Single-cell landscape of peripheral and tumor-infiltrating immune cells in HPV-negative HNSCC

Abstract

Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide. HPV-negative HNSCC, arising in diverse upper airway mucosal niches, is particularly aggressive, with poor 5-y survival and a limited response to immune checkpoint inhibitors. A deeper understanding of the tumor-localized immune landscape is essential to uncover actionable immunotherapeutic targets. Here, we integrated two single-cell RNA sequencing (scRNA-seq) datasets from 29 samples totaling nearly 300,000 immune cells to dissect immune rewiring during tumor progression and lymph node metastasis in HPV-negative HNSCC. We identified distinct shifts in adaptive immune cell populations across 14 peripheral blood mononuclear cell (PBMC) and 21 tumor-infiltrating immune cell (TIC) states. Notably, TICs exhibited enriched interferon response and immunomodulatory gene signatures, in contrast to PBMCs, indicating tumor-specific immune imprinting. Ligand-receptor analysis revealed that immunosuppressive crosstalk between macrophages and cytotoxic cells was associated with advanced disease. To spatially validate these transcriptional states, we conducted multiplexed immunofluorescence profiling on nine locally invasive HPV-negative HNSCCs, all from the ventrolateral tongue mucosa. Spatial proteomics confirmed peritumoral enrichment of activated (CD107a⁺, ICOS⁺) NK and CD8⁺ T cells and intratumoral accumulation of exhausted (PD-1⁺, PD-L1⁺) phenotypes, mirroring pseudotime trajectories inferred from scRNA-seq. These findings highlight spatially localized cytotoxic cell exhaustion as a key immune evasion mechanism in HPV-negative HNSCC and underscore the value of integrating spatial and single-cell data to reveal therapeutic vulnerabilities.

Keywords: HPV-negative HNSCC; immune exhaustion; single-cell RNA sequencing (scRNA-seq); spatial proteomics; tumor immune microenvironment; tumor-infiltrating immune cells.

Figure 1.

Integrated atlas of PBMCs and TICs in HNSCC samples. UMAPs colored by cell type (a), pathological node (b) and pathological stage (c) in PBMCs. The bar graphs show the average proportions of cell types across pathological nodes (d) and pathological stages (e). Box and whiskers show the proportions of Tem/Temra cytotoxic T cells (f) and CD16⁺ NK cells (g). UMAPs colored by cell type (h), pathological node (i) and pathological stage (j) in TICs. The bar graphs show the average proportions of cell types across pathological nodes (k) and pathological stages (l). Box and whiskers show the proportions of Tem/Trm cytotoxic T cells (m), gamma delta T cells (n) and regulatory T cells (o). HNSCC—head and neck squamous cell carcinoma; PBMC—peripheral blood mononuclear cells; NK—natural killer; pDC—plasmacytoid dendritic cell; Tcm—T central memory; Tem—T effector memory; Trm—T resident memory; Temra—T effector memory expressing CD45RA; DC—dendritic cell; MAIT—mucosal associated invariant T cell; UMAP—uniform manifold approximation and projection; TIC—tumor infiltrating cells.

Figure 2.

Differentially expressed genes at the single-cell level in TICs. Jitter plots indicating differentially expressed genes between N status (a) and pathological stages (b). The size of each dot represents the percentage of cells expressing the gene. The cell types are highlighted by color. Genes were ranked based on the average Log2FC. Highlighted genes are those that showed the highest differential expression with known immunological function among the 20 genes with the highest Log2FC in each cell annotation. The black line represents a Log2FC of zero. The dotted line represents the Log2HR threshold of 0.25. HNSCC—head and neck squamous cell carcinoma; NK—natural killer; pDC—plasmacytoid dendritic cell; Tcm—T central memory; Tem—T effector memory; Trm—T resident memory; Temra—T effector memory expressing CD45RA; DC—dendritic cell; MAIT—mucosal associated invariant T cell.

Figure 3.

Gene Set Enrichment Analysis (GSEA). The terms were determined using the MSigDB_Hallmark library. (a) GSEA analysis was conducted per cell type, comparing N⁺/N0 cells (A) or T3 + T4/T1 + T2 cells (b) and considering all DEGs. The graphical representation illustrates the enrichment of cell types present in TICs from the integrated datasets GSE139324 and GSE164690. The size of the points in the figure represents −Log padj value of each enrichment in that specific cell type, and the color represents the normalized enrichment score (NES) from lowest (blue) to highest (red). NK—natural killer; pDC—plasmacytoid dendritic cell; Tcm—T central memory; Tem—T effector memory; Trm—T resident memory; Temra—T effector memory expressing CD45RA; DC—dendritic cell; MAIT—mucosal associated invariant T cell.

Figure 4.

Cell‒cell communication in TIC. Circular plots illustrating cellular communication among major cell types, representing the numbers of interactions between these cells in TIC from N0 (a) of N⁺ (b). Bar plots representing the Log2 fold change (N⁺ over N0) in outgoing interactions from intermediate macrophages and alveolar/TREM2⁺ macrophages (c) and incoming interactions from Trm/Tem and Temra/Tem cytotoxic T cells (d). HNSCC—head and neck squamous cell carcinoma; NK—natural killer; pDC—plasmacytoid dendritic cell; Tcm—T central memory; Tem—T effector memory; Trm—T resident memory; Temra—T effector memory expressing CD45RA; DC—dendritic cell; MAIT—mucosal associated invariant T cell; UMAP—uniform manifold approximation and projection; TIC—tumor infiltrating cells.

Figure 5.

Gene trajectories in TICs. Jitter plots with average line illustrating the relative expression of AREG (a), KLRG1 (b), TCF7 (c), GZMA (d) and KIR2DL4 (e) in NK cells. The blue dots represent cells from N0, and the red dots represent cells from N⁺. PCA representing gene trajectories of CHUK, HLA-DRB5, HLA-DQA1, HLA-DQB1, CD3D, HLA-DRA, NFKBIA, HLA-DRB1, TRAC, and UBE2N in cytotoxic T cells over pseudotime (f), pathological stage (g) and PCDC1 expression (h). PCA—Principal component analysis.

Figure 6.

Spatial proteomics of HPV-negative HNSCC. The spatial multi-IF assay at lower resolution presents a representative area from one of the tumors analyzed in the spatial proteomics study. In this image, intratumoral and peritumoral regions are highlighted, with analyses focused on compartmentalizing these two subsets within each tumor. The intratumoral region (left) shows NK cells and CD8⁺ T cells (dashed white circle in the central panel) with lower concentrations of immune markers compared to the peritumoral region (right) (a). Z score heatmap representation of immune marker expression levels in NK and CD8⁺ T cells across different tissue compartments (whole tissue, peritumoral, and intratumoral regions). (b) Spatial distribution of immune markers in tumor regions. The left panel shows CD107a enrichment, with peritumoral areas noted in red and tumor-enriched regions marked in blue. The right panels display PD-1, PD-L1, HLA-A, and ICOS expression maps according to TACIT annotation. For each marker, the peritumoral and intratumoral regions are outlined with dashed lines (c). NK—natural killer.