flow cytometric analysis of tumor-infiltrating CD8+ T cells. F Flow cytometric analysis of tumor-infiltrating CD8+ T cells apoptosis. G Flow cytometric analysis of PD-1 and TIM-3 double-positive tumor-infiltrating CD8+ T cells. H Flow cytometric analysis of IFN-γ-positive tumor-infiltrating CD8+ T cells. I-J Humanized mouse model with subcutaneous PDOs transplantation after CD8+ T cells depletion.Tumor growth curves of indicated groups (I) and tumor weights at day 40 (J) are shown. Data in A, D-H, J were presented as mean ± SD and analyzed using two-tailed, unpaired Student’s t test. *, **, *** and **** indicated p < 0.05, 0.01, 0.001 and 0.0001. Data in C and I were measured by two-way ANOVA and post hoc analysis to test mouse tumor growth between groups Multi-cohort Analysis Identifies HCAR2+ Epithelial Cells as a Predictive Biomarkers for Immunotherapy Response To evaluate the predictive value of HCAR2+ epithelial cell infiltration in immunotherapy response, we assessed immune dysfunction in ESCC patients stratified into the low-HCAR2 and high-HCAR2 groups, and observed that T cell dysfunction scores were significantly higher in tumors with high HCAR2 expression relative to those with low HCAR2 expression (Fig. 7A). At the pan-cancer level, analysis of three independent immunotherapy-treated GEO datasets (GSE126044, GSE35640, and GSE99070) revealed consistently improved objective response rates (ORR) in patients with high-HCAR2 expression (Figs. 7B-D). Furthermore, the high-HCAR2 group exhibited reduced immune exclusion scores (Fig. 7E) and enhanced immunotherapy efficacy (Figs. 7F–H). ESCC Patients in high-HCAR2 group also had lower TIDE (Tumor Immune Dysfunction and Exclusion) scores (Fig. 7I), which indicates a diminished potential for immune evasion. Validation in additional immunotherapy-treated pan-cancer cohorts (GSE111636, GSE115821, and GSE135222) confirmed the association between high-HCAR2 expression and favorable immunotherapy outcomes (Figs. 7J–L). In the IMvigor210 cohort of patients with urothelial carcinoma, those with high-HCAR2 expression showed superior clinical responses (Fig. 7M) and prolonged progression-free survival (PFS) following anti-PD-L1 therapy (Fig. 7N). These findings were further validated in the SRP067586 dataset (Fig. 7O). Collectively, these results establish HCAR2+ epithelial cells as a promising predictive biomarker for immunotherapy sensitivity across multiple cancer types. HCAR2+ epithelial cell infiltration predicts immune therapy response. A Box plot showing T cell dysfunction in groups with high and low HCAR2 expression. B-D Violin plots showing HCAR2 expression levels in immune therapy response groups from GEO datasets (B) GSE126044, (C) GSE35640, and (D) GSE99070. E Box plot showing exclusion in groups with high and low HCAR2 expression. F-H Stacked bar plots showing immune therapy response in indicated groups from the GEO datasets (F) GSE126044, (G) GSE35640, and (H) GSE99070. I Box plot showing TIDE scores in indicated groups. J-L Stacked bar plots showing immune therapy response in indicated groups from GEO the datasets (J) GSE111636, (K) GSE115821, and (L) GSE135222. M Stacked bar plot showing immune therapy response in indicated groups from the IMvigor210 cohort. (N) Survival curves showing survival prognosis in indicated groups. O Stacked bar plot showing immune therapy response in indicated groups from the SRP067586 dataset. Data in A-E, I, were analyzed using Wilcoxon rank-sum test. *, **, *** and **** indicated p < 0.05, 0.01, 0.001 and 0.0001 High HCAR2 Expression Predicts Enhanced Response to Immune Checkpoint Blockade (ICB) Therapy We hypothesized that HCAR2+ epithelial cells, due to their capacity to upregulate PD-L1, exhibit increased sensitivity to PD-L1/PD-1-targeted immunotherapy. To test this, HCAR2-KD and Scramble PDOs were subcutaneously implanted in humanized NSG mice. Starting on day 3 post-engraftment, mice received intraperitoneal injections of either IgG isotype or atezolizumab (anti-PD-L1, 100 mg/kg every 3–4 days for 3 weeks) (Fig. 8A). Longitudinal monitoring revealed that Scramble exhibited accelerated growth compared to HCAR2-KD under IgG treatment, while Atezolizumab treatment induced significant tumor regression in Scramble group, but had minimal effect on HCAR2-KD group, as evidenced by growth curves (Fig. 8B) and terminal tumor weight at day 45 (Fig. 8C). Flow cytometric analysis of tumor-infiltrating lymphocytes demonstrated that atezolizumab significantly increased CD8+ T cells infiltration, reduced apoptosis and enhanced effector function in Scramble, whereas these effects were abrogated in HCAR2-KD (Figs. 8D-F). These data establish a paradoxical role for HCAR2: while it drives PD-L1-mediated immunosuppression to promote tumor progression, its high expression simultaneously confers heightened sensitivity to PD-L1 blockade therapy. This suggests HCAR2 could serve as both a prognostic marker for immune evasion and a predictive biomarker for atezolizumab responsiveness in PD-L1-high cancer. HCAR2-high tumors are more sensitive to anti-PD-L1 therapy. A Schematic diagram of the treatment protocol. B Tumor growth curves of indicated groups were shown. C Tumor images and weights at day 45. D Flow cytometric analysis of tumor-infiltrating CD8+ T cells. E Flow cytometric analysis of apoptosis in tumor-infiltrating CD8+ T cells. F Flow cytometric analysis of IFN-γ-positive tumor-infiltrating CD8+ T cells. Data in B were measured by two-way ANOVA and post hoc analysis to test mouse tumor growth between groups. Data in C-F were presented as mean ± SD and analyzed using two-tailed, unpaired Student’s t test. *, **, *** and **** indicated p < 0.05, 0.01, 0.001 and 0.0001 Discussion In this study, we identified a distinct HCAR2+ epithelial cell subpopulation (CNVhigh_Epi5) in ESCC through single-cell RNA sequencing. This subset exhibits a unique transcriptional profile and preferentially communicates with PDCD1+ TIGIT+ NK-like exhausted T cells via the PD-L1/PD-1 axis, forming an immunosuppressive niche that dampens antitumor immunity. Our findings align with emerging evidence of a “malignant cell-exhausted T cell axis” in tumor immune evasion [17, 18]. Notably, HCAR2 + epithelial cell infiltration inversely correlates with immune activation markers and antigen presentation signals, suggesting these cells may orchestrate immune editing during ESCC progression-a mechanism analogous to clonal evolution-driven immune evasion in melanoma [19, 20]. Furthermore, genomic instability likely amplifies HCAR2-driven immunosuppression, as high HCAR2 expression strongly associates with CNV burden in oncogenic regions. This vicious cycle mirrors colorectal cancer models where chromosomal instability activates NF-κB to suppress antitumor immunity [21]. We demonstrated that HCAR2 suppresses CD8+ T cells function through STAT3-mediated PD-L1 upregulation, thereby accelerating ESCC progression. STAT3 activation not only directly binds to the PD-L1 promoter to enhance its transcription [22] but also promotes tumor survival via proliferation, anti-apoptosis, and angiogenesis pathways [23–25]. Consistent with this, HCAR2 knockdown in vivo significantly inhibited tumor growth while enhancing CD8+ T cells infiltration and effector function. Clinically, high HCAR2 expression correlates with PD-L1/PD-1 pathway activation, increased immunosuppressive myeloid cell infiltration, and CD8+ T cell exhaustion-phenotypes concordant with studies linking myeloid-derived suppressor cells to T cell dysfunction [26, 27]. These data suggest that HCAR2 shapes an immunosuppressive microenvironment through upregulating PD-L1 to inhibit T cell function and recruiting immunosuppressive myeloid cells to further promote immune evasion. Our discovery of the HCAR2-STAT3-PD-L1 regulatory axis provides a rationale for novel combination therapies. While STAT3 inhibitors effectively block HCAR2-induced PD-L1 upregulation in vitro, their synergy with ICB warrants further exploration. Preclinical studies have shown STAT3 inhibitors enhance ICB efficacy in melanoma and breast cancer models [28–30], Future research could explore the combination of HCAR2 inhibitors with immune checkpoint inhibitors to enhance anti-tumor immune responses. Paradoxically, although high HCAR2 expression predicts poor prognosis, it also associates with improved response to anti-PD-L1 therapy. This dichotomy mirrors the “double-edged” role of PD-L1 across cancers: elevated PD-L1 signifies active immune evasion (poor prognosis) yet indicates susceptibility to PD-1/PD-L1 blockade [31–34]. In humanized mouse models, high-HCAR2 tumors showed pronounced sensitivity to atezolizumab, with marked tumor regression and restored CD8+ T cell cytotoxicity-likely due to addiction to PD-L1-mediated immune escape. Similar mechanisms are observed in EGFR-mutant non-small cell lung cancer, where PD-L1 upregulation drives both immunosuppression and ICB responsiveness [35]. Additionally, reduced immune exclusion scores and lower TIDE (Tumor Immune Dysfunction and Exclusion) scores in HCAR2-high tumors suggest a transition from immune-excluded to immune-inflamed TMEs, which are more amenable to ICB [36]. Our study positions HCAR2 as a dual biomarker: a prognostic indicator for high-risk ESCC requiring aggressive intervention and also a predictive marker for ICB sensitivity, akin to tumor mutational burden or microsatellite instability [37]. However, validation in larger prospective cohorts is essential to confirm HCAR2’s clinical utility. Mechanistically, the interplay between HCAR2 and lactate metabolism merits investigation, given HCAR2’s role as a hydroxycarboxylic acid receptor and lactate’s known capacity to activate STAT3-PD-L1 signaling [38–40]. Targeting the metabolic-immune crosstalk may offer novel therapeutic avenues. Conclusion By elucidating HCAR2’s central role in ESCC immune evasion and treatment response, our work advances molecular stratification strategies for precision immunotherapy. Dual targeting of the HCAR2-STAT3-PD-L1 axis with small-molecule inhibitors and ICB could redefine therapeutic paradigms for ESCC patients. Supplementary Information Supplementary Material 1. Authors’ Contributions Z.L. is responsible for Conceptualization, Investigation, Writing - Original Draft, Review and Editing; G.S. is responsible for Investigation and Resources; X.M. is responsible for Methodology and Visualization; C.M. is responsible for Data Management; D.A. and R.Z. are responsible for Formal Analysis; Z.C. and Q.S. are responsible for Software - related work; X.G. and T.Z. wrote the original draft; C.X. and B.N. are responsible for Review and Editing; Y.C. and Z.J. are responsible for Validation; Y.X. is responsible for Project Management; K.W. is responsible for Funding Acquisition; L.W. is responsible for Supervision; X.W. is responsible for Supervision, Project Management and Funding Acquisition. Funding The research is supported by grants from the National Natural Science Foundation of China (No. 82273362, No. 82472973), the Natural Science Foundation of Tianjin, China (No. 22ICQNIC00100, No. 24JCQNJC00390); Joint Funds of the Natural Science Foundation of Tianjin (No. 25JCLMJC00500); The Science & Technology Development Fund of Tianjin Education Commission for Higher Education (2022K1221); Tianjin Health Research Project (No. TJ WJ2024QN016); The Science Technology Development Fund of the State Administration of Traditional Chinese Medicine in Hebei Province (No. T2025063); Inner Mongolia Autonomous Region Science and Technology Innovation Guidance Project (No. CXYD 2020BT09). Data Availability Research data can be obtained from the Genome Sequence Archive for Human (GSA-Human, https://bigd.big.ac.cn/gsa-human/) under the China National GeneBank DataBase (CNGB). The relevant original sequencing files have been archived with the accession number HRA003312. This data includes the single-cell transcriptome data of tumor specimens and peripheral blood mononuclear cells (PBMCs) from 12 esophageal squamous cell carcinoma (ESCC) patients who received neoadjuvant treatment at the Tianjin Medical University Cancer Institute and Hospital. It covers the transcriptomic characteristics of subpopulations such as epithelial cells, T cells, macrophages, and stromal cells. The specific content includes raw single-cell RNA sequencing data, gene expression matrices, cell clustering results, differential expression analysis results, etc. The article clearly states that “data, analytical methods, and research materials will be made available to other researchers.” For further resource or methodological support, you can contact the corresponding author’s team (such as Xiuchao Wang, email: wangxiuchao2008@163.com) for assistance. Declarations Ethics Approval and Consent to Participate The use of patient-derived tissues and clinical data in this study was approved by the Ethics Committee of Tianjin Medical University Cancer Institute and Hospital, China (Approval No.: EK20240211), conducted in strict accordance with the principles of the Declaration of Helsinki, and all patients provided written informed consent for the collection and use of their tumor specimens and associated clinical information for research purposes. Additionally, animal experiments involving humanized mouse models were reviewed and approved by the Animal Ethical and Welfare Committee (AEWC) of Tianjin Medical University Cancer Institute and Hospital (Approval No.: 2024001), with procedures adhering to national guidelines for the care and use of laboratory animals to minimize discomfort and ensure strict compliance with ethical standards for animal welfare. Consent for Publication The manuscript does not contain any individual person’s data, images, or case reports that could potentially identify patients. All studies involving human tissues were conducted with anonymized samples, and no personal or identifiable information is included. Therefore, formal consent for publication from individual patients or their guardians is not required. Competing Interests The authors declare no competing interests. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data Transparency Statement Data, analytic methods, and study materials will be made available to other researchers. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Zekun Li, Guangcong Shen, Xiaoqing Ma and Chenyang Meng contributed equally to this work. Contributor Information Yongjie Xie, Email: xieyongjie@tjmuch.com. Kaiyuan Wang, Email: kywang@tmu.edu.cn. Liangliang Wu, Email: wuliangliang830906@126.com. Xiuchao Wang, Email: wangxiuchao2008@163.com. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement Research data can be obtained from the Genome Sequence Archive for Human (GSA-Human, https://bigd.big.ac.cn/gsa-human/) under the China National GeneBank DataBase (CNGB). The relevant original sequencing files have been archived with the accession number HRA003312. This data includes the single-cell transcriptome data of tumor specimens and peripheral blood mononuclear cells (PBMCs) from 12 esophageal squamous cell carcinoma (ESCC) patients who received neoadjuvant treatment at the Tianjin Medical University Cancer Institute and Hospital. It covers the transcriptomic characteristics of subpopulations such as epithelial cells, T cells, macrophages, and stromal cells. The specific content includes raw single-cell RNA sequencing data, gene expression matrices, cell clustering results, differential expression analysis results, etc. The article clearly states that “data, analytical methods, and research materials will be made available to other researchers.” For further resource or methodological support, you can contact the corresponding author’s team (such as Xiuchao Wang, email: wangxiuchao2008@163.com) for assistance.