Single-cell resolution profiling of the immune microenvironment in primary and metastatic nasopharyngeal carcinoma

Abstract

Background

Nasopharyngeal carcinoma (NPC) is an assertive malignancy with partially understood underlying mechanisms, urging further study into its diverse and dynamic tumor microenvironment (TME) to bolster diagnosis, treatment, and prognostic accuracy.

Aims

To track the evolutionary route of metastasis, here we perform a yielding scRNA-seq data from 24 primary carcinoma, 7 peripheral blood mononuclear cell (PBMC) nasopharyngeal carcinoma, and 7 metastatic carcinoma patients’ sample.

Materials and methods

Following high quality control and filtration, a total of 292,298 cells from these tumors were classified into 10 clusters: T cells, B cells, Macrophages/Monocytes, Natural Killer (NK) cells, Plasma cells, plasmacytoid Dendritic Cells, Migratory Dendritic Cells, Mast cells, Cancer-Associated Fibroblasts, and Epithelial cells.

Results

By comparing and analyzing the different functional capacities of cellular entities within primary and metastatic nasopharyngeal carcinoma, coupled with a detailed investigation into the heterogeneity and differential fate trajectories of T cells, B cells, and myeloid cells, as well as assessing the interactions of cell–cell communicative heterogeneity between these carcinogenic states, we established single-cell atlases for primary and metastatic tumors and identified a large number of potential therapeutic targets.

Conclusion

This comprehensive analysis significantly advances our understanding of nasopharyngeal carcinoma (NPC) metastasis by detailing the evolutionary dynamics and the impact of the tumor microenvironment at a single-cell resolution, thereby laying a crucial foundation for future metastatic tumor research and providing new insights into immune heterogeneity, molecular interactions, and potential therapeutic strategies for NPC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00432-024-05900-5.

Introduction

Nasopharyngeal carcinoma (NPC) is a type of cancer that originates from the epithelial cells of the mucosal lining in the nasopharynx. The tumor is commonly found in the pharyngeal recess, also known as the fossa of Rosenmüller, in the nasopharynx (Chen et al. 2019). NPC is most commonly found in southeastern Asia, particularly among the Cantonese people in Guangdong, China, where it has the greatest recorded occurrence (Chang et al. 2021; Chua et al. 2016). Although nasopharyngeal carcinoma and other epithelial head and neck tumors originate from comparable cell or tissue lineages, they exhibit significant differences. Epstein-Barr virus (EBV) infection is strongly associated with the development of NPC. Pathologically, NPC is characterized by the presence of a large number of immune cells both surrounding and inside the tumor lesions. This indicates the presence of a highly intricate tumor microenvironment (TME) in NPC (Lin et al. 2023; Lv et al. 2023; Young et al. 2016). One important aspect of NPC is the role of the immune system in causing disease. The immunosuppressive microenvironment creates a favorable condition for tumor cells to grow (Gong et al. 2021a). More than 70% of patients are diagnosed with intermediate to advanced disease. This immune contribution is strongly associated with epithelial cancers, as well as malignant lymphoid and natural killer/T cell lymphomas (Li et al. 2022). More than 90% of individuals diagnosed with undifferentiated NPC have been found to be infected with the EBV. Significant advancements have been achieved in comprehending the molecular mechanisms of NPC and formulating therapeutic strategies in the past several decades. Radiotherapy and chemotherapy are the primary therapeutic modalities for NPC. However, their effectiveness is restricted in individuals with locally advanced or distant metastatic tumors (Chen et al. 2021; Sidaway 2020).

TME is an intricate milieu comprised mostly of tumor cells, immune cells in close proximity to the tumor, tumor-associated fibroblasts, vascular endothelial cells, and other components (Lu et al. 2022). Cellular contacts facilitate the ability of tumor cells to avoid detection by the immune system, and serve as the cellular process responsible for tumor advancement and spread to other parts of the body. There is increasing evidence that certain types of immune cells, such as macrophages, neutrophils, natural killer cells, dendritic cells, and bone marrow-derived suppressor cells, as well as T cells and B cells, play crucial roles in the TME. These immune cells are involved in the development and progression of cancer (Lei et al. 2020; Rajbhandary et al. 2023). Generally, the goal of cancer immunotherapy is to train host immune cells in lymphoid tissue and antitumor immune cells in the TME to recognize and destroy tumor cells (Tan et al. 2020). In addition, the process of immunotherapy-induced antitumor immune response has the potential to enhance systemic immune monitoring, which in turn could eradicate both local and distant metastases (Zhang et al. 2020). Furthermore, immunotherapy has the potential to modulate immune protection, avoid tumor recurrence, and develop long-term immunological memory (Peng et al. 2022). The NPC microenvironment still lacks adequate identification and characterization of other lymphocytes, myeloid cells, and fibroblasts that reside in the TME. In NPC, for instance, little is known about the phenotype and function of invading B cells, despite the fact that new data suggests that B cells are a key component in the success of immunotherapy for a number of cancers (Helmink et al. 2020).

Hence, to comprehensively decipher the complexity of the stromal infiltration and reveal NPC-specific features, we collected paired primary and metastatic tumour samples from several previous trials (GSE150825, GSE162025, HRA000036) in de novo metastatic NPC and conducted integrative genomic and transcriptomic sequencing, including scRNA-seq, to trace the evolutionary history of distant metastases in NPC (Gong et al. 2021a; Lin et al. 2023; Liu et al. 2021). We reasoned that through integrative analysis of the large-scale bulk and single-cell sequencing data of paired primary and metastatic NPC tumours, a finer understanding of the evolutionary history of NPC metastases and the mechanism underpinning the effect of locoregional treatment on metastatic NPC could be obtained.

Methods

Data availability

The single-cell data from 10 nasopharyngeal carcinoma (NPC) primary tumor-blood pairs, sourced from the published dataset in the Gene Expression Omnibus (GEO) under accession number GSE162025, involves individuals aged between 22 to 65 years. Additionally, metastatic tumor samples were acquired from the dataset under accession number HRA000036, which includes samples from two patients, one male and one female, each with a primary tumor, regional lymph nodes, and distant lymph node metastases. The tumor metastasis sites are specified as left and right regional lymph nodes, inguinal lymph nodes, and left and right axillary lymph nodes. (Supplementary Table 1).

Single cell RNA-seq data processing

Paired-end sequencing reads in FASTQ format underwent processing via the Cell Ranger pipelines (version 7.0.1, 10 × Genomics), employing standard parameters for the demultiplexing of samples, mapping of reads to human genome (GRCh38), barcode analysis, quantification of single-cell transcripts, and the generation of expression matrices. Consequently, this procedure culminated in the acquisition of refined single-cell matrix data for downstream analyses.

Dimensionality reduction and clustering

The outputs, initially pre-processed by Cell Ranger (version 3.1.0), were subsequently imported into the Seurat package (version 5.0.2) for high quality control and advanced downstream analysis. This process entailed the exclusion of cells of low quality, characterized by fewer than 200 genes per cell, fewer than three cells per gene, Unique Molecular Identifiers (UMIs) per cell ranging from less than 500 to under 7500, a mitochondrial gene composition exceeding 10%, and a ribosomal gene composition surpassing 15%. The sctransform (SCT) assay facilitated normalization, the identification of the top 2000 highly variable genes, scaling, and dimensionality reduction. Peripheral Blood Mononuclear Cells (PBMCs), Primary, and Metastatic Nasopharyngeal Carcinoma were intergrated using the harmony package (version 0.1). Principal Component Analysis (PCA) was applied to refine the dataset, with the leading 30 PCs being harnessed for subsequent analytical processes. Upon the construction of a shared nearest neighbor graph, clustering was executed via the Louvain Method, followed by a two-dimensional visualization employing Uniform Manifold Approximation and Projection (UMAP). In order to pinpoint marker genes for each cluster, a Wilcoxon rank sum test was employed to ascertain differentially expressed genes (DEGs) in comparison to other clusters. Finally, cell types were delineated based on the expression of cancer cell markers, adhering to classifications from a preceding study.

Analysis of variance in gene expression across diverse populations

The FindMarkers function within the Seurat package serves to identify differential expression markers among distinct cell clusters in single-cell RNA sequencing datasets, focusing on genes present in a minimum of 25% of the cells within a cluster and displaying an expression level change of at least 0.25 on the log fold change (logFC) threshold. Leveraging the group_by and top_n functions of the dplyr package, the top 10 upregulated genes are earmarked for the visualization of textual markers. Ultimately, ggplot2 is employed to graphically represent these findings.

GO and reactome analyses

Leveraging the FindMarkers() function in conjunction with the group_by and top_n methods, we delineated the top 50 differentially expressed genes (DEGs) within each subgroup. Subsequently, we evaluated the functional discrepancies among various groups employing the compareCluster function within the clusterProfiler package. This analysis particularly focused on enriching Biological Processes (BP) within the Gene Ontology (GO) and Reactome databases. To visually present the outcomes of these enrichment analyses, the dotplot function was utilized.

RNA velocity analyses

To delineate the cellular fate differentiation in both in primary and metastatic nasopharyngeal carcinoma. The spliced and unspliced reads was conducted by using the velocyto Python package (version 0.17.17), based on previously aligned BAM files from scRNA-seq data. This analysis, which included the calculation of RNA velocity for each gene in each cell and the projection of RNA velocity vectors into a low-dimensional space, was performed using the scVelo Python pipeline (version 0.3.2). Following default parameters and preprocessing steps outlined in the scVelo package, we engaged in dynamical modeling for cellular trajectory inference through RNA velocity, offering a refined generalization of the original method to identify various transcriptional states. Latent time, pseudotime analyses, and velocity confidence levels were determined using standard parameters, with extrapolated states subsequently mapped onto the UMAP embedding generated in the initial phase of analysis. Clustering and visualization for identified stromal clusters were reiterated with identical parameters to emphasize the continuum. Preprocessing steps for velocity analysis included detection of minimum count thresholds, filtering, and normalization, executed via scv.pp.filter_and_normalize and scv.pp.moments functions. Gene-specific velocities were calculated in stochastic mode using scv.tl.velocity and depicted through scv.pl.velocity_embedding. Additionally, the scv.tl.latent_time function was employed to deduce a collective latent time from splicing dynamics, with relevant genes plotted along a temporal axis according to dynamic expression patterns using the scv.pl.heatmap function.

TF network analyses

To elucidate the transcriptional regulation mechanisms of metastatic tumors at the cellular level, the single-cell RNA-seq dataset was initially preprocessed using the R programming language and Seurat libraries. Subsequently, the data were transposed and exported in TSV format, facilitating further analysis. The activity score of each transcription factor (TF) regulatory module was computed using the pySCENIC Python package (version 0.12.1). Following this, a gene regulatory network (GRN) was constructed employing grn command, leveraging the derived expression matrix and a predefined list of transcription factors. This step was succeeded by the utilization of the ctx command in conjunction with the cisTarget database to identify regulons and evaluate transcription factor activity. After these processes, the aucell command was executed to calculate the regulator activity score for each cell, with the results being stored as a CSV file. Subsequently, within the R environment and utilizing libraries such as Seurat, ComplexHeatmap, and pheatmap, heatmaps were generated to illustrate the variances in regulator activities across different cell types, derived from the regulator activity scores.

Cell–cell interaction analysis

To conduct a comprehensive analysis of cell communication in primary and metastatic tumor samples, the Cellphonedb Python package (version v5.0.0) was employed. This database encompasses 2,912 cell-to-cell ligand-receptor interactions, significantly enhancing the scope and precision of the study. Following this, Cellphonedb was utilized to perform statistical analysis on preprocessed and normalized single-cell RNA-seq data. The findings were graphically represented using dot plots and heatmaps, unveiling intricate communication networks among various cell types and their evolutionary dynamics across tumor stages.Next, CellChat (version 2.1.0) was used to generate circular network diagrams, illustrating the intricate patterns of intercellular interactions and quantifying the influence of specific ligand-receptor pairs to delve deeper into the cellular communication framework. Furthermore, using the Seurat package alongside ggplot2 package, a comparative analysis of cell communication in primary versus metastatic tumor cells was undertaken. This investigation focused particularly on the roles of chemokines, co-inhibitory, and co-stimulatory activities, highlighting their differential expression and impact across varied tumor microenvironments.

Result

Profiles of nasopharyngeal carcinoma cells in 10 distinct clusters identified through scRNA-seq analysis

To generate a deep transcriptional atlas of nasopharyngeal carcinomas (NPCs), high-resolution 10 × genomics scRNA-seq was performed on data from 24 primary, 7 peripheral blood mononuclear cell (PBMC), and 7 metastatic nasopharyngeal carcinoma (Fig. 1A). Following high quality control and filtration, a total of 292,298 cells from these tumors were classified into 10 clusters: T cells (177,664 cells), B cells (56,496 cells), Macrophages/Monocytes (14,506 cells), Natural Killer (NK) cells (27,534 cells), Plasma cells (6,347 cells), plasmacytoid Dendritic Cells (pDC, 1,251 cells), Migratory Dendritic Cells (805 cells), Mast cells (855 cells), Cancer-Associated Fibroblasts (CAFs, 807 cells), and Epithelial cells (6,033 cells) (Fig. 1B–C).

Fig. 1.

Dissection of the tumor microenvironment in Nasopharyngeal Carcinoma (NPC) with scRNA-seq. A Workflow diagram showing the samples from 24 primary NPC tumors, 7 PBMC and 7 metastatic tumors for scRNA-seq. B UMAP plots of cells from the 38 samples profiled in this study, with each cell color-coded to indicate its associated cell types. C Using Harmony to integrate three sets of scRNA data, primary tumors (blue), PBMC (green), and metastatic tumors (red). D A dotsplot illustrates marker genes exhibiting high expression across 10 clusters, high expression (Red), low expression (blue). E, F FeaturePlot and ViolinPlot depict the distribution and quantification of specific marker genes within each subpopulation, respectively

We used batch correction and data integration to remove batch effects across datasets. Moreover, each group was identified according to marker genes. T cells exhibited significantly higher CD3D, CD3E, CD8A, and GZMK. B cells were responsible for producing antibodies and participating in the humoral immune response, such as MS4A1, CD79A, CD79B, and CD19. Macrophages and monocytes demonstrated high levels of LYZ, CD14, S100A8, C1QA, CD163, and MRC1. Natural Killer (NK) cells expressed GNLY, NKG7, KLRD1, and PRF1. Plasma cells showed increased expression of MZB1, IGHG1, IGHA1, and TNFRSF17. Plasmacytoid dendritic cells (pDCs) were defined by IL3RA, LILRA4, and CLEC4C. Migratory dendritic cells associated with ability to migrate from peripheral tissues to secondary lymphoid organs, such as CCR7, LAMP3, EBI3, and PDCD1LG2. Mast cells were characterized by the presence of TPSAB1, CPA3, MS4A2, and TPSB2. Cancer-associated fibroblasts (CAFs) were identified by COL1A1, COL3A1, DCN, and FAP. Epithelial cells were characterized by the expression of specific markers, such as CD24, KRT19, EPCAM, and KRT18. (Fig. 1D).

We also found that PTPRC was highly expressed in immune cell-related fractions, indicating its importance in these groups. CD3D was present in the majority of T cells, whereas CD8A selectively identified a subset, highlighting the diversity of T cell functions by distinguishing CD4⁺ and CD8⁺ cells. CD19 gene distinguished B cell subtypes, and myeloid cells were detected by LYZ, revealing their immunological involvement. Due to their exclusive relationship with GNLY, NK cells were crucial to innate immunity. Plasma, pDCs, Migratory DCs, and Mast cells were specifically defined by IGHG1, IL3RA, LAMP3, and TPSAB1, highlighting their diverse and specialized functions. COL1A1 was mostly identified in fibroblasts, while EPCAM was only found in epithelial cells, demonstrating their vital involvement in tissue integrity. This marker gene distribution mapping illuminates the complex interaction and functional distinctions that distinguish healthy and pathological tissue states (Fig. 1E) and enhances our understanding of cellular architecture. Violin plots show similar patterns and distributions. (Fig. 1F).

Each groups exhibited diverse functional potentiality in different cancer types

We examined nasopharyngeal cancer primary, metastatic, and peripheral blood mononuclear cell heterogeneity in three datasets. The PBMC group had significantly more B and NK cells than the primary and metastatic groups. There were no significant differences in T cell populations. Epithelial cells were observed solely in primary and metastatic samples, with metastatic samples exhibiting a higher proportion. This study illustrates the complicated immunological and epithelial cell patterns of nasopharyngeal cancer phases and components (Fig. 2A).

Fig. 2.

Comparison of single-cell data from three groups. A The proportion of cells from three groups (primary NPC, PBMC, and metastatic NPC) among 10 populations. B A scatter plot displaying differential expression across multiple groups, with the top 10 upregulated genes marked for each group. The selection standard include |avg_log2FC|> = 1 and p_val_adj < 0.05. Upregulated genes are depicted in red, while downregulated genes are shown in green. C A violin plot show the top 10 genes with the most significant differences among primary NPC, PBMC, and metastatic NPC. D, E The dotplot illustrating the gene ontology and reactome pathway enrichment analysis conducted on the top 50 differentially expressed genes across the three groups, respectively

Volcano plot showed up-regulated and down-regulated genes across dataset, including the top10 differential genes in each group (Fig. 2B), with these same genes highlighted in the violin diagram (Fig. 2C). The upregulation of HSPA1A and HSPA1B in primary tumor tissues emphasizes their role in protein folding and cellular stress responses. EGR1 and CDKN1A (p21), which regulate cell proliferation, apoptosis, and cycle regulation, are also elevated. NR4A1, DUSP4 and GZMK show cellular differentiation, survival, and immune response, and ITM2C, CD83 and FABP5 show signaling and metabolic pathways. In PBMCs, high expression of ADGRG1, CX3CR1, S1PR5, and KLRD1 indicates substantial immune system interaction with the tumor, whereas SPON2, FGR, FAM65B, and FCGR3A indicate active immune cell signaling and response SNHG series expression increases in metastatic tumors, affecting RNA modification and cell metastatic pathways. GAS5 may reduce cancer, while SEPTIN6, PCED1B-AS1 may help cells migrate and adapt to different microenvironments. Across tumor phases, a complex regulatory network controls tumor formation, immune system, and cellular activity (Fig. 2C).

The functional enrichment analysis of the 50 most variably expressed genes reveals diverse biological activities across primary, PBMC, and metastatic tumor samples. Primary tumor immune activation requires T cell responses and apoptotic signals, therefore viral interactions may help them hide. These malignancies address protein stress and cell adhesion, suggesting ERK1/2 for proliferation, differentiation, invasiveness, and migration. Adhesion molecules and cytokines help myeloid and lymphocytes defend pathogens and detect tumors in PBMCs. Cancer-specific cytoskeletal alterations and autophagy help metastatic tumors move and adapt to stress. Survival, proliferation in new habitats, and immune evasion are promoted by these adaptations. Additionally, TGF-β pathways boost cell differentiation and metastatic properties, showing the complex link between tumor evolution and host immune responses (Fig. 2D-E).

Variation of T cells sub-populations across primary and metastatic tumors

T cells from nasopharyngeal carcinoma were isolated, re-clustered, and divided into 10 populations (Fig. 3A). All T cells in these subpopulations express CD3E. Two CD8⁺ groups, CD8_T-GZMK and CD8_T-GNLY, express GZMB, a cytotoxic effector marker, suggesting they may destroy tumor cells. CD8_T-GNLY have high level GZMK, GNLY, and PRF1 expression boosted their effector T cell status. However, CD8_T-NELL2, expressing NELL2 and CCR7, resembles naïve T cells, indicating an unstimulated CD8⁺ T cell cluster. Gene expressions show CD4⁺ T cell variety. CD4_T-SPDYA and CD4_T-FHIT subsets, expressing CCR7, may constitute naïve CD4⁺ T cells. Their functions match T cell phenotypes, with TXNIP and TOX2 involvement suggesting metabolic and transcriptional regulation. FOXP3 and IL2RA-high Tregs regulate tumor immune evasion and immunological homeostasis. Treg-RBMS3 elevated expression suggests a regulatory role, possibly necessary for tumor-mediated immunosuppression. In response to the immune response, pro-T cells with high MKI67 and TOP2A expression may proliferate rapidly (Fig. 3B, C).

Fig. 3.

T cells undergo dynamic re-clustering within NPC microenvironment. A UMAP plot displaying 10 distinct clusters of T cells, including CD4 T cells, CD8 T cells, Treg cells, and proliferating T cells, each uniquely represented by different colors. B The dotsplot illuminates the expression patterns of marker genes across 10 distinct T cell clusters. C The FeaturePlot visualizes the distribution of specific gene expression within each group, effectively illustrating how characteristic genes are expressed across different cell populations. D A violin plot visually highlights the variation in gene expression levels across groups within T cells, focusing on the top 10 genes that exhibit the most significant differences. E RNA velocity delineates the fate trajectories of T cells within primary and metastatic tumors, with purple denoting the early stages of differentiation and yellow signifying the terminal stages. F The heatmap displays the differential transcriptional regulation among T cell subpopulations. G The heatmap exhibits genes of primary and metastatic tumors as they vary along the differentiation trajectory

An analysis of primary and metastatic cancers revealed distinct biological variations in the top 10 genes that were expressed differently. The immune response and regulation of the microenvironment were primarily influenced by genes expressed in the original tumor. Tumors exhibit the expression of immunoglobulin genes IGHA1 and IGHG1, which suggests the presence of immunological responses mediated by antibodies. Elevated concentrations of CCL4, CCL4L2, CCL3, and CCL3L1 suggest that the tumor is attracting T cells and macrophages to carry out immunological surveillance. The activity of the MAPK signaling pathway is crucial for the growth and survival of cancer cells, and it may be inhibited by DUSP4. The increased expression of GZMB indicates that cytotoxic T lymphocytes specifically attack tumor cells, whereas the transcriptional regulation of BHLHE40 may modify the development of tumor cells and their biological cycles. Protein stability and signal transmission are regulated by PTMS and ubiquitination. Genes that are highly expressed in metastatic tumors enhance the migration, invasion, and ability to adapt to stress in the surrounding microenvironment. BACH2, a regulatory protein that controls gene expression, has the potential to modify the process of cell specialization and the body's defense mechanisms. Elevated levels of LEF1 and PLAC8 expression indicate the presence of tumor cells that are likely to adhere and migrate, leading to metastasis. MT1X, MT1E, and MT2A can aid metastatic cells in managing oxidative stress in unfamiliar settings by sequestering and detoxifying metal ions. Metastatic tumors employ SESN3 and HSPH1 to adjust to harsh circumstances, hence controlling protein quality and folding (Fig. 3D).

The transition of pro-T cells to effector cells in primary tumors of nasopharyngeal cancer involved the maturation of undifferentiated T cells into functional T cells (Fig. 3E). The trajectory is demonstrated through signal transmission, immunological stimulation, inflammatory reactions, and the activation of metabolic genes during the process of cell differentiation. The genes ACTG1 and ITGAE govern cell adhesion and cytoskeleton. PCLAF, IL2RA, and CTLA4 are crucial immune response genes that work in conjunction with FOXP3 and IFI16. NR4A2 promotes immune homeostasis and prevents autoimmune diseases by enhancing immunological tolerance and Tregs. ALOX5AP and AOAH regulate lipid metabolism and inflammation, whereas PPP2R2B and TBC1D4 govern transcription and cell differentiation (Fig. 3G). Treg cells undergo a conversion process from FOXP3 to RBMS3, whereas metastatic cancers undergo differentiation from CD4_T_SPDYA to CD4_T_TOX2. The differentiation process starts with the CD4_T_TXNIP molecule (Fig. 3E). This may facilitate tumor cells in adapting to new surroundings by boosting their ability to survive, multiply, and spread. The genes UBA52 and IL7R are involved in cellular metabolism and proliferation, and tumor-immune interactions, respectively. The genes YPEL5, NR4A2, and ZNF331 have a role in signal transduction and stress response, whereas NKG7 and MT1E are involved in immunological control. These genes showcase the intricate behavior of tumor cells. The genes ATP1B3 and ARHGEF39 influence the movement of cells, whereas CD58 and ICOS enhance the ability of cells to evade the immune system, suggesting strategies for metastasis. The study of RNF125 in protein ubiquitination provides insights into how tumor cells adapt to maintain protein balance across different organs (Fig. 3G). Metastatic tumor cells undergo differentiation by evolving their gene expression, which emphasizes metabolic pathways for growth, modulation of stress response, and the intricate immunological milieu.

Discrepancies are observed when comparing transcriptional regulation across in primary and metastatic tumor groupings. The increase of DDIT3 indicates the presence of metastatic stress, whereas the elevation of NFATC1 supports the control of immune response to promote tumor growth. These cancerous growths increase the expression of CCNT2 and CREB1, necessitating improved regulation of the cell cycle in new environments to facilitate rapid proliferation. The expression of HOXB8 enhances the process of tumor cell differentiation and proliferation. Immunological tolerance and evasion in in situ cancers are affected by elevated levels of immunomodulatory genes, such as FOXP3 and GATA3. HIVEP1 and PBX4 enhance cellular differentiation and maintain structural integrity, whereas CEBPD, FOSL2, KLF13, and MAFF have the potential to exacerbate inflammation and stimulate tumor growth. The findings illustrate the intricate and dynamic interaction of regulatory systems in tumors, encompassing processes such as development, immune evasion, and environmental adaptability (Fig. 3F).

Elucidating heterogeneity and fate differentiation of B cells in primary and metastatic nasopharyngeal carcinoma

We employed unsupervised clustering to classify B cells in nasopharyngeal cancer in order to get insight into their mechanisms. This method detected 8 distinct groups, with showing the presence of typical B cell markers such as MS4A1, CD79A, CD79B, and CD19. Naive B cells had higher levels expression of BACH2, TCL1A, and FCER2, suggesting the presence of immature B cells. The B_CLECL1 and memory B DNAH8 exhibited the expression of CLECL1, whereas the latter demonstrated an upregulation of DNAH8, FCRL4, and CCR1, suggesting the presence of a distinct subtype of memory B cells. The presence of notable IL7R and BCL11B expression in the B_IL7R subgroup indicates a more favorable response to the IL7 signal. Both the proliferative_germinal_center_B and germinal_center_B subsets had elevated levels of AICDA, RGS13, and GCSAM, which suggests that there is active germinal center B cell activity. The former also had elevated MKI67 gene, suggesting ongoing cellular proliferation. The Plasma_cells_FNDC3B exhibited substantial expression of FNDC3B, MZB1, IGHG1, IGHA1, and TNFRSF17, indicating the presence of plasma cell characteristics (Fig. 4A–C).

Fig. 4.

B Cells manifest Dynamic Reorganization and Diversity within the Microenvironment of NPC. A UMAP plot reveals 8 clusters of B cells, including naïve B cells, Memory B cells, Germinal Center B cells, and Plasma Cells, each uniquely identified by varied colors. B The dot plot elucidates the expression profiles of marker genes within eight unique B cell clusters. C The FeaturePlot offers a visualization of the expression distribution of particular genes across groups, clearly depicting the manifestation of characteristic genes within diverse cell populations. D A violin plot graphically emphasizes the fluctuations in gene expression levels among groups within B cells, centering on the top 10 genes showcasing the most pronounced differences. E RNA velocity traces the progression trajectories of B cells across primary and metastatic tumors, with purple representing the onset of differentiation and yellow denoting the final stages. F The heatmap delineates the differential transcriptional landscapes among B cell subpopulations, highlighting the distinct regulatory mechanisms and gene expression profiles that define the heterogeneity within these cellular subsets. G The heatmap shows the changes in gene expression related to the trajectory across various B cell subgroups

Immunoglobulin genes such as IGHG2 and IGKC exhibit potent antibody-mediated protection by B cells in the early stages of cancer. This activity highlights the tactics of the B cell brigade in combating tumor invasion. TRIB1 and GAS6 collaborate to control cell viability and promote cellular communication, which is crucial for the proliferation and dissemination of tumors. During the metastatic phase, the tumor exhibits resilience to metastatic stress through the expression of heat shock proteins HSPH1 and HSPE1. This resilience enables the tumor to survive and adapt to new surroundings. The presence of metal ion processing genes MT2A and MT1X contributes to the narrative by demonstrating the requirement for equilibrium in order to prevent cellular demise caused by toxicity. The gene regulation and DNA repair mechanisms of ZNF10 contribute to the survival of tumors, whereas the immune evasion abilities of CD72 and FCER2 complicate the progression of the tumor. The emphasis of TCL1A on cell division and inhibition of programmed cell death, along with NIBAN3 in managing low oxygen levels, demonstrates the ability of tumor cells to adjust to challenging surroundings. (Fig. 4D).

The gene expression in primary tumors progresses from Naive B cells to B-CLEC1L, and finally transitions to Memory B cells. This process involves the control of genes such as WDR76, PCLAF, and CYTOR. The process demonstrates the development and functional change of B cells in the primary tumor microenvironment. The genes involved in this process are responsible such as repairing DNA, controlling transcription, regulating the cell cycle, and modulating the immune response. In contrast, the pattern of gene activity in metastatic cancers is more intricate, progressing from B-CLEC1L to Memory B cells, then through Proliferative germinal center B cells to Germinal center B cells, and finally culminating in the formation of Plasma cells. The key genes involved in this process are DERL3, PRDM1, CD247, and others. These genes primarily play a role in B cell proliferation, differentiation, immunological control, and cell death pathways. This alteration exposes B cells may undergo additional modifications during the metastatic process to enhance the survival and spread of tumor cells (Fig. 4E, G).

During the progression of primary tumors and their spread to other parts of the body, transcriptional regulatory factors play a crucial role by impacting different stages of tumor growth in a distinct manner. During the first stage, ELF3, a transcription factor that is exclusive to epithelial cells, has a vital function in maintaining cell identity while promoting cell growth and specialization. Simultaneously, SPI1 and BCL11B play crucial roles in coordinating the differentiation and activity of immune cells, so regulating the dynamic interaction between the tumor and the immunological defenses of the host. SREBF2 regulates lipid metabolism, providing tumor cells with vital energy and the necessary components for cellular membranes. Moreover, a combination of TGIF2, ZNF394, STAT4, TCF7, and NR2F1 contributes to the development of tumor by influencing several signaling pathways and gene expression patterns. On the other hand, metastatic tumors focus on improving DNA repair, speeding up the cell cycle, and effectively utilizing genetic information. The increased expression of BRCA1 serves as evidence of the improved repair mechanisms that are triggered in response to DNA damage. The E2F family members, such as E2F1, E2F2, E2F4, and E2F8, promote fast growth of cancer cells, ensuring their ability to adapt and withstand the challenging process of metastasis. YBX1 plays a critical role in enhancing the stability of mRNA and the efficiency of translation. This narrative provides a detailed description of the regulatory landscape of transcription, highlighting the significant impact these factors have on tumor growth at every stage (Fig. 4F).

Diversity of myeloid sub-populations in primary and metastatic tumors

In order to observe the effect of myeloid cells on the tumor microenvironment, they were divided into 9 groups (Fig. 5A). The Mac1, Mac2, Mono, and Neutrophil subgroups expressed the CD68, C1QA, and C1QB genes, while the LYZ gene was widely activated. The CD163 and MRC1 genes in Mac2 promote tissue healing and anti-inflammatory response. In addition, the Mono, which expresses more S100A12, VCAN, and FCN1, regulates the immune system and inflammation. Neutrophils expresses the S100A8 gene, indicating their synergy with dendritic cells. Moreover, cDC, and mDC activate T cells via CD1C and CLEC10A. FCER1A, LILRA4, and CLEC4C distinguish the mDC, which modulates the immune system. LILRA4 and CLEC4C are critical for virus defense and autoimmune balance. The migratory DC, with proliferation genes MKI67 and TOP2A and expression of CCR7, LAMP3, EBI3, and PDCD1LG2, shows how myeloid cells shape tumor progression and the immunological environment (Fig. 5B, C).

Fig. 5.

Exploring the Diversity and Subgroup Dynamics of Myeloid Cells in NPC Across Primary and Metastatic Tumors. A UMAP display reveals 9 unique subsets of myeloid cells, encompassing macrophages, monocytes, neutrophils, dendritic cells, and proliferative cells, each distinguished by a unique color palette. B The dot plot offers a detailed view of the expression patterns for marker genes within the eight delineated myeloid cell clusters, shedding light on their distinct genetic identities. C The FeaturePlot presents the distribution of specific gene expressions among various groups, effectively illustrating the unique gene signatures across different myeloid cell populations. D A violin plot highlights the variance in gene expression across myeloid cell groups, focusing on the top 10 genes with the most notable differences. E RNA velocity charts the developmental paths of myeloid cells in both primary and metastatic tumors, with purple indicating the early stages of differentiation and yellow marking the advanced stages. F A heatmap outlines the diverse transcriptional regulation within myeloid cell subsets, emphasizing the unique regulatory mechanisms and gene expression patterns that contribute to their heterogeneity. G A heatmap captures the dynamic gene expression changes associated with the developmental trajectories among myeloid cell subgroups, showing the evolution of gene expression as myeloid cells progress through their differentiation pathways

Tumor biology is diverse and multifaceted, as shown by primary and metastatic tumor gene expression profiles. High gene expressions of IL6, CCL2, and SLC25A37 in primary tumors indicate chronic inflammation in the tumor microenvironment, which affects immune system modulation. IL6 and CCL2, key regulators, boost tumor immune evasion and promote a tumor-promoting milieu. Furthermore, PTGS2 expression is closely connected to pain, inflammation, and tumor growth. High levels of IGLC3, TREM1, and SPP1 indicate immune cell activation and tumor cell contacts. MT1E and MT1X upregulation increases oxidative stress resistance, while NR3C1 emphasizes hormone signaling in tumor survival and adaptation. SOX4, a transcription factor, may promote phenotypic alterations and increase invasiveness. Metastatic tumors maintain genomic integrity and cope with stress by upregulating POLB and ITM2C, which are involved in DNA repair and the endoplasmic reticulum stress response. The increased expression of PLD4 and AFF3 genes suggests specialized cellular signaling and lipid metabolism transcriptional control (Fig. 5D).

In primary and metastatic tumors, Mono differentiates into Mac2, proliferative cells mature into mDC, and cDC with neutrophils in the late stage (Fig. 5E). At the start of tumor growth, FCRL1, TESPA1, RND3, and NFKBIA regulate inflammatory responses, cell polarity, and immune evasion. In particular, NFKBIA regulates NF-κB signaling pathway activity for tumor cell survival, whereas CCL3 and L1B genes highlight the importance of inflammatory mediators in the tumor microenvironment. ADAM12 and WNT5B expression affects cell adhesion, migration, and wnt signaling pathways, which are crucial for tumor cell invasiveness and microenvironment interactions. These findings reveal distinct molecular mechanisms for cell differentiation, immunoregulatory signaling, and cell–cell interactions in primary tumors. In metastatic samples, elevated SPIB and POLB expression is associated to immune responses and DNA repair, and LILRA4 and IRF4 may boost immunomodulation. ADAM12 shows extracellular matrix remodeling and cell migration during tumor growth. CCL18 and NFKBIA activation during metastasis may also boost tumor microenvironment immunosuppression and inflammation. These findings illuminate the molecular adaptations to new surroundings, enabling their survival, invasion, and dissemination in metastatic cancers (Fig. 5G).

BX1, HMGA1, and TEAD4 may promote tumor growth in transcriptional regulatory networks. In primary tumors, IRF8 promotes cell proliferation, preserves tumor stem cell features, and regulates cell–cell interactions. IRF7 and NFATC2 expression modulates the immunological landscape and immune evasion techniques during tumor cell interactions with the host's immune system, and spreading ability and adaptation depend on RUNX2, NR3C1, ZEB1, and NFKB1. Metastatic cancers express ZEB1, a key Epithelial-mesenchymal transition (EMT) regulator, to increase migration and invasion. ETS2, CEBPB, and MEF2A alter cell differentiation and cell–cell signaling to facilitate tumor metastasis. These factors regulate immunological modulation, cell cycle, cell fate determination, DNA damage response, and cell migration and invasion in primary and metastatic cancers (Fig. 5F).

Cell–cell communication heterogeneity in primary and metastatic nasopharyngeal carcinoma

Cell communication across 30 tumor cell populations (Fig. 6A) showed substantial differences in CAF, pro-T, and B-IL7R cell contacts in primary versus metastatic tumors (Fig. 6B). This discovery brought chemokine, co-inhibitory, and co-stimulatory molecule communications under scrutiny. The interaction between TNF superfamily members, their receptors, and immunomodulatory molecules is crucial in both tumor phases. These receptor pairs shape the tumor microenvironment, cell signaling, immunological control, and life cycles, as well as tumor cell motility and metastasis. TNFSF14-TNFRSF6B and TNFSF13-TNFRSF17 receptor pairings may boost local immune responses, helping tumor cells survive and proliferate. Additionally, the TNF-TNFRSF interaction can increase inflammation via the NF-κB pathway, promoting tumor growth. ICAM1 and VCAM1 linked to integrins make tumor cells tightly integrated with the milieu, facilitating local tumor propagation (Fig. 6C).

Fig. 6.

Cell–cell communications in primary and metastatic NPC. A The network diagram compares cell–cell communications among 30 different groups in primary and metastatic tumors, with the thickness of the lines indicating the weight of interaction between subgroups: thicker lines represent more communication, while thinner lines indicate less. B The heatmap illustrates the volume of cell–cell communication among multiple populations within two tumor groups, with red indicating relatively higher levels of communication and blue denoting relatively lower levels. C The dot plot presents the populations with the greatest quantitative differences in receptor-ligand interactions, namely CAFs, pro-T, and B-IL7R cells, highlighting the marked variations in chemokine, co-inhibitory, and co-stimulatory receptor dynamics among these subgroups

Tumor cells need CXCL12-CXCR4/CCR7 axis activation to proliferate, survive, and metastasis to new settings. The CXCL12-CXCR4 axis drives tumor cell migration to CXCL12-rich locations like lymph nodes, lungs, and bone marrow, facilitating metastasis (Fig. 6C). Thus, cell communication receptors are crucial to initial tumor genesis, maintenance, and metastasis. They enhance tumor cell survival and dissemination by changing signaling pathways and immunological responses, enabling tumor cells to thrive in new surroundings. Understanding these mechanisms provides tumor therapy molecular targets for customized treatment.

Exploring cellular heterogeneity and molecular characteristics of metastatic and non-metastatic primary tumors

To explore the differences between primary tumors with and without metastasis, we analyzed single-cell RNA sequencing data from 2 metastatic primary tumors and 10 non-metastatic primary tumors, categorizing them into 10 distinct clusters (Fig. 7A, B). Analysis of the cell type proportions revealed that metastatic primary tumors have higher proportions of myeloid cells, dendritic cells (DCs), epithelial cells, and plasma cells, whereas non-metastatic primary tumors predominantly contain T cells and B cells (Fig. 7C). The cell types identified include: CD4⁺ T cells (CD3D, CD3E), CD8⁺ T cells (CD3D, CD3E, CD8A, CD8B, GZMK), B cells (MS4A1, CD79A, CD79B, CD19), macrophages/monocytes (CD14, S100A8, C1QA, CD163, MRC1), plasma cells (MZB1, IGHG1, IGHA1, TNFRSF17), plasmacytoid DCs (IL3RA, LILRA4, CLEC4C), migratory DCs (LAMP3, EBI3, PDCD1LG2), mast cells (TPSAB1, CPA3, MS4A2, TPSB2), cancer-associated fibroblasts (CAFs) (COL1A1, COL3A1, DCN, FAP), and epithelial cells (KRT19, EPCAM, KRT18) (Figure D). In examining the differential gene expression between metastatic and non-metastatic primary tumors, we focused on the top 10 genes with the largest expression differences (Fig. 7E). Notably, CCL3L1 is highly expressed in metastatic primary tumors and plays a crucial role in regulating immune cell responses to tumors, affecting cell migration and invasion within the tumor microenvironment, potentially promoting or inhibiting tumor metastasis (Fig. 7F). Additionally, we conducted Gene Ontology (GO) enrichment analysis on the top 100 differentially expressed genes for both groups. The findings suggest that metastatic tumor cells are more active in energy and nucleotide metabolism, likely correlating with their enhanced proliferation and survival capabilities in new environments; meanwhile, non-metastatic primary tumor cells are more focused on maintaining protein stability and modulating functions related to immune interactions with the host, which may support their prolonged survival at the primary site (Fig. 7G). In our transcription factor analysis, several key transcription factors, including MYC, SOX2, ETS2, GATA1, and GATA2, are found to be highly expressed in metastatic primary tumors. These factors are involved in regulating essential processes such as cell proliferation, migration, and angiogenesis, and are closely linked to the aggressiveness and metastatic potential of various cancers (Fig. 7H). In summary,these results elucidate the cellular and molecular distinctions between metastatic and non-metastatic primary tumors, revealing potential targets for therapeutic intervention.

Fig. 7.

Comparative analysis of cellular and molecular profiles bewteen metastatic and non-metastatic primary tumors. A The umap plot displaying 10 distinct clusters, identifying various cell types such as CD4⁺ T cells, CD8⁺ T cells, B cells, Macrophages/Monocytes, plasma cells, pDCs, migratory DCs, mast cells, cancer-associated fibroblasts (CAFs), and epithelial cells. B Integration of single-cell RNA sequencing data, with metastatic primary tumors represented in red and non-metastatic primary tumors in green. C Proportions of cell populations from metastatic and non-metastatic primary tumors across the identified 10 clusters. D A dot plot illustrating the expression levels of marker genes across the 10 clusters, with high expression shown in red and low expression in blue. E A violin plot highlighting the top 10 genes exhibiting the most significant expression differences between metastatic and non-metastatic primary tumors. F A FeaturePlot visualizing the distribution of CCL3L1 expression within the two groups, high expression (Red), low expression (blue). G A heatmap displaying the differential transcriptional regulation between the two tumor types

Discussion

The study offered comprehensive insights into the evolutionary path and features of NPC metastasis. The current work has demonstrated that the microenvironment of NPC is actually more intricate and diverse than previously documented. In order to create a comprehensive record of gene activity in NPCs, we utilized a 10 × Genomics single-cell product library and Illumina's high-throughput sequencing technology. This allowed us to obtain scRNA-seq data from 24 primary carcinoma samples, 7 PBMC nasopharyngeal carcinoma samples, and 7 samples from patients with metastatic carcinoma. The samples included various cell types such as fibroblasts, NK cells, and different subtypes of T cells, B cells, and myeloid cells. We depicted a thorough genetic overview of main, regional lymph node, and metastatic tumors in NPC. Using phylogenetic analysis and scRNA-seq analysis, we compared the heterogeneity and fate differentiation of T/B cells in primary and metastatic nasopharyngeal carcinoma. Our investigation focused on the different functional potential of each cell in each type of nasopharyngeal carcinoma and the heterogeneity of cell–cell communication in primary and metastatic nasopharyngeal carcinoma.

Research findings have shown that there is a correlation between the longevity of patients with NPC and particular immunological signatures. This suggests that the number of different types of immune cells and the activation status of the immune system in NPC patients directly affect the formation of tumors and the outcome of treatment (Xu et al. 2022). Within the NPC microenvironment, a significant number of T lymphocytes that infiltrate the tumor are highly specialized in recognizing and targeting both tumor and EBV antigens (Nilsson et al. 2020). Nevertheless, study at the level of individual cells has revealed that these T cells, which fight against tumors, are experiencing disrupted balance in the immune system. They are demonstrating impaired ability to multiply and destroy cells (Gong et al. 2021b; Huang et al. 2023). We have detected a notable increase in fatigued CD8⁺ T lymphocytes in the tumor microenvironment of patients with metastases through the bloodstream, as determined by scRNA-seq. The recruitment of immunosuppressive cells to the initial tumor site shields cancer cells from cytotoxic cell-induced death and induces vascular permeability, resembling lymphatic channels. This process enhances the likelihood of hematogenous diffusion (Jain et al. 2016; Kitamura et al. 2015). Furthermore, the correlation analysis of B cells in primary tumors and metastatic tumors of nasopharyngeal carcinoma revealed that the gene expression pattern progresses from Naive B cells to B-CLEC1L in primary tumors, ultimately transitioning to Memory B cells. This process involves the regulation of genes such as WDR76, PCLAF, and CYTOR. The process described involves the development and functional change of B cells in the primary tumor microenvironment. The genes involved in this process are responsible for DNA repair, transcriptional control, cell cycle regulation, and modulation of the immune response. This exemplifies the dynamic involvement of the immune system in fighting against the tumor and how B cells adapt their functional status in response to alterations in the microenvironment, ensuring ongoing immunological defense (Miao et al. 2015; Sobti et al. 2023). Longitudinal study of data from patients with nasopharyngeal cancer revealed that an elevation in the peripheral monocyte-to-lymphocyte ratio was associated with a decrease in overall survival. Immunophenotyping at several time intervals revealed that individuals at low risk exhibited a notable reduction in the levels of monocytic myeloid-derived suppressor cells after chemotherapy. This reduction had a subsequent impact on the effectiveness of cytotoxic T-lymphocyte (CTL) immunotherapy(Hopkins et al. 2021; Li et al. 2023). We conducted a thorough examination of orthotopic tumors and metastatic tumors of nasopharyngeal carcinoma, revealing the alterations in different types of myeloid cells and their functions in tumor formation and metastasis. Furthermore, the myeloid cells primarily originated from the TME, suggesting that the recruitment of myeloid cells was specific to NPC. The myeloid-derived suppressor cells (MDSCs) and tumor-associated macrophages (TAMs) present in the TME act as a link to promote immune suppression. This is because several characteristic genes in these subtypes have previously been identified as being connected with immune cell exhaustion and cell cycle arrest. Overall, the single-cell data from our clinical cohort and the bulk RNA data from the GEO cohort offer valuable and distinct information about the identification and characterization of the stromal landscape in the NPC microenvironment. This data also provides strategies for making treatment decisions for NPC patients with distant metastasis, which could potentially enhance the survival outcomes of NPC.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

QL, JX and XD contributed to the conception and design of the present study, analyzed and interpreted the data, and critically revised the manuscript for important intellectual content. QL, JX, BD, DG, CS and XD contributed to designing the study, analyzed the data, and drafted and revised the manuscript. QL, JX, BD and XD confirm the authenticity of all the raw data. All authors read and approved the final manuscript. All the authors declare that they have no conflict of interest.

Funding

The authors received no financial support for the research authorship and/or publication of this article.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Conflict of interest

The authors declare no competing interests.