Macrophages are among the most abundant immune cells in the pancreatic ductal adenocarcinoma (PDAC) tumor microenvironment (TME) and play a key role in regulating the immunosuppressive niche that facilitates tumor growth. Although recent three-dimensional (3D) culture systems using patient-derived materials have advanced our understanding of tumor biology, most models lack key cellular TME components and thus fail to capture tumor-immune cell interactions. To address this gap, we developed an in-vitro 3D co-culture model incorporating PDAC patient-derived organoids (PDOs) and macrophages within a synthetic hydrogel matrix. We optimized culture conditions by tuning medium and matrix conditions to support both cell lineages. Flow cytometry and transcriptomic analyses revealed that initially undifferentiated macrophages adopt an M2-like profile upon exposure to PDAC PDOs in starPEG-heparin hydrogels, mirroring the macrophage phenotypes observed by multiplex immunohistochemistry in the matched primary PDAC tissues. Cytokine secretome profiling revealed PDO-specific differences, indicating distinct underlying macrophage polarization subtypes. Collectively, our starPEG-heparin hydrogel-based 3D co-culture enables hypothesis-driven and physiologically relevant studies of tumor-macrophage interactions and may advance immune-modulatory treatment strategies in patients with PDAC.
Patients with high-risk endometrial carcinoma (HR-EC) experience substantial mortality despite multimodality therapy, molecular classification and contemporary care. We determined whether proteomic tumor programs capture lethal risk beyond established clinical and genomic classifiers and may inform translational investigations. We performed integrated clinical, genomic, and quantitative proteomic profiling of archival FFPE primary tumors from 274 patients with stage I-III HR-EC treated with curative intent and long-term follow-up. Molecular subtyping, clinical-genomic risk modeling, and proteomic analyses evaluated relationships with endometrial cancer (EC)-specific death and progression. A proteomic risk score for progression was developed and validated in 73 independent HR-EC patients. Among patients with non-POLE tumors, TCGA molecular subtype and TP53 mutation status did not significantly stratify progression or EC-specific mortality. Clinical-genomic classification and regression tree analysis identified distinct risk groups with divergent outcomes. Quantitative proteomic profiling identified reproducible tumor programs independently associated with lethal outcomes after adjustment for clinicopathologic and molecular subtypes. Semi-supervised protein clusters classified high vs. low risk of progression in either serous or grade 3 endometrioid carcinoma. A proteomic risk score stratified progression risk across training, testing, and independent cohorts and recapitulated adverse proteomic biology. Proteomic integration improved prognostic risk stratification and warrants further validation and translational investigation in HR-EC.
Accurate tumor-node-metastasis (TNM) staging is essential for treatment decision-making, prognostic stratification, and outcomes modeling in oncology. However, real-world clinical databases often extend over decades and inherently encompass multiple successive staging systems, introducing substantial heterogeneity that undermines the stability and interpretability of survival analyses, prognostic modeling, and staging system evaluation. To address this challenge, we developed an automated framework that combines TNM staging criteria and staging-relevant anatomical knowledge with a large language model (LLM) and a second-stage reflection step to convert narrative radiology reports into unified AJCC/UICC 9th edition TNM staging. In an expert-annotated cohort of 340 NPC patients, we evaluated multiple LLMs and found that the GLM4.5-based framework achieved the highest accuracy for T (0.95), N (0.86), and M (0.99) classification. Based on this optimal performance, the framework was applied to a real-world cohort of 51,242 NPC patients between 2010 and 2025. A stratified random validation subset of 200 cases showed high concordance with the expert reference standard. In downstream utility analyses, the unified staging was associated with a higher C-index and more distinct survival-curve separation than existing electronic health record staging. Collectively, this study provides a scalable framework for unifying heterogeneous historical cancer staging data, supporting longitudinal analyses in large real-world oncology cohorts.
Immunotherapy efficacy in non-small cell lung cancer (NSCLC) remains limited, and aberrant activation of the Wnt/β‑Catenin pathway plays a significant role. However, the precise involvement of its receptor, Frizzled7 (Fzd7), in immune checkpoint inhibitor (ICI)-resistance remains unexplored. Here, we elucidate the role of Fzd7 in shaping the immunosuppressive tumor microenvironment (TME) and develop a novel bispecific antibody (BsAb, named SHH-YN-Bi) that co-targets Fzd7 and programmed cell death ligand 1 (PD-L1). This study reveals that Fzd7 is remarkably enriched in human NSCLC tissues and correlates with PD-L1 expression and immune exclusion. SHH-YN-Bi was generated via “Knob-into-Hole”, exhibited effective dual binding to both Fzd7 and PD-L1, and demonstrated superior tumor targeting compared with Atezolizumab. Mechanistically, SHH-YN-Bi concurrently recruited and activated CD8+ T cells into the TME in a basic leucine zipper transcriptional factor ATF-like 3 (Batf3)+ conventional type 1 dendritic cells (cDC1s)-dependent manner, leading to robust anti-tumor activity in ICI-resistant orthotopic NSCLC models. Furthermore, scRNA-seq analysis uncovered a potent correlation between Fzd7 and cancer-associated fibroblast (CAF)-mediated extracellular matrix (ECM) deposition. Notably, SHH-YN-Bi treatment induced a phenotypic switch of CAFs from a highly oncogenic to a normal state, which was associated with elevated CD8+ T cell infiltration. Collectively, SHH-YN-Bi overcomes ICI resistance in NSCLC by simultaneously blocking the oncogenic Wnt3a/Fzd7/β-Catenin signaling and PD-1/PD-L1 axis. Our findings highlight the critical role of Fzd7 in immunosuppressive TME and offer an innovative approach for NSCLC treatment.
Gut microbiota shapes cancer initiation, progression, immune evasion, and therapeutic response. Dysbiosis remodels the tumor microenvironment through innate and adaptive immune regulation, microbial metabolites, pattern-recognition receptor signaling, epigenetic and metabolic reprogramming, extracellular vesicles, and intratumoral microbes. This review summarizes microbiota–immune interactions in immune checkpoint inhibitor therapy and discusses microbiome-targeted strategies, including fecal microbiota transplantation, probiotics, and diet, to improve immunotherapy outcomes.
Spatial organization in esophageal squamous cell carcinoma resection specimens after chemoradiation remains poorly characterized. We profiled 56 patients using Visium spatial transcriptomics and identified eight niches. All 37 evaluated transcriptional signatures showed marked variation across the five niches with adequate patient representation, revealing structured biological compartmentalization. ΔNp63/NRF2-associated enrichment was highest at the Tumor Leading Edge and lowest at the Stromal Boundary; within Fibro-Immune Niche spots, it was also higher in perineural-invasion-positive tumors. In exploratory analyses of patients with pathological regression Grades 1–2, higher inflammation-, ΔNp63/NRF2-, SPP1-, multigene radiotherapy-resistance-, and oncogenic-transcription-factor-associated enrichment in the Fibro-Immune Niche was associated with shorter failure-free survival. Related Fibro-Immune programs also showed concordant adverse trends in the full cohort, although independent prognostic value was not established in either analysis. Post-treatment Fibro-Immune programs provide biologically coherent candidates for independent validation.
EGFR-mutant non-small cell lung cancer (NSCLC) shows limited benefit from immune checkpoint inhibitors (ICIs), but the spatial immune mechanisms associated with resistance remain unclear. We applied a single-cell-resolution multimodal spatial proteomic–transcriptomic framework integrating imaging mass cytometry (IMC; n = 12) and Xenium-based spatial transcriptomics (n = 8) to map the tumor microenvironment of EGFR-mutant NSCLC after chemoimmunotherapy. Single-cell RNA sequencing datasets (n = 11) were incorporated for cell–cell interaction inference and pathway analysis. Spatial profiling showed that non-major pathologic response (non-MPR) tumors were enriched for immunosuppressive cellular neighborhoods containing SPP1⁺ macrophages and CD1C⁺ conventional dendritic cells (cDCs). These CD1C⁺ cDCs exhibited reduced immune activation and antigen-presentation programs. Spatial distance analysis showed closer proximity between SPP1⁺ macrophages and CD1C⁺ cDCs in non-MPR tumors than in MPR tumors. Ligand–receptor analysis identified candidate macrophage-derived signals associated with altered CD1C⁺ cDC states, which was supported by Xenium-based co-localization and reduced pro-inflammatory programs in spatially interacting CD1C⁺ cDCs. Together, these findings suggest that an SPP1⁺ macrophage–CD1C⁺ cDC spatial niche is associated with reduced immunotherapy response and may inform future strategies to restore DC immunogenicity in EGFR-mutant NSCLC.
In this exploratory epigenome-wide association study (EWAS), we sought to investigate associations of pre-treatment blood-derived DNA methylation with survival and response outcomes among patients with recurrent/metastatic (R/M) head and neck squamous cell carcinomas (HNSCCs) who received immune checkpoint inhibitor (ICI) therapy. Patients with R/M HNSCC were recruited from the Dana-Farber Cancer Institute, Dartmouth Cancer Center, and Rhode Island Hospital. Peripheral blood was collected prior to the start of ICI, and DNA methylation was measured using the Illumina EPIC V1 and V2 BeadArrays. After strict quality control, the top 10% most variable CpG sites (n = 67,712) were included in an EWAS of methylation with overall survival (OS) and progression-free survival (PFS). Models were adjusted for key covariates, including the DNA methylation-derived immune cell-type proportions. Among the 116 patients enrolled, 64 (55.2%) died during follow-up, and 22 (19.0%) experienced cancer progression. Of the 30 patients (25.9%) who remained progression-free, 13 showed a complete response to the ICI therapy. We identified two CpG sites (cg04079758 and cg10654272) that were associated with improved PFS (false discovery rate [FDR] < 0.05). The CpG sites cg04079758 and cg10654272 are located in the South Shelf region on chromosome 19 and in an OpenSea region on chromosome 6, respectively. Additionally, one of the CpG sites, cg04079758, was located in Kallikrein 9 (KLK9), a gene involved in tissue remodeling and inflammation within the tumor microenvironment. No CpG site was statistically significantly associated with OS. Distinct blood DNA methylation signatures may be associated with ICI therapy-related outcomes in patients with R/M HNSCC.
Combination chemotherapy with immune checkpoint blockade has reshaped the treatment landscape for recurrent endometrial carcinoma (EC). However, subgroup analyses have demonstrated limited efficacy in patients previously treated with platinum-based adjuvant chemotherapy. To investigate how prior adjuvant chemotherapy affects the tumor microenvironment (TME) at recurrence, we analyzed 32 matched primary–recurrent EC pairs by performing multi-omic profiling in 20 pairs and multiplex immunohistochemistry (mIHC) in all cases. Recurrent tumors with prior platinum-based adjuvant chemotherapy exhibited increased variant allele frequencies of mutations in MYC signaling-related genes, with enrichment of glycolysis, regulatory T (Treg) cells, and M2 macrophage polarization. mIHC demonstrated increased Treg-cell infiltration and the M2/M1 macrophage ratio in recurrent tumors compared with matched primary tumors. Moreover, progression-free survival after recurrence correlated more strongly with the TME of recurrent tumors than with that of the primary tumors. Platinum-based adjuvant chemotherapy is associated with the emergence of immunosuppressive TME features in recurrent EC.
Myeloid lineages exhibit prominent enrichment and profound heterogeneity within the pancreatic ductal adenocarcinoma (PDAC) microenvironment, significantly impacting clinical outcomes. Here, we analyzed 45,705 high-quality myeloid cells from patient-matched tumor and blood samples, with normal blood and tissues as controls. This revealed two tumor-promoting subsets: GBP5⁺ neutrophils sustaining a persistent pro-inflammatory circuit, presumably mediated by STAT3, and THBS1⁺ monocytes putatively orchestrating immunosuppression via Amphiregulin-driven differentiation blockade. We developed moGT, a heterogeneous graph transformer integrating multi-omic profiles with myeloid cellular abundance. moGT deconvolutes PDAC into four molecular subtypes governed by specific myeloid ecosystems (ECs), architectures supported by spatial transcriptomics. Notably, Subtype 2 exhibits an unexpectedly poor prognosis despite a favorable ‘classical’ phenotype, characterized by a GBP5⁺ neutrophil-enriched EC and NAMPT inhibition vulnerability. Conversely, aggressive Subtype 4 is defined by a THBS1⁺ monocyte-dominated EC, showing predictive sensitivity to microtubule-targeting agents. Our cross-compartment analysis links myeloid architecture to clinical heterogeneity, providing a precision oncology framework.
Metastatic relapse after curative-intent resection suggests that dissemination-associated programs may precede clinically detectable spread. We integrated clinical stratification with whole-exome sequencing, bulk transcriptomics, proteomics, single-cell transcriptomics and digital pathomics in a lung adenocarcinoma-predominant non-small cell lung cancer cohort. Relapse-prone primary tumors showed chromosomal instability, APOBEC-associated mutational features, immune attenuation and recurrent HIF/hypoxia-associated transcriptional programs, supported by higher CA9 and HIF-1α staining in the pooled progression/metastasis group. Single-cell trajectories showed progressive acquisition of related malignant states. Integrative prioritization identified STEAP2 as a candidate effector associated with HIF-1α stability and selected invasive phenotypes. Structure-guided screening nominated ZD93 as a candidate STEAP2-directed compound. ZD93 attenuated hypoxia-associated outputs and migration in vitro and was associated with fewer histologically confirmed lesions and longer humane-endpoint survival in exploratory xenograft analyses. Together, these findings delineate a HIF-associated progression-prone state that precedes relapse and nominate STEAP2 as a candidate vulnerability requiring mechanistic and therapeutic validation.
Immunological biomarkers are increasingly relevant for personalized cancer treatment, but peripheral blood-derived biomarkers are not yet used to guide therapy in head and neck squamous cell carcinoma (HNSCC). The prospective non-randomized DIREKHT study (ClinicalTrials.gov: NCT02528955, 2015-08-19) therefore integrated immune monitoring into postoperative radio(chemo)therapy (R(C)T) to explore blood-based biomarkers. In 70 oral cavity and oropharyngeal cancer patients receiving curative R(C)T, the peripheral immune status was assessed before and after therapy and during follow-up by flow cytometry-based immunophenotyping of 45 immune parameters. A machine learning workflow identified predictors of disease-free survival (DFS), using Repeated Elastic Net Technique (RENT) feature selection within repeated stratified K-fold cross-validation and nested cross-validation for tuning and assessment. This approach identified a 29-parameter immune signature from pre- and post-therapeutic profiles, with key contributors including HLA-DR + T cells, HLA-DR+ monocytes, and basophils. The best model achieved a Matthews correlation coefficient of 0.681, with pre- and post-therapeutic parameters contributing equally, highlighting immune dynamics during R(C)T. Adding clinical parameters did not improve performance (MCC = 0.678), but yielded a comparable model integrating immune and clinical variables. Blood-based immune signatures may have prognostic relevance for DFS after R(C)T in HNSCC. Validation in larger cohorts is required to confirm clinical applicability and reduce the signature.
High grade serous ovarian cancer (HGSC) has low survival partly due to the lack of methods for detection, diagnosis, and risk prediction. TP53 mutations, which drive HGSC, are found in gynecological tissues as the result of somatic evolution, but it is unknown whether an excess of mutations is linked to ovarian cancer. Here we investigate if TP53 mutation burden measured in uterine lavage, a minimally invasive gynecological liquid biopsy, can discriminate between patients with and without HGSC. We used ultradeep TP53 duplex sequencing (>15,000x duplex depth) to detect TP53 mutations in uterine lavage collected pre-operatively in 278 patients undergoing gynecological surgery for pelvic masses (average risk) or cancer risk-reduction (high risk). All lavages contained multiple TP53 mutant clones, which were used to quantify TP53 mutation burden frequency (MBF). Average risk patients with HGSC had significantly higher TP53 MBF independently of age and other risk factors (77% sensitivity, 89% specificity, AUC = 0.88). Excluding tumor TP53 clonal mutations from the lavage MBF calculation maintains this association, suggesting that it is the overall TP53 somatic mutation burden (rather than the discovery of the specific tumor driver mutation) that identifies HGSC. These results demonstrate that TP53 somatic mutations are common in uterine lavage but more abundant in patients with HGSC, highlighting a connection between TP53 somatic evolution and ovarian cancer. Uterine lavage offers a minimally invasive approach that could be valuable to identify patients with HGSC.
Abstract Pancreatic ductal adenocarcinoma (PDAC) is characterized by chemotherapy resistance, partly driven by its dense and heterogeneous tumor microenvironment (TME). Since most preclinical PDAC models inadequately capture the tissue architecture, their translational value for therapeutic testing remains limited. This study investigated organotypic tissue slice cultures (OTSCs), which preserve the multicellular tissue architecture, as a rapid platform for personalized ex vivo drug response profiling. OTSCs were generated from 27 resected PDAC specimens. Following workflow establishment and quality control, ex vivo drug profiling was performed in 15 patients using gemcitabine, gemcitabine plus paclitaxel, and FOLFIRINOX. Treatment response was assessed by quantitative digital pathology, and an ex vivo sensitivity score (EVSS) was defined. Clinical correlations with longitudinal follow-up were assessed in ten patients. OTSCs preserved tissue architecture and revealed interpatient heterogeneity. In exploratory analyses of clinically matched patients, ex vivo sensitivity was associated with prolonged PFS (median 445 vs. 141 days, log-rank p = 0.0027; HR per 10 EVSS points 0.625; 95% CI 0.484–0.807, p = 0.00032), while lower preoperative CA 19-9 levels and higher GATA6 expression were associated with higher EVSS and longer overall survival. OTSC-based profiling enables rapid, patient-specific drug response assessment in PDAC and supports evaluation as a functional stratification tool.
Targeted sequencing approaches, such as conventional cancer panels, often lack coverage of intronic regions. Consequently, gene-disrupting complex genomic rearrangements (CGRs) may go undetected, leading to false-negative results in hereditary cancer testing and limiting access to targeted therapies. Here we report two unrelated cancer patients carrying a novel germline CGR event truncating BRCA1, identified through whole-genome sequencing (WGS). The CGR designated BRCA1 del(e12-14)-inv, comprising deletion and inversion events, removed BRCA1 exons 12–14. Matched tumor genomes further showed loss-of-heterozygosity (LOH) at the locus and mutational signatures supporting homologous recombination deficiency (HRD), confirming the pathogenicity of the germline variant. Our findings suggest that WGS can identify gene-truncating germline CGRs that remain undetected by conventional target-exon–based testing. Of note, detecting these events would be clinically important because they may identify cancer patients who could benefit from PARP inhibitor therapy and may also inform genetic counselling for familial cancer predisposition. These cases highlight the value of WGS for detecting clinically actionable structural variants and support its integration into diagnostic workflows to enhance the clinical utility of genomic medicine.
Neoadjuvant immune checkpoint blockade is becoming the standard of care for stage III melanoma, and biomarkers to refine these regimens are urgently needed. Intratumoral tumor-infiltrating lymphocytes (iTIL) are essential for response, but their clinical scoring is hampered by high inter-observer variability, particularly in lymph node biopsies. Here, we present a pathologist-in-the-loop deep learning workflow, to quantify iTIL in hematoxylin and eosin-stained biopsies from melanoma lymph node metastases. Our model was trained on 127,278 pathologist-annotated cells, and externally validated on 159 whole-slide images from three pioneering phase I/II trials testing neoadjuvant ipilimumab and nivolumab in stage III melanoma, with exceptionally long-term follow up (>5 years). AI-assisted iTIL% correlated strongly with pathologist-based iTIL% (Spearman ρ = 0.76, P < 0.0001) and was predictive of major pathological response (P = 0.0008). A high AI-assisted iTIL% was associated with longer event-free survival (HR = 0.40, P = 0.010), especially when combined with a high tumor mutational burden (TMB; HR = 0.18, P = 0.0056); in the double high subgroup (n = 33), only one melanoma-related event occurred in 5 years of follow up. This work provides a roadmap towards standardised, AI-assisted iTIL scoring in melanoma lymph node metastases. Prospective studies are indicated to confirm that iTIL metrics, particularly with the TMB, could inform standard of care protocols and future trials.
Head and neck squamous cell carcinoma (HNSCC) remains a major global health burden with limited targeted therapeutic options. Here we present an integrated artificial intelligence (AI)-driven drug discovery framework combining graph neural networks (GAT, GCN, AttentiveFP and GIN), classical machine learning models (Random Forest, XGBoost and SVM), and structure-based simulations to identify potential inhibitors associated with HNSCC. Specifically, the GCN model converts SMILES strings into molecular graphs, utilizing atom features (e.g., atomic number, hybridization) and bond features (e.g., bond type, conjugation) as inputs to learn graph-level representations. Using curated bioactivity datasets from ChEMBL and PubChem, we trained seven predictive models and constructed a consensus ensemble framework that achieved high precision, recall, and an AUROC exceeding 0.90 on independent validation datasets. Based on ensemble consensus scores from seven predictive models, large-scale virtual screening of the ChEMBL and DrugBank libraries yielded 36,771 and 391 high-confidence hits, respectively, with chemical space and scaffold analyses revealing substantial structural diversity and significant overlap with DrugBank (sharing 85.9% of its scaffolds), demonstrating effective coverage of drug-relevant chemical space. Through HNSCC-associated gene screening, PPI network analysis, MCC-based hub gene ranking, and machine learning-based feature selection, PIK3CA was prioritized as a key therapeutic target for subsequent validation. Molecular docking and 200-ns molecular dynamics simulations further suggested stable binding interactions for several top candidates, including CHEMBL4454174, CHEMBL4458803, and CHEMBL4591345. In addition, clinical tissue validation showed increased PI3Kα, p-AKT, and p-S6 expression in HNSCC tumor tissues compared with adjacent tissues, supporting activation of the PI3Kα/AKT/S6 signaling axis. Functional validation using siRNA-mediated PIK3CA knockdown further indicated reduced PI3Kα expression, attenuated AKT/S6 signaling, and impaired cell viability, proliferation, and migration in HNSCC cells. These findings highlight the potential of integrating AI-based prediction, structural validation, and biological validation to accelerate the discovery of candidate therapeutics for precision oncology.
We report a novel in-frame CD274::PDCD1LG2 fusion in primary mediastinal B-cell lymphoma. Functional studies using expression constructs and CRISPR-engineered lymphoma cells confirmed formation of a PD-L1::PD-L2 chimeric protein and marked upregulation of PD-L1 surface expression, consistent with loss of 3′-UTR-mediated repression. These findings expand the genomic landscape of structural mechanisms driving PD-L1 activation and may inform variant interpretation and patient selection for PD-1 blockade in lymphoma.
Early-onset gastric cancer (EOGC) is characterized by aggressive clinical behavior and poor prognosis; however, the immunological features underlying its progression remain incompletely understood. Here, we performed single-cell RNA sequencing on EOGC tumors, matched adjacent mucosa, and traditional gastric cancer (TGC) samples, integrating these data with publicly available gastric cancer datasets to comprehensively characterize the EOGC tumor microenvironment. Our analyses revealed that EOGC exhibits an immunosuppressive “cold” microenvironment characterized by increased exhausted CD4+ T cells, impaired anti-tumor effector T-cell activity, and enhanced B-cell isotype switching from IgG to IgA, suggesting immune evasion. Notably, we identified a THBS1-expressing macrophage subset with prominent M2-like polarization that was associated with tumor progression and unfavorable prognosis. Functional experiments demonstrated that THBS1-overexpressing THP-1-derived macrophages promoted M2 polarization and enhanced malignant phenotypes in gastric cancer cell lines, whereas inhibition of THBS1 suppressed tumor growth in vivo. Together, these findings delineate a distinct immunosuppressive ecosystem in EOGC and provide insight into the potential role of THBS1+ macrophages in EOGC progression.
Head and neck squamous cell carcinoma (HNSCC) is a highly aggressive malignancy with limited therapeutic options. Here we combine single-cell RNA sequencing, spatial transcriptomics, clinical cohorts, and experimental models to dissect malignant epithelial heterogeneity and its impact on disease progression and treatment response. We identify a malignant subtype characterized by partial epithelial–mesenchymal transition (pEMT), enriched at the invasive front and associated with poor prognosis. This pEMT subtype exhibits pronounced vasculogenic mimicry (VM) potential and resistance to anlotinib-based neoadjuvant therapy, driven by high expression of the extracellular matrix component LAMC2. Functional assays demonstrate that LAMC2 promotes proliferation, VM formation, and resistance to anti-angiogenic therapy. Mechanistically, TGF-β signaling enhances LAMC2-driven VM, while blockade of TGF-β synergizes with anlotinib to suppress tumor growth. In addition, LAMC2–CD44 interactions shape an immunosuppressive tumor microenvironment enriched in M2 macrophages, cancer-associated fibroblasts, and regulatory T cells. Together, these findings define a LAMC2+ pEMT subtype that mediates therapeutic resistance through the TGF-β–pEMT/LAMC2–VM axis, highlighting a potential strategy to overcome resistance and reprogram the HNSCC microenvironment.
