Integrating single-cell and spatial transcriptomics reveals endoplasmic reticulum stress-related CAF subpopulations associated with chordoma progression

Abstract

Background: With cancer-associated fibroblasts (CAFs) as the main cell type, the rich myxoid stromal components in chordoma tissues may likely contribute to its development and progression.

Methods: Single-cell RNA sequencing (scRNA-seq), spatial transcriptomics, bulk RNA-seq, and multiplexed quantitative immunofluorescence (QIF) were used to dissect the heterogeneity, spatial distribution, and clinical implication of CAFs in chordoma.

Results: We sequenced here 72 097 single cells from 3 primary and 3 recurrent tumor samples, as well as 3 nucleus pulposus samples as controls using scRNA-seq. We identified a unique cluster of CAF in recurrent tumors that highly expressed hypoxic genes and was functionally enriched in endoplasmic reticulum stress (ERS). Pseudotime trajectory and cell communication analyses showed that this ERS-CAF subpopulation originated from normal fibroblasts and widely interacted with tumoral and immune cells. Analyzing the bulk RNA-seq data from 126 patients, we found that the ERS-CAF signature score was associated with the invasion and poor prognosis of chordoma. By integrating the results of scRNA-seq with spatial transcriptomics, we demonstrated the existence of ERS-CAF in chordoma tissues and revealed that this CAF subtype displayed the most proximity to its surrounding tumor cells. In subsequent QIF validation involving 105 additional patients, we confirmed that ERS-CAF was abundant in the chordoma microenvironment and located close to tumor cells. Furthermore, both ERS-CAF density and its distance to tumor cells were correlated with tumor malignant phenotype and adverse patient outcomes.

Conclusions: These findings depict the CAF landscape for chordoma and may provide insights into the development of novel treatment approaches.

Keywords: cancer-associated fibroblasts; chordoma progression; endoplasmic reticulum stress; scRNA-seq; spatial transcriptomics.

Figure 1.

The single-cell atlas of chordoma tissues. Diverse cell types in chordoma delineated by single-cell transcriptomic analyses. (A) UMAP of 49 cell clusters identified from the scRNA-seq data. (B) UMAP of cell subclusters identified from the scRNA-seq data of 6 chordoma tissues and 3 NP tissues (left). The corresponding UMAP of cell subclusters for each sample (right). (C) The proportion of each cell type from each sample. (D) Expression of selected marker genes in the major cell types of chordoma tissues and NP tissues. (E) The cell-type marker genes expression matrix in the 10 cell types isolated from chordoma tissues and NP tissues.

Figure 2.

Identification of CAFs by single-cell RNA-seq. Fibroblast subclustering reveals distinct CAFs populations. (A) Two-dimensional UMAP projection of 6 fibroblast clusters across all samples. (B) UMAP plot of fibroblast specific markers for all fibroblast clusters. (C) Heatmap of top 10 DEGs per cluster for each fibroblast subpopulation. (D) The proportion of each fibroblast population from each sample. (E) UMAP of normal fibroblast clusters, CAFs, and cancer cells (Left); the cell cycle score and stemness score for the 3 cell types (Right). (F) GSVA results of fibroblasts, CAFs, and chordoma cells. (G) Heatmap of DEGs identified in fibroblasts, CAFs, and chordoma cells.

Figure 3.

Identification of CAF subtypes. CAF subclustering reveals 4 CAFs subpopulations. (A) Two-dimensional UMAP projection of 4 CAF clusters across chordoma samples. (B) UMAP plot of CAF-specific markers for 4 CAF clusters. (C) Heatmap of top 10 DEGs per cluster for each CAF subpopulation. (D) The proportion of each CAF subpopulation from each chordoma sample. (E) Violin plot showing the expression levels of DEGs per cluster for each CAF subpopulation. (F-I) Gene ontology. (GO) enrichment anlayses of CAF-2 (F), CAF-1 (G), CAF-3 (H), and CAF-4 (I).

Figure 4

Association between ERS-CAF signature scores and chordoma progression. (A) The heatmap shows the expression level of ERS-CAF marker genes among chordoma patients. (B) The heatmap shows the distribution of clinicopathological characteristics and ERS-CAF signature scores among chordoma patients. (C) Kaplan–Meier curves of LRFS and OS of chordoma patients stratified by ERS-CAF signature score. (D–E) Multivariate Cox regression model including factors that were significant in univariate analysis for LRFS (D) and OS (E) of chordoma patients.

Figure 5.

Cell communication prediction. (A) Circle plot showing the number of interactions and interaction weight/strength among different cell types. The line thickness is proportional to the number of ligands-receptors pairs. (B) Detailed view of ligands broadcast by each major cell population. (C) Ligand–receptor communication network between CAFs and other major cell types predicted by scRNA-seq data. Right, heatmap of the top ligands expressed by 3 CAF subpopulations. Middle, heatmap of predicted ligand-receptor pairs between CAF subpopulations and different cell types in chordoma. Bottom, expression heatmap of top predicted receptors regulated by CAF subpopulations in different cell types. (D) Heatmap of receiving signal patterns and sending signal patterns of all cell types identified by CellChat analyses. (E) Circle plot showing the inferred JAM, THBS, and GDF signaling networks among different cell types.

Figure 6.

Spatial distribution of CAF subpopulations. (A) Spatial transcriptomics (ST) map of primary chordoma tissue. (B) The correlation between the genes and UMIs. (C) Spatial plots showing the marker genes for epithelial-like cells, mononuclear phagocytes, T cells, endothelial cells, neutrophils, B cells, chondrocytes, and mural cells. (D) T-SNE plot showing the ST clusters of primary chordoma tissues. (E) Violin plots showing the expression of CAF marker genes across the ST clusters for primary chordoma. (F) Heatmap showing the distribution of cell types identified by scRNA-seq in ST map. G–L for recurrent chordoma using the same presentation form as primary chordoma.