Cell type diversity and cuproptosis specificity in HNSCC

Cuproptosis is characterized by lipid peroxidation production, excessive accumulation of copper ions, and abnormal expression of cuproptosis-related genes. We measured the intracellular reactive oxygen species (ROS) by DCFH-DA fluorescent probe in HNSCC samples and found that the ROS content was much higher in HNSCC tissue samples than in HC tissue samples (Fig. 1a). Additionally, rubeanic acid copper staining indicated that HNSCC tissues remained similar copper ion content with HC tissue (Fig. 1b). On the contrary, immunofluorescence results showed that the expression levels of the anti-cuproptosis protein lipoic acid synthetase (LIAS) and ferredoxin 1 (FDX1) were upregulated in HNSCC tissue sample (Fig. 1c and Supplementary Fig. S1a). These findings supported the hypothesis that the abnormal expression of cuproptosis-related genes was the most important factor affecting cuproptosis in HNSCC.

Fig. 1

Cuproptosis-related lncRNAs (CRLs) and major cell types associated with HNSCC. a DCFH-DA fluorescence image for HNSCC clinical samples. n = 7 per group. Quantitative analysis of intensity is also shown. 40×, scale bar = 50 μm. b Rubeanic acid copper staining for HNSCC clinical samples. 40×, scale bar = 50 μm. c Immunofluorescence images for the localization and the expression of LIAS and FDX1 in HNSCC clinical samples. n = 7 per group. Quantitative analysis of intensity is also shown. 40×, scale bar = 50 μm. d UMAP representation of major cell types identified by scRNA-seq in HC (left) and HNSCC (right). e Expression distribution of cuproptosis-related genes (CRGs) in major cell populations. f Univariate Cox regression analysis of 19 HNSCC-related prognostic CRLs

To further probe for the cuproptosis-related gene expression specificity in different cell types, we performed scRNA sequencing data analysis. After quality control to remove low-quality cells expressing high mitochondrial gene signatures and exclude doublets, our dataset included 93 345 core cells for subsequent analysis (Supplementary Fig. S1b–j). The core cells were classified into 4 major cell compartments by the UMAP algorithm comprised of epithelial, endothelial, fibroblast, and immune cells (Fig. 1d). Analysis of the transcriptomic signatures for the major cell types documented differential expression of subcluster-defining genes for the 4 major cell compartments (Supplementary Fig. S1h). Representation of these 4 main cell types was different between HNSCC and healthy control (HC), as the HNSCC contained expanded immune and epithelial compartments (Supplementary Fig. S1g). Therefore, we surmised that these two compartments of cells play a crucial role in the occurrence and development of HNSCC. To explore the cuproptosis specificity among different cell populations in HNSCC, we selected ten hub cuproptosis-related genes (CRGs) by the PPI database and visualized the expression levels of these hub CRGs in the identified cell populations of HNSCC. The results indicated that CRGs were significantly modulated in the epithelial cell compartment of HNSCC, but the concrete mechanism remains unclear (Fig. 1e).

TME characteristics of two HNSCC subtypes based on prognostic CRLs

A total of 118 CRLs were identified by co-expression analysis with CRGs (R > 0.4, FDR < 0.001) (Supplementary Fig. S2a). Then, 19 HNSCC-related prognostic CRLs were selected by univariate Cox regression analysis for the following study (Fig. 1f). Within the 19 prognostic CRLs, 11 CRLs were found to be significantly up- or downregulated in HNSCC cohort (logFC > 0, FDR < 0.05) (Supplementary Fig. S2b).

We conducted an unsupervised consensus analysis to understand the molecular heterogeneity of HNSCC from the perspective of CRLs. The results indicated that k = 2 was more reasonable, and all the samples were divided into two HNSCC subtypes (Subtype A and Subtype B), with less correlation between the two subtypes (Supplementary Fig. S3a). We compared the overall survival among the two subtypes of HNSCC patients via the K-M survival analysis to detect whether there was an association between the different subtypes and clinical outcomes. Overall survival of Subtype A was significantly longer than that of Subtype B (P = 0.002) (Fig. 2a).

Fig. 2

JPX is notably more concentrated within epithelial compartments. a Kaplan–Meier curves for differential survival of 2 CRL clusters. b Heat map of immune cells in 2 CRL clusters. c Immune cell-related KEGG enrichment levels comparison between 2 CRL clusters. d The illustration showed that CRLs acted as pivotal nodes in the molecular mechanism of HNSCC, by Figdraw. HC, healthy control; HNSCC, head, and neck squamous cell carcinoma. e 19 CRLs were identified as finial markers with importance > 10. f Expression of JPX is shown in UMAP. Wilcoxon signed-rank test: *P < 0.05, **P < 0.01, and ***P < 0.001

Then, we further investigated whether CRL-based HNSCC subtypes showed different immune patterns, and the results indicated that Subtype A showed a distinctly different immune cell landscape and chemokine categories than Subtype B (Fig. 2b, c and Supplementary Fig. S3d, e, g, h). Among them, suppressive immune cells such as regulatory T cells and myeloid-derived suppressor cells (MDSC) were significantly reduced in Subtype B. Next, we compared the expression of immune checkpoint molecules between two HNSCC subtypes and found that out of ten differentially expressed immune checkpoints, CD276, CD47, PDCD1LG2, PVR had a higher expression in Subtype B (Supplementary Fig. S3f). Furthermore, tumor angiogenesis and extracellular matrix-related pathways and -related genes were both more enriched in Subtype B (Supplementary Fig. S3b, c). These results revealed that the key CRLs could regulate the TME characteristics of HNSCC. The results indicated that CRLs acted as pivotal nodes in the molecular mechanism of HNSCC (Fig. 2d), but the concrete mechanism remains to be explored.

JPX regulates the malignant cell behaviors of CAL27 cells

The 11 CRLs that expand in cancer were included for the Random Forest model construction (Supplementary Fig. S3i, 4b). We found that JPX was the most decisive in both the external test set and the training set (Fig. 2e and Supplementary Fig. S4a-e). Hence, we chose JPX for subsequent UMAP projection and found that JPX is more significantly concentrated in epithelial compartments of the HNSCC atlas compared to HC (Fig. 2f and Supplementary Fig. S4f). Focusing on one of the major cell types that expand in HNSCC, we reclustered the only cells identified as epithelial cells to search for subpopulations and obtained a more refined view of this population (Fig. 3a, b and Supplementary Fig. S5a, b). The results showed that Epi 1.4 was significantly expanded in HNSCC than HC and dominated the interaction with other epithelial cell subtypes (Fig. 3c and Supplementary Fig. S5d–i). Analysis of the transcriptomic signatures for the epithelial cell subtypes documented differential expression of their cell-defining genes. The heatmap showed the expression levels of the top 5 DEGs of each cell population. The top 5 DEGs of Epi 1.4 included efferocytosis-related gene (SLC2A1) and oxidation-reduction-related genes (AKR1C2 and AKR1C3) (Supplementary Fig. S5c). The cancer cell phenotype was determined by copy number variation (CNV) analysis and keratin expression, and Epi 1.4 contained substantial malignant epithelial cells (Fig. 3d). Compared to the normal epithelial cell clusters, the malignant epithelial cell cluster showed a higher correlation with focal adhesion, glycolysis, ECM organization, and NF-κB signaling (Supplementary Fig. S6a, b). Interestingly, the overlay of JPX distribution onto epithelial cell subpopulation projections indicated that JPX was more likely to distribute in HNSCC atlas rather than HC and mainly expressed in clusters of Epi 1.4 and malignant epithelial cells (Fig. 3e and Supplementary Fig. S5j). We pursued further cytology experiments to determine whether the CRL JPX would mediate the abnormal behaviors of malignant epithelial cells. Primarily, JPX was knockdown or overexpressed in CAL27 cells with siRNA segments targeting JPX or JPX overexpression plasmid (Fig. 3f). Low expression of JPX increased cell death in CAL27 cells, as detected by live-death staining (Fig. 3g). The results of Ki-67 staining demonstrated that knockdown of JPX inhibited CAL27 cell proliferation. The Ki-67 expression level in L- group was 1.8 times lower than that in control group (Fig. 3h). On the contrary, overexpression of JPX produced the opposite result, confirming that the malignant cell behaviors of CAL27 cells could be inhibited by reducing the expression level of JPX. We then divided HNSCC samples into JPX high-expression and JPX low-expression subtypes, and survival analysis showed that JPX expression was negatively correlated with overall survival (Fig. 4a). Subgroup analysis was stratified by age (>60, ≤60), TNM stage (I-II, III-IV), Grade (I-II, III-IV), gender (male, female) and JPX expression level. An internal verification was performed by comparing the riskScore of the different subgroups using the Wilcoxon signed-rank test. JPX expression level has the most important impact on survival in both subgroups (Fig. 4b). The related differentially expressed genes of JPX are shown in Supplementary Fig. S7a. The GO enrichment analysis, KEGG pathway analysis, and GSEA analysis results indicated that JPX was associated with keratinization, inflammation, proliferation, and invasion of HNSCC (Supplementary Fig. S7b, c). Then, we draw a mutational landscape to display the differentially mutated genes and their mutation types. Proved to be the high-frequency mutated gene in JPX high-expression subgroup, NOTCH1 and CASP8 might mediate the aggressive cell behaviors induced by JPX (Supplementary Fig. 7d). Subsequently, we performed a correlation analysis between JPX and predicted IC50 drug data from the GDSC, CTRP and PRISM databases and found that JPX high-expression HNSCC subclass was more sensitive to Tozasertib, Brivanib and Axitinib. (Fig. 4c and Supplementary Fig. S7e).

Fig. 3

JPX regulates the malignant cell behaviors of CAL27 cells. a UMAP of epithelial cell populations in HC. b UMAP of epithelial cell populations in HNSCC. c Proportion plots of epithelial cell populations in HC and HNSCC. d UMAP plots comparing normal and malignant epithelial cells in HNSCC. e Expression of JPX is shown in UMAP of epithelial cell populations in HNSCC. f The relative expression levels of JPX in CAL27 cells with JPX knockdown or overexpression. g Live/dead staining of CAL27 cells with JPX knockdown or overexpression. Quantitative analysis of intensity is also shown. 20×, scale bar = 150 μm. h Ki-67 cells staining of CAL27 cells with JPX knockdown or overexpression. Quantitative analysis of intensity is also shown. 40×, scale bar = 50 μm. Con, control group; neg, empty vector group; pos, positive control group; L-, JPX knockdown group; V., vehicle group; L+, JPX overexpressed group. Wilcoxon signed-rank test: *P < 0.05, **P < 0.01, and ***P < 0.001

Fig. 4

Functional analysis of JPX. a Kaplan–Meier Curves for differential survival of JPX high-expression and JPX low-expression subgroups. b Cox regression analysis of age, stage, grade, gender, and JPX. c Drug sensitivity analysis of JPX high-expression and JPX low-expression subgroups. d JPX-based ceRNA network. e 8 miRNAs were identified as finial markers with importance > 24. f Spearman’s correlation coefficient analysis of the miRNAs and JPX. Spearson correlation analysis: *P < 0.05, **P < 0.01, and ***P < 0.001

The overexpression of miR-193b-3p relieved JPX-induced abnormal cell behaviors of CAL27 cells

We constructed a ceRNA network to further investigate how the lncRNA JPX regulated pivotal mRNA expressions by targeting miRNAs, ultimately leading to the aggressive cell behaviors in HNSCC (Fig. 4d). A total of 8 miRNAs and 15 mRNAs were included in the JPX-based ceRNA network. We used the 8 miRNAs for the Random Forest model construction and found that miR-193b-3p was the most effective miRNA in both the external test set and the training set (Fig. 4e and Supplementary Fig. S8a). The Spearman’s correlation coefficient analysis of miR-193b-3p showed that miR-193b-3p functioned relatively independently from the other miRNAs and was negatively correlated with JPX (Fig. 4f). Therefore, we further focused on the downstream mechanism of miR-193b-3p in HNSCC. Downregulation of miR-193b-3p in JPX-overexpressed CAL27 cells and upregulation of miR-193b-3p in JPX-knockdown CAL27 cells were verified by RT-qPCR, indicating that JPX was located upstream of miR-193b-3p in the regulatory network (Fig. 5a). Results of Ki-67 staining, CCK8 assay and live-death staining indicated that hyperexpressive miR-193b-3p relieved JPX overexpression-induced abnormal cell proliferation of CAL27 cells (Figs. 5b, c and 6b). We also found that hyperexpressive miR-193b-3p effectively inhibits the migration of CAL27 cells in transwell assay (Fig. 5d). Consistent with this, in wound healing assay, miR-193b-3p overexpressing caused the healing percentage decline from 45% to 22% (Fig. 5e). CAL27 cells modified with both JPX overexpression and miR-193b-3p overexpression exhibited the comparable cell behaviors as the control group (Fig. 5a–e).

Fig. 5

The overexpression of miR-193b-3p relieved JPX-induced abnormal cell behaviors of CAL27 cells. a The relative expression levels of miR-193b-3p in CAL27 cells with JPX/miR-193b-3p knockdown or overexpression. b Ki-67 cells staining of miR-193b-3p knockdown or overexpressed in CAL27 cells. Quantitative analysis of intensity is also shown. 40×, scale bar = 50 μm. c Live/dead staining of miR-193b-3p knockdown or overexpressed in CAL27 cells. Quantitative analysis of intensity is also shown. 20×, scale bar = 150 μm. d Migration assays in CAL27 cells with JPX/miR-193b-3p knockdown or overexpression. Quantitative analysis is also shown. 40×, scale bar = 50 μm. e Wound healing in CAL27 cells with JPX/miR-193b-3p knockdown or overexpression. Quantitative analysis of healing percentage is also shown. 10×, scale bar = 200 μm. Con control group, neg empty vector group, pos positive control group, L- JPX knockdown group, V. vehicle group, L+ JPX overexpressed group, NCM- miR-193b-3p inhibitor empty plasmid group, NCM+ miR-193b-3p mimic empty plasmid group, M- miR-193b-3p knockdown group, M+ miR-193b-3p overexpressed group, L+M+ JPX overexpressed and miR-193b-3p overexpressed group. Wilcoxon signed-rank test: *P < 0.05, **P < 0.01, and ***P < 0.001

Fig. 6

JPX upregulates the expression level of PLAU by inhibiting miR-193b-3p. a DEGs genes downstream of miR-193b-3p include PLAU, LAMC1, STMN1, and TGFBR3. b CCK-8 assay of CAL27 cells with JPX/miR-193b-3p knockdown or overexpression. c The relative expression levels of PLAU, LAMC1, STMN1, and TGFBR3 in CAL27 cells with JPX/miR-193b-3p knockdown or overexpression. d Immunofluorescence images for the localization and the expression of PLAU in CAL27 cells with JPX/miR-193b-3p knockdown or overexpression. 40×, scale bar = 50 μm. e Protein expression and mRNA expression of RRM2 based on CPTAC database. f Expression of PLAU is shown in UMAP of epithelial cell populations in HNSCC. g Available pathology images of PLAU localization and expression in HC and HNSCC were obtained from the HPA database. h Different expression levels of PLAU between the HC and HNSCC groups. Con control group, neg empty vector group, pos positive control group, L- JPX knockdown group, V. vehicle group, L+ JPX overexpressed group, NCM- miR-193b-3p inhibitor empty plasmid group, NCM+ miR-193b-3p mimic empty plasmid group, M- miR-193b-3p knockdown group, M+ miR-193b-3p overexpressed group, L+M+ JPX overexpressed and miR-193b-3p overexpressed group. Wilcoxon signed-rank test: *P < 0.05, **P < 0.01, and ***P < 0.001

JPX upregulates the expression level of PLAU by inhibiting miR-193b-3p

In order to determine the downstream genes of miR-193b-3p, we detected the expression level of PLAU, LAMC1, STMN1 and TGFBR3, which were contained in the ceRNA network (Fig. 6a). Among them, only PLAU showed an enhanced expression trend after JPX overexpression or miR-193b-3p knockdown, and an inhibited expression trend after JPX knockdown or miR-193b-3p overexpression in line with our previous prediction (Fig. 6c). Consistently, immunofluorescence results also indicated that the expression level of PLAU is downregulated in CAL27 cells after JPX knockdown or miR-193b-3p overexpression (Fig. 6d and Supplementary Fig. S8b). In addition, the RT-qPCR results confirmed that the difference in expression of PLAU in HNSCC and HC was the most significant, compared to LAMC1, STMN1, and TGFBR3 (Fig. 8a and Supplementary Fig. S9b). This may imply that PLAU was the critical functional target of JPX in malignant epithelial cells. The mRNA expression and protein expression of PLAU were higher in HNSCC than in HC, based on the CPTAC database (Fig. 6e). Besides, PLAU expression was higher in HNSCC in the external test set and the training set from TCGA-HNSCC cohort (Fig. 6h and Supplementary Fig. S8c). Analysis of immunohistochemistry based on the online platform HPA (https://www.proteinatlas.org/) also suggested that PLAU was highly expressed in HNSCC malignant tissue samples (Fig. 6g). Based on PLAU expression in our scRNA-seq dataset, we found that PLAU was mainly expressed in the Epi 1.4 and malignant epithelial cell cluster (Fig. 6f and Supplementary Fig. S8d). Subsequently, we integrated the spatial transcriptome data with the scRNA-seq dataset. We mapped the five types of cells and PLAU to HNSCC tissue sections, which showed that the PLAU mapping range was essentially the same as that of the epithelial cell mapping range, suggesting that PLAU was predominantly expressed in the epithelial cell, in agreement with our previous prediction (Fig. 7a, b and Supplementary Fig. S8e, f). K-M survival analysis showed that patients with high PLAU expression had a poorer clinical prognosis (Supplementary Fig. S9a). Additionally, patients with higher T stages of HNSCC had higher PLAU expression (Fig. 7c). So, we came to the conclusion that abnormal activation of JPX/miR-193b-3p/PLAU signaling axis resulted in the aberrant cell proliferation, migration, and invasion of the malignant epithelial cells in HNSCC (Fig. 7d).

Fig. 7

JPX/miR-193b-3p/PLAU signaling axis entwined with malignant cell behaviors and drug resistance in HNSCC. a, b The spatial distribution of epithelial cells and PLAU. c The expression levels of PLAU with T-stage. Kruskal−Wallis test: *P < 0.05, **P < 0.01, and ***P < 0.001. d Schematic diagram of the mechanism by which the JPX/miR-193b-3p/PLAU axis mediates HNSCC. e Immunofluorescence images for the localization and the expression of Ki-67 and PLAU in HNSCC clinical samples. n = 7 per group. 40×, scale bar = 50 μm. f Immunohistochemical images for the localization and the expression of PLAU in HNSCC clinical samples. n = 7 per group. Quantitative analysis of positive areas is also shown. 10×, scale bar = 200 μm. 40×, scale bar = 50 μm. Wilcoxon signed-rank test: *P < 0.05, **P < 0.01, and ***P < 0.001

For further confirmation, we performed Ki-67 staining in clinical tissue samples, and the results suggested that the cell proliferation in HNSCC tissue samples was 7.6 times more vigorous than that in HC tissue samples (Fig. 7e and Supplementary Fig. S9d). Meanwhile, the expression level of PLAU in tissue samples was examined by immunofluorescence and immunohistochemistry. In line with expectations, PLAU was highly expressed in all the HNSCC samples tested (Fig. 7e, f and Supplementary Fig. S9d). Additionally, the results of RT-qPCR showed that JPX and PLAU were 35.4 and 184.5 times, respectively, more highly expressed in HNSCC tissue samples than HC, while miR-193b-3p expressed 2.74 times lower (Fig. 8a). Expression of JPX, miR-193b-3p and PLAU, and clinical survival status were described using a Sankey-diagram. The results reconfirmed that JPX high-expression led to a higher proportion of miR-193b-3p low expression and PLAU high-expression, and the PLAU high-expression samples higher proportion of deaths (Fig. 8b). We draw a mutational landscape to display the differentially mutated genes and their mutation types. NSD1, a tumor-related gene encoding methyltransferase, was observed to be the high-frequency mutated gene in PLAU high-expression subgroup, while TP53, a common suppressor gene, was found to be the high-frequency mutated gene in PLAU low-expression subgroup (Fig. 8c). Subsequently, we performed a correlation analysis between PLAU and predicted IC50 drug data from the GDSC, CTRP and PRISM databases and found that PLAU expression significantly enhanced the chemosensitivity of Dasatinib, Midostaurin and Palbociclib in HNSCC (Fig. 8d and Supplementary Fig. S10a).

Fig. 8

Drug sensitivity analysis of PLAU. a The relative expression levels of JPX, miR-193b-3p, and PLAU in HNSCC clinical samples. n = 7 per group. b Sankey diagram showing JPX, miR-193b-3p, PLAU expression, and clinicopathologic characteristics. c Somatic mutation characteristics of PLAU high-expression and PLAU low-expression subgroups in HNSCC. d Drug sensitivity analysis of PLAU high-expression and PLAU low-expression subgroups. Wilcoxon signed-rank test: *P < 0.05, **P < 0.01, and ***P < 0.001

PLAU mediates the epithelial–immune interactome in HNSCC

Based on CellChat analysis within the 4 major cell compartments in our scRNA-seq dataset, we found that epithelial cells dominated the interaction with other cell compartments, and the largest number of interactions between epithelial cells and immune cells was noted in HNSCC (Fig. 9a and Supplementary Fig. S1j). From Fig. 9b, the enrichment of GO analysis of malignant epithelial cell clusters of epithelial cells listed antigen processing and presentation of peptide antigen, regulation of immune effector process, cellular response to copper ion, positive regulation of leukocyte activation, positive regulation of cytokine production and leukocyte cell–cell adhesion. This would indicate that epithelial cells shift to an inflammatory state in HNSCC, implying an interaction between malignant epithelial cells and immune cells in spite of the unclear main function bearer of this process. GO enrichment analysis, KEGG pathway analysis, and GSEA analysis results indicated that PLAU was associated with aggressive tumor behavior and tumor immune microenvironment (Fig. 9c, d and Supplementary Fig. S11). Key genes related to immune function, including MMP14, ITGA6, IL-1A, TGFB1, CSF2, etc., were involved in the PLAU downstream pathways interaction network (Supplementary Fig. S9c). PLAU was positively correlated with the abundance of several immune cells analyzed (Fig. 9e). Therefore, it is speculated that PLAU may mediate the functional connection between malignant epithelial cells and immune cells.

Fig. 9

Cellchat analysis in HNSCC. a Strength of ligand-receptor interactions between cell population pairs of HNSCC based on CellChat analysis. Edge width is proportional to the number of ligand-receptor pairs. Circle sizes are proportional to the number of cells per cluster. b Chord diagrams of pathway enrichment. c Dot plots of GO enrichments associated with PLAU in HNSCC. d Dot plots of KEGG pathways associated with PLAU in HNSCC. e The abundance of immune cells varied a lot between PLAU high-expression and PLAU low- expression subgroups

Among the cells identified in the atlas of HNSCC, the immune cell cluster formed the largest population. We extracted immune cells and subdivided them into eight subcategories: B/plasma, CD4 + T, CD8 + T, DC, macrophages, mast, monocytes, and Tregs cells (Fig. 10a, b and Supplementary Fig. S12a–c). We found that the receptor of PLAU, named PLAUR, was expressed in those immune cell clusters (especially in monocyte/macrophage clusters), suggesting that epithelial cells utilize intercellular signaling to drive immune cell recruitment via PLAU-PLAUR interaction in HNSCC (Fig. 10c and Supplementary Fig. 12d). The results of RT-qPCR and immunofluorescence in clinical tissue samples confirmed that PLAUR was 1.8 times more highly expressed in HNSCC tissue samples than HC and was mainly expressed in macrophages (Fig. 10d, e). Consistent with the predicted results, the mRNA and protein expression of PLAUR were higher in HNSCC patients, suggesting that high PLAU-PLAUR expression may mediate malignant cellular behaviors, leading to a worse clinical outcome in HNSCC patients (Supplementary Fig. S12e–g). Moreover, the results of the AUCell analysis found that multiple tumor-related signaling pathways were enriched within the expressed genes for monocytes and macrophages (Supplementary Fig. S13). Whereafter, CellChat was performed to analyze intercell communication within the immune cells and noted that macrophages/monocytes dominated the interaction with other immune cells (Fig. 11a, b). This intercell communication might be mediated mainly by SPP1 regulation (Fig. 11c, d). SPP1, also known as osteopontin, is upregulated in various cancers and involved in various biological processes, such as inflammation, fibrosis, cancer progression, and metastasis.25 Our result indicated that macrophages were the main source of SPP1 and various kinds of immune cells, including B cells, CD4 + T cells, CD8 + T cells, mast cells, and neutrophils, could serve as the receiver or the influencer of SPP1 (Fig. 11c, d and Supplementary Fig. S14a). SPP1 plays an important role among immune cells. The mRNA and protein expression of SPP1 were elevated in patients with HNSCC, and patients with high expression of SPP1 have a worse clinical prognosis (Supplementary Fig. S14e–g). Of all the SPP1 ligands, SPP1–CD44 interaction most commonly exists in immune cells, especially between macrophages and neutrophils/mast cells (Fig. 11e and Supplementary Fig. S14a–d). Collectively, these data suggest that the overexpressed PLAU in tumor epithelia could interact with its receptor PLAUR located on the surface of immune cells (especially macrophages), to influence the interaction mode of aberrant epithelial cells and immune cells in HNSCC (Fig. 11f).

Fig. 10

PLAU mediates the epithelial–immune interactome in HNSCC. a UMAP of immune cell populations in HNSCC. b Proportion plots of immune cell populations in HC and HNSCC. c Expression of PLAUR in UMAP of immune cell populations in HNSCC. d The relative gene expression levels of PLAUR in HNSCC clinical samples. n = 7 per group. e Immunofluorescence images of PLAUR (green) and CD68 (red, characterizing macrophages) in HNSCC clinical samples. n = 7 per group. Quantitative analysis of intensity is also shown. 20×, scale bar = 150 μm. Wilcoxon signed-rank test: *P < 0.05, **P < 0.01, and ***P < 0.001

Fig. 11

The intercell communication within the immune cells is mediated by SPP1 regulation. a Number of ligand-receptor interactions between immune cell population pairs of HNSCC based on CellChat analysis. b Strength of ligand-receptor interactions between immune cell population pairs of HNSCC based on CellChat analysis. c Bubble plots of the significant differentially expressed ligand-receptor pairs of HNSCC immune cell types. The dot color reflects communication probabilities, and the dot size represents computed P-values. Empty space means the communication probability is zero. d Heatmap of CellChat analysis depicting dominant cell types involved in SPP1 signaling. e Violin plot showing expression of SPP1 receptor by cell type from the scRNA-seq data. f Schematic representation of the interaction mode of aberrant epithelial cells and immune cells in HNSCC

In conclusion, this study confirms that Cuproptosis influences the malignant cell behavior and epithelial–immune interaction in HNSCC. We integrated scRNA-seq analysis of 6 HNSCC tissues and bulk genomic information of 502 HNSCC tissues to comprehensively assess the underlying cellular distribution of CRL and its correlation with major non-immune and immune features in HNSCC. The key CRLs could regulate the TME characteristics of HNSCC. Using a random forest (RF) model, we selected the key lncRNA JPX among these CRLs and implemented a ceRNA network. The upregulation of the JPX/miR-193b-3p/PLAU axis in malignant epithelial cells promoted cell proliferation, migration, and invasion in HNSCC. PLAU overexpressed in tumor epithelia could increase its binding with the receptor PLAUR, expressed mainly on macrophages, to ultimately influence the aberrant epithelial–immune interactome in HNSCC. We propose the combination of JPX and its downstream gene urokinase-type plasminogen activator (PLAU) inhibitors as a novel therapeutic strategy to target HNSCC and provide new insights into precision oncology (Supplementary Fig. S15).