Abstract: Objective To explore the heterogeneity of pancreatic cancer and new methods for tumor cell molecular subtyping and identify the signature genes in pancreatic cancer progression. Methods Based on the single-cell sequencing data of 16 pancreatic cancer tissues from the GSE155698 dataset, the single pancreatic cancer cells were classified according to EPCAM gene expression after preliminary clustering, re-clustering, and subgrouping to identify the signature genes, followed by pathway enrichment analysis and pseudo-time analysis. The key genes identified were validated using the clinical and tissue gene and protein expression data from 179 pancreatic cancer patients and 171 healthy controls. The impact of CEACAM5, LGALS1, and CENPF on proliferation, migration and invasion of pancreatic cancer cells were analyzed. Results Analysis of 48 570 cells from 16 pancreatic cancer samples revealed a total of 22 clusters, including 5 clusters of pancreatic cancer cells, which were classified into Subtype 1, Subtype 2, and Subtype 3, each exhibiting distinct gene expression patterns and functions. The signature genes were enriched in negatively regulated protein metabolic processes, ferroptosis, and antigen processing and presentation-related pathways in Subtype 1 pancreatic cancer cells; in peptide synthesis processes, translation, and ribosome-related pathways in Subtype 2; and in ATP metabolic processes, glycolysis/gluconeogenesis, and cell cycle-related pathways in Subtype 3. Subtypes 2 and 3 were potentially derived from Subtype 1, and Subtype 3 possibly represented the final developmental stage of pancreatic cancer cells. The key signature genes (CEACAM5, LGALS1, and CENPF) also exhibited different expression patterns in the developmental trajectory and showed high expressions in pancreatic cancer in association with poor prognoses. In pancreatic cancer cells, downregulation of CEACAM5, LGALS1, and CENPF significantly inhibited the proliferation, migration, and invasion abilities of the cells (P<0.05). Conclusion Pancreatic cancer cells exhibit significant heterogeneity, and CEACAM5, LGALS1, and CENPF gene expressions, which affect pancreatic cancer cell proliferation, invasion, and metastasis, can be used to identify distinct molecular subtypes during tumor cell development.

Key words: pancreatic cancer; molecular subtypes; heterogeneity; single cell sequencing