Single-cell transcriptomic analysis of canine insulinoma reveals distinct sub-populations of insulin-expressing cancer cells

Abstract

Canine malignant insulinoma is a rare, highly metastatic and life-threatening neuroendocrine tumour of pancreatic beta cells. To map the single-cell transcriptomic landscape of canine insulinoma for the first time, transcriptomic profiles of 5,532 cells were captured from two spontaneous insulinomas (Patient 1 and 2) and one associated metastasis (Patient 2) in two Boxer dogs. Distinct cancer, endocrine, and immune cell populations were identified. Notably, all three tumour samples contained two transcriptionally distinct insulin-expressing tumour cell populations (INS⁺ and INS⁺FOS ^low ), characterised here for the first time. These two cancer cell populations significantly differed by ~ 8,000 differentially expressed genes (DEGs), particularly tumour suppressor genes (e.g. TP53, EGR1) and cancer-related pathways (e.g., MAPK, p53). In contrast, COX7A2L was one of a few genes ubiquitously expressed and significantly upregulated (> 20-fold) in both insulin-expressing tumour populations compared to other captured populations. Both populations were also characterised by expression of chromogranin/secretogranin neuroendocrine tumour marker genes (e.g. CHGA, SCGN). There were far fewer gene expression differences observed between insulin-expressing tumour cells from the two patients (~ 600 DEGs) than between the two cancer cell populations within each patient. These DEGs included CLTRN, TMSB4X, CSRP2, LGALS2, and C15orf48. Unexpectedly for a tumour of endocrine origin, the metastasis in Patient 2 exhibited > 20-70 fold upregulation of exocrine pancreatic genes including CLPS, PRSS2, PRSS and CTRC. Immune cell analyses identified distinct infiltrating immune populations, including memory T cells and macrophages and revealed likely tumour-immune interactions, including the CD40-CD40L interaction. This study provides the first single-cell RNA sequencing (scRNA-seq) analysis of naturally occurring insulinoma in any species, revealing tumour cell heterogeneity, novel immune microenvironment features, and potential therapeutic targets. Despite its small scale, the findings highlight the utility of scRNA-seq in veterinary oncology and its translational potential for pancreatic neuroendocrine tumours across species.

Supplementary information: The online version contains supplementary material available at 10.1186/s44356-025-00026-3.

Keywords: Boxer; EGR1; P53; canine; insulinoma; scRNA-seq.

Fig. 1

Canine insulinoma study overview. A Primary tumour tissue from an 8-year-old female neutered Boxer diagnosed with insulinoma (Patient 2) received after tissue had been removed for diagnostic purposes. B Metastatic lesion tissue from Patient 2 received after tissue had been removed for diagnostic purposes. C Study overview depicting excision, cell preparation, scRNA-seq, and bioinformatic analyses of tumour samples from Boxers diagnosed with insulinoma

Fig. 2

Cell populations captured from scRNA-seq of canine insulinoma samples. A (Left) UMAP showing cells captured from tumours. The cell type corresponding to each cluster of cells was determined through standard marker expression, top distinguishing markers, and automatic classifiers (scMCA 0.2.0, scATOMIC 2.0.3). (Right) Doughnut plots showing the proportions of each cell type captured from each tumour sample. See Supplementary Table 3 for details. B UMAPs showing levels of select standard marker or distinguishing marker genes for cell type classification: Acinar (CEL), Alpha cell (GCG), B cell (MS4A1), Delta cell (SST), Ductal cell (KRT19), Endothelial cell (PLVAP), Gamma cell (PPY), INS⁺ FOS^low (INS, FOS, EGR1), INS⁺ (INS), Macrophage (CD86), S100A12^high (CD86, S100A12), Nerve (PLP1), Pancreatic stellate (COL1A2), T (CD3E), T CCL5^high (CD3E, CCL5), Muscle (ACTG2). See Supplementary Table 4 for cluster distinguishing markers and Supplementary Table 10 for scATOMIC classification

Fig. 3

Characterization of two distinct insulin-expressing cancer cell populations found within each tumour. A Heatmap of the top 50 DEGs between the INS⁺ FOS^low and INS⁺ clusters (adj.p <.05, avg log2 FC > 1). B KEGG over-representation pathway analysis of the DEGs between the INS⁺ FOS^low and INS⁺ clusters (adj.p <.05). C Violin plots of TP53, FOS, EGR1, and DUSP1 expression between the insulin-expressing clusters, showing significant downregulation in the INS⁺ FOS^low cluster. D UMAP plots showing downregulation of select DEGs in the INS⁺ cluster. E UMAP plots showing expression of chromogranin and secretogranin family genes in both the INS⁺ FOS^low and INS⁺ clusters

Fig. 4

Heterogeneity between primary tumours. A Violin plots of select genes that were differentially expressed in both insulin-expressing populations when comparing the primary tumours of Patient 1 to the primary tumour of Patient 2. B UMAP plots showing upregulation of immune/inflammatory-related genes OAS2, CCL2, and AKAP13 in Patient 1 compared to Patient 2 across the cell populations captured and upregulation of CD40LG in Patient 2 in both populations of insulin-expressing cancer cells, compared to these cells in Patient 1. C Heatmap of beta cell, glucose, and insulin-related DEGs between the primary tumours. Expression in the metastasis of Patient 2 is also shown for comparison. D Violin plots of select glucose and insulin-related DEGs from Fig. 4 C: SLC2A2, PIK3R1, IGFBP7, and SGK1 across samples – Patient 1 (red), Patient 2 primary (green), Patient 2 metastasis (blue)

Fig. 5

Differentially expressed genes in metastatic insulin-expressing tumour cells compared to primary tumour in canine insulinoma. A Heatmap of INS⁺ cluster (left) and INS⁺ FOS^low cluster (right) DEGs between primary tumour and metastasis (adj.p <.05, avg log2 FC > 1). B Violin plots showing upregulation of pancreatitis markers in both insulin-expressing cancer cell populations in the metastasis compared to primary tumour. P-values shown reflect the analysis between the primary tumour and metastasis for the INS⁺ cell cluster. For the INS⁺ FOS^low cluster analysis and p-values, see Supplementary Table 11. C Violin plot showing that the insulin-expressing cancer cell populations have significantly higher stemness scores compared to the INS⁺ FOS^low population in the primary tumour of Patient 1. Shown are Holm-adjusted p-values from Games-Howell pairwise tests

Fig. 6

Characterisation of tumour-infiltrating immune cells and cancer cell response in canine insulinoma. A Heterogeneity of macrophage immune response between patients is shown in a heatmap of a subset of DEGs involved in immune-related processes. B Cell communication analysis showing significant interactions between insulin-expressing cancer cells and tumour-infiltrating immune cells. C Circle plots showing significant interactions of PTPRC—MRC1, CLEC2D—KLRB1, and SELPLG – SELL between immune cell populations. D Circle plots showing significant interactions of CD40LG − CD40 and APP − TREM2 + TYROBP between insulin-expressing cancer cells and immune cells