Figures and data
Cellular composition of SiNETs as determined by single cell and single nuclei sequencing.
(A,B) UMAP plots showing the diversity of single cells from SiNET2 (A) and SiNET5 (B), colored by their cluster assignment. (C,D) Cluster annotations (top bar) in SiNET2 (C) and SiNET5 (D) are supported by the expression of canonical cell type markers (rows). Also shown are three cell cycle markers (bottom rows) (E) Cell type frequencies in each of the 10 SiNETs that we profiled, along with one SiAdeno sample.
SiNETs broadly classify into two major subtypes
(A) Barplot showing the number of upregulated genes against a common threshold, number of genes (y-axis) vs number of tumors (x-axis). (B) Heat map showing a list of 25 representative genes that define SiNET signature of our scRNA seq cohort used to cluster NET samples from a bulk-seq dataset [15]. Type of NET is color coded on the top panel, with P-NET and RE-NET referring to pancreatic and rectal NETs. (C) Heatmap representing clustering of SiNET samples in our cohort based on genes that were differentially expressed and shared between 2-5 samples, showing two major variable gene programs (D) Correlation heat map between the NET samples. (E) Heatmap showing average expression of epithelial and neuronal gene-sets (rows) in the neuroendocrine and epithelial cells from our SiNET samples (columns). Epithelial gene-sets include signatures of multiple cell types from the small intestine [17], and neuronal gene-sets include three clusters of neurons [36]. (F) Heatmap of the SiNET dominated cluster from the bulk dataset [15] was subjected to differential expression analysis using the same set of genes as (C).
Heterogeneity in the SiNET tumor microenvironment.
For each of three non-malignant cell types, the diversity of that cell type is shown in one exemplary tumor: fibroblast heterogeneity in SiNET8 is shown in (A-C), endothelial cell heterogeneity in SiNET5 is shown in (D-F), and B-cell heterogeneity in SiNET2 is shown in (G-I). For each cell type, three panels depict three types of analyses. The first panel (A, D, G) is a UMAP plot of the respective tumor, where only the respective cell type is colored, and distinct colors highlight the clusters of that cell type. The second panel (B, E, H) shows differential expression analysis between the first two clusters using heatmaps, with labeling of selected genes. The third panel (C, F, I) shows clustering of cells from that cell type (columns) based on their relative expression of previously defined [18] signatures of diversity in that cell type (rows); the top panel shows assignment of cells to clusters.
Cell cycle analysis reveals proliferating B cells in SiNETs.
(A) Bars show the percentage of cycling cells (y-axis) per cell type and per tumor (x-axis). Tumors are color-coded, the two subtypes, Epithelial-like and Neuronal-like, are differentiated by distinct shapes, represented as square and circle, respectively. Information regarding the presence of cell types with zero percentage of cycling cells is provided along the x-axis. Horizontal lines indicate average percentages of cycling cells per cell type. (B) Correlation between cell type and cell-cycle program as computed from an SiNET bulk RNA-seq dataset [15]. Score for each cell type is represented on the top of individual panels. (C) Boxplot depicting the expression of MIF in each SiNET cell type, for each of the two SiNET subtypes.
A putative progenitor population in mixed LCNEC.
(A) Copy number variation (CNV) profiles inferred from scRNA-seq data for all cells from the LCNEC sample. Malignant and non-malignant cells are annotated based on their CNV profiles, using the same color codes as in next panels. (B) tSNE plot showing the diversity of single cells from the mixed lung tumor, colored by their clustering. (C) Heatmap shows relative expression of differentially expressed genes (rows), separated by horizontal lines into programs that distinguish between the four populations of cells detected in the LCNEC sample. Also included are two cell cycle programs (G1/S and G2/M). Columns correspond to malignant cells, separated into the four populations by vertical lines and as indicted by color at the top. Selected genes are labeled for each program. (D) Bars show the percentage of cycling cells in each malignant cluster. (E) Malignant cells scored against an epithelial vs. neuroendocrine programs (gene set), colored by their assignment into four populations.
Cellular composition of SiNETs as determined by single cell and single nuclei sequencing.
(A-H) UMAP plots showing the diversity of single cells from each sample, colored by their clustering.
Specific upregulated genes expressed in neuroendocrine cells per siNET sample.
Heterogeneity in the SiNET tumor microenvironment.
For each of three non-malignant cell types, the diversity of that cell type is shown in one exemplary tumor: fibroblast heterogeneity in SiNET3 is shown in (A, B), fibroblast cell heterogeneity in SiNET5 is shown in (C, D), and endothelial heterogeneity in SiNET2 is shown in (E, F). For each cell type, panels depict types of analyses. The first panel (A, C, E) is a UMAP plot of the respective tumor, where only the respective cell type is colored, and distinct colors highlight the clusters of that cell type. The second panel (B, D) shows differential expression analysis between the first two clusters using a heatmap, with labeling of selected genes. The third panel (F) shows clustering of cells from that cell type (columns) based on their relative expression of previously defined [18] signatures of diversity in that cell type (rows); the top panel shows assignment of cells to clusters.
Heat map illustrating the expression of G1/S and G2/M genes across various cell types in the Epithelial-like Subtype (A), Neuronal-like Subtype (B) and the siAdeno sample.
The annotated cell types include Epithelial cells, Macrophages, T cells, and Fibroblasts that were sampled for illustrative purposes.
Scatter plot illustrating the percentage of cycling B/Plasma cells and the correlation between the germinal center signature and cycling B/Plasma cells signature.
In SiNET1 and SiNET2 we observe high correlation between cell cycle and GC score, but not in SiNET9.
