Soils and Seeds That Initiate Pancreatic Cancer Metastasis

The metastasis of pancreatic cancer cells is a terminal event in many patients. As such, substantial attention has been directed toward the characterization of the molecular and cellular features of pancreatic cancer metastasis in human samples and animal models. Metastasis can be functionally partitioned into an initiation stage, whereby neoplastic cells acquire the migratory properties to expand beyond the primary tumor, a dissemination stage denoting the systemic and intraluminal spreading of neoplastic cells, and an assimilation stage when metastatic cells occupy and expand in new tissue niches. Throughout this spectrum, metastatic cells are faced with new metabolic and immunologic challenges that must be overcome for the cells to survive these transitions.

Understanding the mechanisms of metastasis during the evolution of pancreatic ductal adenocarcinoma (PDAC) should help guide the development of meaningful interventions for patients. Studies using patient autopsy samples have demonstrated that metastasis is a late clinical event, with a surprising lack of genetic alterations between primary tumors and metastases (1, 2). Rather, primary tumors and corresponding metastases demonstrated dramatic differences in chromatin structure and metabolic pathways (3). In contrast, genetically engineered mouse models (GEMM) of PDAC showed that metastases can consist of polyclonal collections of disseminated primary PDAC cells (4). More recent work in GEMMs has implicated epigenetic alterations in metastases compared with the corresponding primary PDAC specimen (5). Collectively, these results provide the impetus to evaluate the cell-intrinsic and cell-extrinsic factors that promote metastasis initiation in pancreatic cancer.

In this issue of Cancer Discovery, Chiou and colleagues used GEMMs to demonstrate that hypoxia initiates a transient metastatic program in primary PDAC cells (6). They first determined that expression of the chromatin interacting protein HMGA2 was highly increased in more advanced and metastatic human pancreatic cancer. By incorporating an endogenous HMGA2-GFP reporter allele into the PDAC GEMMs, they discovered a subpopulation of highly transformed pancreatic cancer cells within primary tumors. Cellular assays with freshly isolated cells showed that the HMGA2-expressing subpopulation formed bigger spheres in culture and more metastases following transplantation, in comparison with the reporter-negative pancreatic cancer cells. Interestingly, the HMGA2 subpopulation was not stable, as both reporter-negative and reporter-positive cells were observed following transplantation. The transient nature of HMGA2 expression may underlie the seemingly surprising finding that very few gene expression changes were observed when bulk primary and metastatic PDAC cells were profiled. Instead, the direct molecular analysis of HMGA2-expressing PDAC cells revealed gene expression alterations that were most consistent with the activation of Hif1/2 due to intratumoral hypoxia. Colocalization of hypoxic regions with HMGA2-expressing cells was confirmed in the GEMMs, and the transcription factor BLIMP1 was identified as a HIF1/2 target gene that acts as a necessary effector in the metastatic process by driving its own transcriptional network. BLIMP1 simultaneously attenuated the cell cycle and promoted cell-invasive behavior, resembling processes previously described that are downstream of RUNX3 and other forms of epithelial–mesenchymal transition (EMT) in PDAC (7). Finally, expression levels of candidate BLIMP1 target genes stratified patients with PDAC to a worse outcome following surgery.

The finding that hypoxia drives BLIMP1 expression to initiate a metastatic program in PDAC opens new avenues of research for the pancreatic cancer field. The hypoxic propensity of PDAC may be ascribed to the tumor outgrowing an effective blood supply, or may more specifically reflect the unusual hypovascular nature of PDAC tumors (8). In addition to the development of direct antagonists of HIF1/2 and BLIMP1, analyzing the BLIMP1 transcriptome may nominate more tractable therapeutic targets to block the sensitization of primary PDAC cells toward metastatic dissemination.

Furthermore, the results implicating hypoxia as an initiating event in PDAC metastasis add to the growing literature of processes that occur in primary tumors to stimulate the metastatic process. For example, in a series of patient PDAC tumors, subclonal epigenetic alterations were reported to increase the expression and activity of the oxidative pentose phosphate pathways to promote metastasis (3). Our recent study reported that enhancer reprogramming activated the expression of a suite of genes involved in foregut endoderm development and occurred in a subset of PDAC cells, and enhancer reprogramming was necessary and sufficient to initiate the metastatic process (5). These developmental pathways were different from those reported in the current studies, persisted in the metastases, and promoted both primary cell proliferation and metastasis without engaging EMT-related programs. Whether enhancer reprogramming, metabolic reprogramming, and hypoxia-induced metastatic pathways occur in separate PDAC cell populations or occur cooperatively in the same cells in a primary tumor is unclear (Fig. 1). It is formally possible that the activation of developmental pathways by enhancer reprogramming may temporally precede these additional mechanisms. Indeed, the study by Chiou and colleagues reported the loss of the wild-type Trp53 allele in 90% of the primary tumor KPC cells, whereas our study first derived organoids from primary tumors and metastases and found only a minimal loss of the wild-type Trp53 gene in primary tumor organoids and uniform loss in organoids prepared from metastases (5). This suggests that the primary tumor samples and organoids used in our work were less transformed than those analyzed by Chiou and colleagues, and further work will clarify this discrepancy.

More generally, these findings provide a new perspective on the role of the microenvironment for providing the milieu or soil that initiates pancreatic cancer metastasis. Beyond hypoxia, nutrient deprivation per se is a well-described feature of PDAC tumors (9), and the adaptive responses to metabolic stress could conceivably also promote cellular behavior leading to metastasis. In addition to adapting to new metabolic challenges, gradients of morphogens and specific tropic factors may also awaken underlying developmental programs, making them more vulnerable for metastasis initiation. In the case of embryonic foregut development, a variety of pathways are known to direct endodermal progression and pancreatic tissue specification, including but not limited to Wnt, Hh, BMP, Notch, FGF, and retinoic acid (10). Although such a long list of potential mechanisms is admittedly speculative and daunting to consider, efforts focused on microenvironmental perturbations and developmental pathways could yield new insights into metastasis initiation and be valuable for interfering with this process in patients with pancreatic cancer.

No potential conflicts of interest were disclosed.

We apologize to colleagues whose work could not be cited due to length limitations. We thank Lindsey Baker for critical comments.

D.A. Tuveson is supported by the Lustgarten Foundation, the NCI, and the V Foundation. C.R. Vakoc is supported by the NCI. Both authors are supported by the NCI Cancer Center CCSG.