Dataset Information

Spatial Transcriptomics Reveals a Molecular Tumor Budding Signature in Head and Neck Cancer

ABSTRACT: Tumor budding is a histopathological feature associated with poor prognosis across multiple cancer types, including head and neck squamous cell carcinoma (HNSCC). Tumor buds represent the earliest traceable local invasion and are considered the origin of minimal residual disease, local recurrence, and metastasis. However, the molecular features underlying this phenomenon in HNSCC remain only partially characterized due to challenges in isolating tumor buds while preserving spatial context. Here, we employed high-plex spatial transcriptomics to specifically isolate and characterize tumor buds, complemented by single-cell RNA sequencing and pharmacogenomic analyses to comprehensively characterize the transcriptional landscape of tumor budding in human papillomavirus (HPV)-negative HNSCC. Our unprecedented spatial resolution revealed that tumor buds exhibited distinct gene expression profiles compared to tumor bulk, with significant upregulation of epithelial-mesenchymal transition (EMT) markers, extracellular matrix remodeling genes (including fibronectin and laminin-332 subunits), and coagulation pathway components. Unlike previous bulk tissue analyses, our approach enabled differential correlation analysis that uncovered extensive gene network rewiring both within tumor cells and between tumor and stromal compartments, highlighting dynamic molecular reprogramming during the budding process. We developed a novel 28-gene tumor budding signature (TBS) that effectively distinguished tumor buds from other compartments (AUC=0.97) and demonstrated independent prognostic value in the TCGA-HNSC cohort (HR=1.54, p=0.02), representing the first molecularly defined biomarker for tumor budding applicable to conventional transcriptomic data in HNSCC. Analyses of single-cell RNA-seq datasets confirmed the tumor-specific expression of the TBS, its association with a hybrid epithelial-mesenchymal state, its expression predominantly at the leading edges of tumors, and its induction via subtypes of epidermal growth factor receptor activities. Uniquely, our pharmacogenomic analysis revealed that SCC cell lines with high TBS scores showed increased sensitivity to Raf-MEK inhibitors, suggesting clinically actionable therapeutic vulnerabilities, which was confirmed in a 3D model of early local invasion. This integrated spatial-molecular approach provides novel insights into tumor budding in HNSCC and establishes a foundation for a better understanding of the molecular mechanisms of tumor budding, as well as improved risk stratification and targeted therapeutic interventions for this aggressive tumor phenotype.

ORGANISM(S): synthetic construct Homo sapiens

PROVIDER: GSE300414 | GEO | 2026/04/17

REPOSITORIES: GEO

ACCESS DATA

Dataset's files

Source:

			Action	DRS
		Other

Items per page:

1 - 1 of 1

Similar Datasets

Project description:Background: Immune-checkpoint inhibition (ICI) only benefits a subgroup of patients with head and neck squamous cell carcinoma (HNSCC). Several molecular and cellular components of the tumor microenvironment (TME) have been hypothesized to drive either response or resistance. Here, spatially defined whole transcriptome data were analysed in search of associations of compartment-specific gene-signatures with HNSCC immunotherapy outcomes. Methods: Pre-treatment biopsy samples from 50 immunotherapy-treated recurrent or metastatic HNSCC patients as well as 12 matched post-treatment biopsies obtained after 4 weeks of treatment, constructed in tissue microarray format (YTMA496), were included in the study. The GeoMx Human Whole Transcriptome Atlas (NanoString Technologies) assay was performed on samples to allow RNA quantification of 18,677 protein encoding genes, using in situ hybridization, in three molecularly defined tissue compartments; tumor (CK), leukocyte (CD45), macrophage (CD68). Differentially expressed genes (DEGs) (P<0.05) between pre- and post-treatment biopsies in each of the tissue compartments were identified. Next, these DEGs were used as a “biological” filter for the initial 18,677 gene set and analysed using LASSO logistic regression models with the aim to obtain pre-treatment gene expression signatures for best overall response (RECIST 1.1.). The performance of each compartment signature was evaluated using the receiver operating characteristic (ROC) curve and AUCs were calculated for Best Overall Response (BOR) to ICI. Results: A six-gene signature (DDX4, COL17A1, HBA1, MMP1, GPNMB, TTN) in the CD45 compartment presented the highest AUC (0.83), followed by signatures in the CK and CD68 compartments (AUC: 0.72 and 0.68, respectively). Cross-testing of the CD45 signature in the other two compartments, as well as in a third, artificially generated “pseudo-bulk” compartment (all compartments combined), showed poor performance, indicating spatial specificity. Interestingly, the CD45 signature included three extracellular-matrix protein-encoding genes (MMP1, COL17A1, TTN), all associated with resistance (negative coefficients in the BOR model) to ICI. Fibroblast and dendritic cell populations, characterized using the CIBERSORTx gene matrix, were the immune phenotypes most closely associated with the CD45 signature. Conclusions: Our results indicate that CD45 molecular tissue compartment gene expression demonstrates increased association with ICI resistance in HNSCC. Extracellular matrix genes rather than immune-cell related genes dominated the CD45 compartment signature, highlighting the importance of non-immune stromal components within the TME and the importance of the use of spatial information in the understanding of ICI resistance.

Dataset Information

Spatial Transcriptomics Reveals a Molecular Tumor Budding Signature in Head and Neck Cancer

Dataset's files

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets