Abstract

Background Lymph node (LN) metastasis is a well-established independent prognostic factor in head and neck squamous cell carcinoma (HNSCC). Formation of suppressive tumour immune microenvironment (TIME) is a major contributor to tumour immune evasion and metastasis. However, the TIME landscape underlying LN-metastatic HNSCC remains poorly elucidated. Methods A total of 688 866 single-cell transcriptomes across 212 HNSCC samples were integrated. Comprehensive bioinformatic analyses on single-cell RNA sequencing and microarray datasets revealed a TREM2 + tumour-associated macrophage (TAM) cluster associated with LN metastasis. The functional role of TREM2 + TAMs was investigated through multiplex immunohistochemistry (mIHC) staining in clinical HNSCC cohort and in vitro co-culture experiments. Furthermore, machine learning algorithms were employed to construct a prognostic model for HNSCC. Results Integrative single-cell analysis revealed the immunosuppressive TIME of LN-metastatic HNSCC, characterised by high infiltration of exhausted CD8 + T cells (CD8 + Tex). We identified a specific TREM2 + TAM cluster that was strongly associated with CD8 + Tex infiltration and LN metastasis. In vitro experiment confirmed that TREM2 + TAMs promoted CD8 + T cell exhaustion. Mechanistically, TREM2 + TAMs exhibited a terminally differentiated phenotype driven by ETV5, and secreted SPP1 to interact with CD44 on CD8 + T cells, thus upregulating BHLHE40 to promote CD8 + Tex formation. Clinically, a prognostic model based on TREM2 + TAM signature genes was trained to independently predict HNSCC outcomes. Conclusions This study delineates the mechanism that TREM2 + TAMs promote LN metastasis in HNSCC by facilitating CD8 + T cells exhaustion via SPP1–CD44–BHLHE40 axis, proposing TREM2 + TAMs as potential therapeutic target for HNSCC.