In this exploratory study, spatial transcriptomic profiling was performed on formalin-fixed, paraffin-embedded (FFPE) tissues obtained from radical resections of palatine tonsil-derived OPC (15). Regions of interest (ROIs) representing tumor tissue regions (TTRs) and lymphoid follicular regions (LFRs) were analyzed using the GeoMx DSP platform to explore immune features associated with HPV status and lymph node metastasis.

Six FFPE samples from primary tonsillar OPCs resected at Kanazawa University Hospital (2019–2020) were analyzed. All patients were male; four were HPV-positive and two HPV-negative, including three metastasis-negative and three metastasis-positive cases. Staging followed the 8th edition of the TNM classification (1). HPV status was determined by p16 immunohistochemistry and confirmed by HPV DNA PCR using type-specific primers for high-risk HPV genotypes. Cases showing diffuse (≥70%) nuclear and cytoplasmic p16 staining were classified as HPV-positive. No patient received adjuvant radiotherapy or chemotherapy. Clinical details are summarized in Supplementary Table S1. Although the cohort was small, age and stage distributions were comparable between HPV-positive and HPV-negative groups. No formal adjustment for clinicopathologic variables was performed due to the exploratory nature of the study.

This study was approved by the Kanazawa University Hospital Institutional Review Board (approval number 2016–033) and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants. Some samples were previously used in other studies, but the present analysis focused on newly generated spatial transcriptomic data addressing distinct immune features.

Spatial transcriptomic profiling was performed using the GeoMx DSP (NanoString Technologies, Seattle, WA), which enables high-plex RNA detection in FFPE samples (11, 16). The GeoMx Whole Transcriptome Atlas (WTA) interrogates ~18,000 protein-coding genes curated by the Human Gene Nomenclature Committee and RefSeq databases. Sections (5 μm) were mounted on Superfrost Plus slides (VWR, Radnor, PA). Morphology markers (PanCK, CD45, SMA) were used to distinguish epithelial, immune, and stromal compartments. Figure 1 shows representative fluorescent staining and ROI selection.

Twelve ROIs were selected per specimen (six LFRs and six TTRs), yielding 36 of each across all six patients. ROI selection was performed independently by two independent researchers (M.M.-K. and N.W.) blinded to HPV and nodal status. Data were processed with the DSP Data Analysis Suite using Q3 normalization. A linear mixed model was applied to detect differentially expressed genes (DEGs) between groups. Statistical significance was defined as |log₂| of fold-change > 0.5 and -log₁₀ P > 1.3. Differential expression was assessed using a threshold of unadjusted p < 0.05, which was considered appropriate for this hypothesis-generating, small-cohort spatial transcriptomics study. FDR correction did not yield significant hits owing to small sample size, but consistent trends were validated across independent analyses (ssGSEA, immunohistochemistry).

Functional annotation of DEGs was performed using the ClueGO plugin for Cytoscape (v3.10.3; https://cytoscape.org) to identify enriched Gene Ontology (GO) categories including Biological Process, Cellular Component, and Immune System Process (17).

Relative immune cell proportions were estimated from ROI expression profiles using CIBERSORTx (https://cibersortx.stanford.edu/) based on 22 immune subsets (18). Outputs were interpreted as relative rather than absolute, given the lack of OPC-specific single-cell reference datasets.

Gene set enrichment was evaluated by single-sample GSEA (ssGSEA) using hallmark gene sets from the Molecular Signatures Database (MSigDB; https://www.gsea-msigdb.org/gsea/index.jsp). Analyses were performed using GenePattern software v3.9.11 and the ssGSEA module v10.1.0 (19).

Consecutive 5 μm sections from FFPE blocks were immunostained as described previously (20, 21). Mouse monoclonal antibodies against eosinophil major basic protein (EMBP; 1:20) and mast cell tryptase (1:10,000; Bio-Rad, Hercules, CA) were used. Visualization was achieved with diaminobenzidine and methyl green counterstaining. Eosinophilic nasal polyps and inflamed tonsils served as positive controls; nonimmune serum was used for negative controls. For each specimen, three high-power fields (200×) were independently reviewed by two blinded observers with >90% concordance. The staining was evaluated qualitatively to confirm spatial expression patterns and cellular localization; no quantitative scoring (e.g., H-score or positive cell ratio) was applied due to the exploratory nature of the study.

Normalized expression data were used for differential analyses. Linear mixed modeling incorporated patient ID as a random effect, with HPV status and tissue region as fixed factors. Principal component analysis (PCA) was applied to normalized LFR profiles to visualize clustering and exclude batch effects. The Mann–Whitney U test was used for two-group comparisons. Correlations between enrichment scores and immune cell proportions were assessed by Spearman’s rank test.

For exploratory DEG analyses using GeoMx whole-transcriptome data, statistical significance was defined as unadjusted p < 0.05 (−log10 p > 1.3) and |log2 fold change| > 0.5. Given the exploratory nature of this analysis, these thresholds were used to identify potential immune-related transcriptional patterns. For subsequent validation analyses, including enrichment score correlations and immune cell density comparisons, statistical significance was defined as p < 0.01 to minimize false-positive findings. Given the small cohort size and exploratory nature of the study, no formal power calculation was performed.

PCA of normalized gene expression from LFRs confirmed the absence of batch effects (Supplementary Figure S1). Within HPV-positive cases, modest separation by metastatic status (F = 0.221, p = 0.028) suggested biologically meaningful heterogeneity rather than technical bias. PCA of TTRs revealed diffuse clustering, consistent with the inherent cellular heterogeneity of tumor compartments.

Spatial transcriptomic profiles differed markedly between HPV-positive and HPV-negative OPCs (Supplementary Figure S2).

In TTRs, 163 differentially expressed genes (DEGs) were detected, with HPV-negative tumors showing upregulation of pathways related to cellular transport and differentiation, while HPV-positive tumors showed no enriched GO terms. This lack of enrichment likely reflects relatively uniform transcriptional profiles among HPV-positive tumors, suggesting lower pathway-level variability or more homogeneous immune activation patterns, rather than technical limitations.

In LFRs, HPV-negative cases demonstrated activation of transcriptional programs related to RNA processing and nucleosome formation, whereas HPV-positive LFRs exhibited enrichment only in the “negative regulation of B cell apoptotic process,” suggesting enhanced B cell survival.

These findings support distinct immune and transcriptional landscapes between HPV-positive and -negative OPCs, consistent with their classification as biologically separate diseases (1–3).

Analyses were next restricted to HPV-positive cases (n = 4) to explore lymphatic dissemination. Given the limited sample size and exploratory aim of this study, differential expression was assessed using an unadjusted p < 0.05 threshold, as FDR correction was overly conservative and yielded no significant hits.

In TTRs, metastasis-positive tumors upregulated genes associated with necrotic cell death and chromatin regulation, while metastasis-negative cases showed enrichment of immune activation and keratin-related pathways (Supplementary Figures S3A–C).

In LFRs, metastasis-positive cases upregulated pathways linked to T cell activation and antigen processing, whereas metastasis-negative cases showed enrichment of cell proliferation and receptor regulation terms (Supplementary Figures S3D–F).

Spatial heatmap of selected DEGs in LFRs, illustrating distinct transcriptional patterns between non-metastatic and metastatic HPV-positive OPCs (Figure 2A).

Immune cell composition estimated by CIBERSORTx revealed a Th2-skewed phenotype in non-metastatic tumors (Figures 2B, C; Supplementary Figure S4).

In TTRs of metastasis-negative cases, naïve B cells, M2 macrophages, monocytes, activated mast cells, and eosinophils were significantly enriched, suggesting a Th2-skewed profile.

In LFRs, plasma cells, naïve B cells, M2 macrophages, activated mast cells, eosinophils, and neutrophils were enriched. In contrast, metastasis-positive cases showed enrichment of activated dendritic cells in LFRs, indicating a shift toward Th1 responses.

ssGSEA analysis using hallmark gene sets revealed gene set enrichment differences based on nodal status (Figures 2D: Supplementary Figure S5).

In LFRs, inflammatory and interferon response signatures predominated in metastasis-positive cases, while “HALLMARK_ESTROGEN_RESPONSE_EARLY” and “LATE” were enriched in metastasis-negative cases. In contrast, these estrogen-response enrichments were not detected in the TTRs, indicating that hormone-responsive signaling was compartment-specific to the peritumoral LFRs. Conversely, bile-acid-metabolism pathways were upregulated in LFRs of metastasis-positive cases but absent in TTRs, suggesting a metabolic activation associated with lymphatic dissemination within the peritumoral compartment.

Notably, early estrogen response scores positively correlated with eosinophil and activated mast cell fractions (Figures 3A, B), whereas late estrogen response scores showed weaker associations (Figures 3C, D). Early and late estrogen responses were positively correlated with each other (Figure 3E). Eosinophil and mast cell fractions were intercorrelated (Figure 3F), suggesting coordinated Th2–hormonal modulation within the peritumoral lymphoid microenvironment.

Immunohistochemistry confirmed spatially distinct eosinophil and mast cell infiltration patterns.

EMBP staining revealed accumulation of eosinophils in TTRs and LFRs of non-metastatic HPV-positive tumors but not in metastatic or HPV-negative cases (Figure 4A). Mast cell tryptase staining showed an apparent increase in mast-cell density in HPV-positive non-metastatic tumors, although this observation was qualitative and not quantitatively assessed (Figure 4B).

These observations are consistent with transcriptomic data, underscoring the involvement of eosinophils and mast cells in the Th2-polarized immune microenvironment that may restrain lymphatic metastasis in HPV-positive OPCs.