Single-cell analysis revealed the complexity of AM
In order to explore the tumor heterogeneity and immune environment at the single-cell level, we collected four tissues from two AM patients with clinical stage IIC T4bN0M0, including SM patient, SM adjacent normal tissues (SMP), PM patient and PM adjacent normal tissues (PMP) (Fig. 1A-B). For data analysis, PM1 stands for the merging of PM and PMP, and SM1 stands for the merging of SM and SMP. As a result, 24,789 individual cells from four samples were passed quality control criteria (Supplementary Table 1) and were divided into 25 major clusters (Fig. 1C). Twelve clusters were explored with unbiased clustering across all cells by Principal Components Analysis (PCA) and visualized by Uniform Manifold Approximation and Projection (UMAP) (Fig. 1D). We annotated the cell type of each cluster with the well-established gene markers, including B cells, plasma cells, endothelials, epithelials, keratinocytes, fibroblasts, vascular smooth muscle cells, myeloid cells, T/NK cells, neutrophils, mast cells and melanocytes (Fig. 1E). The proportion of the 12 cell populations varied among the samples, with more fibroblasts and fewer T/NK cells in PM1 and more melanocytes and keratinocytes in SM1; however, cell types such as B-cells and plasma cells were rare in all four samples (Fig. 1F and Supplementary Table 2).
Single-cell analysis revealed the intertumor heterogeneity of melanocytes
We clustered the melanocyte cells and revealed 5 distinct subpopulations by visualization using UMAP. The cell types were defined based on the specifically expressed genes in each cluster, including CDK4, GAB2, DMKN, IGFBP5 and PTPRC (Fig. 2A). Melanocytes were predominantly found in two tumor tissues, where CDK4 subpopulation was predominantly expressed in SM and GAB2 subpopulation was predominantly expressed in PM (Fig. 2B). It is showed that the distribution of cell clusters from SM and PM were in a high degree of intertumor heterogeneity. Analysis of DEGs showed that PM highly expressed SERPINF1, and SM highly expressed CDK4, COL11A1, and S100A4. Although PM and SM expressed different genes, all of these genes were epithelial mesenchymal transition (EMT) characteristic. (Fig. 2C).
Malignant cells were identified by inferring large-scale CNVs with immune and stromal cells as references (Fig. 2D). We found that SM has higher CNV signal than PM(Fig. 2E). Among the five subpopulations of melanocytes, the CDK4 subpopulation had the highest CNV signal, indicating the highest degree of malignancy (Fig. 2F). Amplification of chromosome 7, 12, 17 and deletion of chromosome 6, 9, 10, 14 were detected in SM. Amplification of chromosome 6, 8, 11 and deletion of chromosome 13 were detected in PM (Fig. 2G). Interestingly, we found that the CDK4 is located on chromosome 12, and both GAB2 and PAK1 are located on chromosome 11. Similarly, the highly expressed pathways in GSVA in SM are quite different from the PM. CDK4 subpopulation is primarily located in SM, highly expressing the NF-kb pathway, whereas the GAB2 subpopulation, which is primarily located in PM, predominantly expresses the Wnt pathway (Fig. 2H).
CAFs were enriched in the AM tumors
The tumor microenvironment (TME) is a complex ecosystem composed of various cell types including malignant and stromal cells (Whiteside 2008). Fibroblasts are common cells of the connective tissue and can be activated in cancer development, which are commonly known as cancer-associated fibroblasts (CAFs). CAFs play an important role in TEM that can interact with cancer cells to promote tumor metastasis and progression (Schoepp et al. 2017). The CAFs gene marker fibroblast activation protein (FAP) was used to distinguish the normal fibroblasts (NFs) and CAFs, and six subpopulations NF-LRP1B, NF-PDE4C, CAF-DSP, CAF-CD70, CAF-NPY1R and CAF-GPM6B were identified (Fig. 3A). Most of the fibroblasts subpopulations are present in the PMP including CAF-CD70, CAF-NPY1R and NFs (Fig. 3B-C).
We performed cellchat analysis to explore the interaction between CAFs and malignant cells. The results showed that the number and strength of fibroblast-melanocytes interactions were higher in PM1, in which the CAF-NPY1R subpopulation was the most abundant (Fig. 3D-F). We found that EPHA3 is expressed only in CAF-NPY1R subpopulation. The ligand EFNA1 is higer expressed than EFNA4 in malignant cells (Fig. 3G and Supplementary Table 3).
For exploring the evolution of EPHA3 in six fibroblast subpopulations, we performed pseudo-time analysis. It showed that above six subpopulations were in different developmental states. As a starting point of NFs, CAFs in PM1 and SM1 differentiate in two directions, subpopulation CAF-NPY1R/CD70 and CAF-DSP were mainly at the two roots of phylogenetic tree (Fig. 3H). The color of CAF-DSP subpopulation which is predominantly present in SM1 is lighter, indicating later differentiation (Fig. 3I). Summarize genes to form 4 modules based on differences in temporal developmental trajectories, and EPHA3 was located in module 3 (Supplementary Table 4). The significance of expression is that EPHA3 expression is elevated with the evolution of fibroblasts (Fig. 3J-K). We verified that the ligand EFNA1 was highly expressed in AM using immunofluorescence staining (Fig. 3L). It suggests that blocking the EPHA3-EFNA1 axis may affect the interaction of CAF with malignant cells and could be an effective therapeutic strategy for PM.
Distinct T/NK subpopulations in AM ecosystem
In tumor immunity, tumor cells act as antigens while immune cells and leukocytes infiltrates the tumor tissue function through chemotaxis for immune defense (Angell and Galon 2013). Immune escape also is an important factor in tumorigenesis. According to gene expression, T/NK cells are divided into 8 subpopulations, namely NK, gdT, Treg, CD8 + Cytotoxic, TNF + CD4+, CD27 + CD4+, CD4 + Naive and CD4 + Memory (Fig. 4A). IL2RA, FOXP3, IKZF2, co-stimulatory (CD28, TNFRSF9 and ICOS) and inhibitory markers (TIGIT, CTLA4 and LAYN) were highly expressed in the Treg subpopulation. CD4-Naïve T cells were marked with expression of CCR7, LEF1, SELL and TCF7 genes. CD4-Memory T cells were featured with high expression of LTB, GPR183 and PASK. CD8-Cytotoxic were characterized with high expression of GZMK, GZMA, IFNG and NKG7. The NK subpopulation was marked with high expression of NKG7, TYROBP, KLRD1 and KLRF1 (Fig. 4B). We found that SM had a stronger immunosuppressive status compared with PM. Unlike fibroblasts, tumor tissues had more T cells infiltration than adjacent normal tissues (Fig. 4C-D). SM1 lacked NK cells, which may be related to the poor prognosis of SM patients.
The TIGIT-NECTIN2 axis was enriched in the interplay between T/NK cells and melanocytes
To explore the interactions between melanocytes and T/NK cells, we conducted cellchat analyses based on ligand-receptor pairs. In cellchat, the number of melanocytes and T/NK cells cellular interactions was greater in PM1, whereas the strength of the effect of SM1 was greater (Fig. 5A). As seen in the information flow, LAMININ, FN1, MK, MIF, MHC-1 and COLLAGEN pathways were highly expressed both in PM1 and SM1. ADGRE5 and CLEC pathways were significantly increased in PM1. CD99 and APP pathways were more expressed in SM1. The differently expressed pathways could provide clues for the therapeutic strategies in PM and SM (Fig. 5B). In addition, the TIGIT/NECTIN2/NECTIN3/PVR axis was enriched in the interactions between T/NK cells and melanocytes, which was a direct cell-cell interaction (Fig. 5C and Supplementary Table 5). TIGIT was predominantly expressed in Treg and CD8-Cytotoxic of PM1, and in NK, Treg and CD8-Cytotoxic of SM1 (Fig. 5D-E). Here we found that TIGIT expression was significantly higher on CD8 + T cells in PM subpopulations than on those in PMP subpopulations. While TIGIT expression in NK cells in PM subpopulations was not significantly different from PMP subpopulations. TIGIT ligands have NECTIN2, NECTIN3, and PVR in PM1 and only NECTIN2 in SM1, in which the TIGIT/NECTIN2 pair is expressed more strongly, which differs from previous TIGIT/PVR expression (Fig. 5F).
In T/NK pseudo-time analysis, naive was used as the starting point to make the proposed time trajectory plot, with darker to lighter colors indicate earlier and later differentiation. It was showed that TNF + CD4 + and CD27 + CD4 + were located at the terminal ends of the 2 branches (Fig. 5G-H). Pooling T cell expressed genes according to differences in temporal developmental trajectories, TIGIT is located in module 4 (Fig. 5I-J and Supplementary Table 6). TIGIT expression was elevated and then decreased in the proposed temporal differentiation, consistent with the aforementioned enrichment in Treg, NK and CD8-Cytotoxic, which is not expressed in CD4 + T in our results. The expression of TIGIT in AM was validated by immunofluorescence staning (Fig. 5K).
Downregulation of CD226 in Treg and CD8 + Cytotoxic cells
CD226 is an inhibitory receptor and competes the same ligands with TIGIT (Worboys et al. 2023). In the interactions between melanocytes and T-NK cells, CD226 is differentially expressed in each subpopulations in AM. In PM1, CD226 was expressed in CD27 + CD4 + and CD4 + Memory subpopulations, and corresponding ligands in tumor cells were TIGB2, NECTIN2 and PVR. NECTIN2 was expressed in a highest intensity. In SM1, CD226 was expressed in NK cells, and the corresponding ligand in tumor cells was only NECTIN2 (Fig. 6A-C). NECTIN2 were highly expressed in AM, but CD226 was down regulated by immunofluorescence staining (Fig. 6D). We found that CD226 was not expressed in Treg and CD8 + Cytotoxic cells both in PM1 and SM1, and it has reported that the activation receptor CD226 is a key component of T cell biology, and its absence impairs the responsiveness of CD 8 + T cells to TCR stimulation (Weulersse et al. 2020).
Beyond that, Eomes is a transcription factor (TF) with a key role in CD8 + T cell differentiation, by binding to the promoter of TIGIT, upregulating its expression. Overexpression of Eomes and losing of CD226 are related. In our SCENIC results, Eomes was enriched in CD8-Cytotoxic (Fig. 6E). In conclusion, CD226 was barely expressed in Treg and CD8 + Cytotoxic cells, while TIGIT was highly expressed in Treg, CD8 + Cytotoxic cells and NK cells (SM1), suggesting that the TIGIT-NECTIN2 axis plays an important role in the immune environment of AM (Fig. 6F).