It is not known whether pancreatic cancers evolve from a single or multiple cells, or from a particular pancreatic lineage. However, in the Pdx1-Cre; LSL-KrasG12D; LSLTp53R172H mouse model of pancreatic cancer, all pancreatic lineages are susceptible to express mutant KRas and p53. Hence, such mouse model implies a scenario of maximal heterogeneity of cancer cell origins. On this basis, I isolated seven subclones of heterogeneous mouse pancreatic cancer cells from a single tumour; each of them had a distinct morphology and gene expression profile. Notably, they possessed different intrinsic phospho-SMADs downstream of the TGFβ receptor (phospho-SMAD2/3) or the BMP receptor (Phospho-SMAD1/5/8). I discovered that SMAD4, a co-SMAD which is frequently found to be lost in pancreatic caner tissues, upregulated HNF4α via the classical BMP-SMAD1 pathway, when cells were experiencing metabolic stress upon deprivation of serum, or in the presence of excess thymidine. Under serum starvation at a hypoglycemic-like glucose concentration, the HNF4α-expressing sub-clones appeared to be more able to sustain an unstressed morphology than other non-HNF4α-expressing sub-clones. Immunohistochemical staining on pancreatic cancer sections revealed nuclear co-localization of SMAD4 and HNF4α in human (half of the cases) and in mouse samples. As a secondary project conducted during characterization of cells, I also found that three of the subclones more robustly proliferated under anchorage independent conditions, and they relied on the MEK-ERK pathway and the canonical Wnt pathway, to a different degree. Both studies demonstrate for the first time in primary cell culture that pancreatic cancer cells within a tumour could be highly heterogeneous in terms of both morphology and signaling pathways.