Pancreatic ductal adenocarcinoma (PDAC) is among the worst diagnoses in medicine, with a 5-year survival rate of 4%, in part because most cases are detected well after metastasis has occurred. Models of PDAC evolutionary history have emerged from DNA sequencing studies of metastases and primary tumours and reveal a progressive elaboration of malignant and metastatic cells that takes 20 or more years.1 Interestingly, at least half of this time is spent in a premalignant state, termed pancreatic intraepithelial neoplasia (PanIN), suggesting that a decade-long window exists for prediction, detection and prevention of pancreatic cancer. Two papers published in Gut use mouse models to derive new insights into human-relevant genetic and environmental influences that drive PanIN formation and progression, and highlight the importance of differentiation as a barrier to tumourigenesis.2 ,3

The first genetic event on the road to invasive pancreatic cancer is mutational activation of KRAS, which occurs in the earliest precancerous PanIN-1 lesions. Experimental research on PDAC initiation was transformed 10 years ago, when a mouse model of PanIN–PDAC progression was developed based on pancreas-specific, Cre recombinase-mediated activation of endogenous mouse Kras.4 This model has been used to validate and dissect both rare genetic risk factors for PDAC, such as Ink4a/Arf deficiency,5 and environmental risk factors such as chronic pancreatitis.6 These particular factors account for only a small proportion of human PDAC, however, leaving open whether more common risk factors can be identified in humans and dissected in mice.

Genome-wide association studies (GWAS) provide one avenue to identify genomic regions containing potential common risk alleles for human diseases. The increased risk attributable to such regions is typically modest (<1.5-fold), as expected for common alleles contributing to complex, non-Mendelian disorders. More frustrating for researchers, the actual single-nucleotide polymorphisms (SNPs) identified by GWAS rarely have obvious effects on protein-coding regions, nor are they linked to clear causal variants in adjacent genes.7 As a result, GWAS often identify candidate genes of uncertain functional importance, whose actual implication in the disease of interest is unclear. Among a small number of genomic regions for which GWAS have recently found non-coding, PDAC risk-associated SNPs, the NR5A2 gene has emerged as an interesting candidate.8 NR5A2 encodes an ‘orphan’ nuclear hormone receptor, also referred to as liver receptor homologue-1 (LRH-1), which has been implicated in a variety of biological processes that include embryonic stem cell pluripotency, cholesterol metabolism and steroidogenesis.9 In the adult pancreas, Nr5a2 cooperates with the acinar-specific transcription factor complex pancreas-specific transcription factor 1 (PTF1) to directly bind and activate a battery of acinar-specific genes.10 Loss of acinar gene expression has been suggested to represent a first step of PDAC development: targeting mutant Kras to mouse acinar cells can cause their dedifferentiation and reprogramming into ductal PanIN lesions, particularly following experimental pancreatitis, while duct cells themselves are remarkably refractory to Kras-induced PanIN formation.11 ,12

Together, the above observations suggest that Nr5a2 might act as an inhibitor of acinar-to-ductal reprogramming, possibly explaining its implication in pancreatic cancer. In putting this hypothesis to the test, the new studies by von Figura et al 2 and Flandez et al 3 provide the first functional validation of a GWAS-identified risk factor for pancreatic cancer, and yet another validation of the Kras mouse model for dissecting mechanisms of tumour susceptibility. Both papers are straightforward and in close agreement on the key details summarised briefly: (1) acinar cells transiently downregulate Nr5a2 during recovery from pancreatitis (during which time they are known to be particularly susceptible to Kras), and this downregulation persists during PanIN and PDAC formation; (2) heterozygosity3 or pancreas-specific deletion2 of Nr5a2 impairs regeneration after pancreatitis, leaving acinar cells in a duct-like metaplastic state; and (3) reduced or eliminated Nr5a2 dosage dramatically sensitises the pancreas to Kras-induced PanIN initiation. Most of these experiments were performed in global heterozygous or pan-pancreatic Nr5a2 knockout mice, leaving open the possibility that Nr5a2 acts non-cell-autonomously to suppress PanIN formation (eg, by preventing transformation of duct cells, or globally reducing the severity of inflammation). von Figura et al,2 however, use an inducible Cre approach to simultaneously activate Kras while deleting Nr5a2 in adult acinar cells only, and demonstrate that such cells undergo rapid, cell-autonomous transformation into ductal PanINs.

Together, these studies leave no doubt that Nr5a2 inhibits the ductal transformation of adult acinar cells by mutant Kras. It is important to avoid the term ‘tumour suppressor’ when describing human NR5A2, however, as a classic tumour suppressor is predicted to undergo somatic mutation and/or loss-of-heterozygosity, which appear to be excluded for NR5A2 based on human PDAC sequencing studies. Nr5a2 is both a target and partner of the PTF1 transcription factor complex, which normally sustains its own activity via autoregulation.10 One can imagine a scenario in which reduced Nr5a2 activity, via heterozygosity in mice3 or potential cis-regulatory SNPs in humans,8 would create instability in the Nr5a2–PTF1 autoregulatory loop, increasing the likelihood of its being shut down entirely by mutant Kras. As important as it will be to validate this model, it is even more important to understand how Nr5a2 normally achieves its inhibition of Kras: is it direct (eg, driving expression of a negative regulator of the RAS pathway) or indirect (eg, suppressing inflammation, which itself synergises with Kras6 ,12), does it act cooperatively with PTF1 or independently and is elevation of its activity sufficient to prevent PDAC initiation? The latter question is particularly important given that Nr5a2, as a nuclear receptor, is a druggable target (albeit of unknown endogenous ligand specificity).9 In a perfect world, Nr5a2 agonists might be useful as chemopreventive agents—to the chorus of theoretical objections to such a scenario, however, must be added the practical fact that NR5A2 expression has been found to promote, rather than inhibit, the growth of several human PDAC cell lines in vitro.13 Perhaps Nr5a2 inhibits Kras early, in the context of PTF1-dependent acinar differentiation, while having an opposing activity later in PDAC progression. Further studies comparing mouse and human Nr5a2 expression, regulation and function should resolve this possible contradiction. In any event, like all the best science, these new studies2 ,3 raise as many interesting questions as they answer, and encourage efforts to use animal models to investigate otherwise intractable problems in this currently intractable human disease.