Inactivation of tumor suppressor gene p16/INK4A and oncogenic activation of KRAS occur in almost all pancreatic cancers. To better understand the roles of p16 in pancreatic tumorigenesis, we created a conditional p16 knockout mouse line (p16 flox/flox), in which p16 is specifically disrupted in a tissue-specific manner without affecting p19/ARF expression. p16 flox/flox; LSL-Kras G12D; Pdx1-Cre mice developed the full spectrum of pancreatic intraepithelial neoplasia (mPanIN) lesions, pancreatic ductal adenocarcinoma (PDA), and metastases were observed in all the mice. Here we report a mouse model that simulates human pancreatic tumorigenesis at both genetic and histologic levels and is ideal for studies of metastasis. During the progression from primary tumors to metastases, the wild-type allele of Kras was progressively lost (loss of heterozygosity at Kras or LOH at Kras) in p16 flox/flox; LSL-Kras G12D; Pdx1-Cre mice. These observations suggest a role for Kras beyond tumor initiation. In vitro assays performed with cancer cell lines derived from primary pancreatic tumors of these mice showed that cancer cells with LOH at Kras exhibited more aggressive phenotypes than those retained the wild-type Kras allele, indicating that LOH at Kras can provide cancer cells functional growth advantages and promote metastasis. Increased LOH at KRAS was also observed in progression of human pancreatic primary tumors to metastases, again supporting a role for the KRAS gene in cancer metastasis. This finding has potential translational implications-future KRAS target therapies may need to consider targeting oncogenic KRAS specifically without inhibiting wild-type KRAS function.