More than acinar identity? A novel cystic phenotype suggests broader roles for NR5A2 in pancreatic cancer

Abstract

The prognosis for pancreatic ductal adenocarcinoma (PDAC) remains dismal. Multiple genome-wide association studies (GWAS) have implicated the nuclear receptor NR5A2 in modulating PDAC risk, but mechanisms for this association are not understood. NR5A2 is a transcription factor that maintains acinar cell identity, and heterozygous loss of Nr5a2 in mice accelerates oncogenic Kras-driven formation of pancreatic intraepithelial neoplasia (PanIN), a PDAC precursor derived from acinar cells. In a recent issue of The Journal of Pathology, Cobo et al characterize a novel mouse model that uses Ptf1a:Cre to drive oncogenic Kras as well as heterozygous Nr5a2 inactivation. In addition to the expected PanIN lesions, these mice exhibited a surprising phenotype: large pancreatic cystic lesions which have not been previously reported. Comparing expression of oncogenic Kras and heterozygous Nr5a2 in various mouse models reveals several possible explanations for these cystic lesions. Importantly, these differences across mouse models suggest that NR5A2 may contribute to PDAC precursors in ways beyond its previously characterized acinar cell-autonomous role. These observations highlight that pathways implicated by GWAS may have roles in unexpected cell types, and an understanding of these roles will be critical to guide new preventive and treatment strategies for PDAC.

Pancreatic ductal adenocarcinoma (PDAC) remains a devastating disease with a 10-year survival rate of approximately 10% [1]. Surgical excision of early disease offers the only chance of a cure, and there are no established chemopreventive strategies. Given the challenges of early diagnosis, mouse models for PDAC are critical to our understanding of the earliest events in cancer initiation. Mirroring the finding of oncogenic KRAS mutations in the vast majority of human PDAC cases, pancreas-specific expression of oncogenic Kras in these mouse models leads to acinar-to-ductal metaplasia (ADM) and the formation of PDAC precursor lesions called pancreatic intraepithelial neoplasia (PanIN). Furthermore, just as pancreatitis is a risk factor for human PDAC, tumorigenesis in these mouse models is greatly enhanced by caerulein-induced acute pancreatitis, which leads to widespread ADM and inflammation.

Multiple genome-wide association studies (GWAS) have implicated the nuclear receptor NR5A2 in modulating risk for PDAC, but the mechanisms for this association are not well understood. Recent single-cell transcriptomic studies have confirmed that NR5A2 is highly expressed in acinar cells [2], where it stabilizes acinar identity, prevents inflammation, and is required for acinar recovery following injury [3,4]. Loss of acinar identity is one of the first steps of PDAC initiation, suggesting that a cell-autonomous mechanism may contribute to PDAC risk. Risk alleles have been associated with loss of NR5A2 protein expression [4], supporting the hypothesis that loss of function of NR5A2 may promote PDAC formation.

To investigate this hypothesis, several studies have tested the impact of Nr5a2 genetic loss of function in mouse models of PDAC. Nr5a2 loss of function accelerates tumor initiation in these models. Constitutive heterozygosity of Nr5a2 cooperates with caerulein-induced pancreatitis to accelerate the formation of Kras-driven PanIN lesions [5]. Similarly, conditional heterozygous knockout of Nr5a2 in pancreas epithelium driven by a Pdx1^late-Cre allele accelerates the formation of Kras-driven ADM and PanIN lesions [3]. These observations support the model that NR5A2 loss of function destabilizes acinar cell identity and promotes PDAC formation, and that risk alleles identified by GWAS may act by diminishing NR5A2 function in acinar cells.

In a recent issue of The Journal of Pathology, Cobo et al [6] expand on these studies by examining the consequences of Nr5a2 conditional loss of function driven by a Ptf1a:Cre allele. While this new model is conceptually similar to those of prior studies, a dramatic new phenotype was observed: large pancreatic cystic lesions. The similarities and differences between these models uncover critical new insights and questions about the formation of PDAC and its precursor lesions.

A tale of three models

Using the Ptf1a:Cre allele, Cobo et al drive conditional heterozygous inactivation of Nr5a2 in a majority of acinar and ductal epithelial cells beginning at embryonic day 10.5. As expected, in mice harboring an oncogenic Kras^G12V allele, inactivation of one Nr5a2 allele (KPN^+/− mice) leads to widespread ADM, focal inflammation, and the formation of low- and high-grade PanIN precursor lesions, none of which are present in mice with both Nr5a2 alleles intact (KPN^+/+ mice). These precursor lesions in KPN^+/− mice were seen at 4–7 weeks, much earlier than seen in previous models (Figure 1). Starting at 20 weeks, KPN^+/− mice manifest a striking cystic phenotype that is macroscopically and histologically visible. This cystic phenotype was accelerated and exacerbated by an episode of caerulein-induced pancreatitis, with cysts replacing over a third of the pancreas in most animals. Furthermore, the cystic lesions progressed and became dysplastic with age. Most of the pancreas was replaced by cystic lesions in mice over 31 weeks of age. Compared with KPN^+/+ mice, KPN^+/− mice had significantly more papillary cystic architecture, dysplasia, high-grade PanIN lesions, and PDAC.

Figure 1.

Timeline comparing mouse models with oncogenic Kras and heterozygous Nr5a2 inactivation reveals intriguing differences that suggest a role for NR5A2 in PDAC that extends beyond acinar cells.

Thus, in mouse models harboring a mutant Kras allele and Nr5a2 heterozygosity, large cystic lesions were virtually unique to mice with conditional heterozygous inactivation of Nr5a2 driven by a Ptf1a:Cre allele but, surprisingly, almost never observed in mice with constitutive heterozygous inactivation of Nr5a2 [5] or conditional heterozygous inactivation of Nr5a2 in pancreas epithelium driven by a Pdx1^late-Cre allele [3]. These differences raise an intriguing model that wild-type expression of NR5A2 outside the pancreas – as would be present with Ptf1a:Cre-driven inactivation but not present with constitutive heterozygous inactivation – may contribute to the formation of large cystic lesions in the pancreas.

What could explain these differences? A clue may come from the observation that conditional heterozygosity of Nr5a2, when compared with constitutive heterozygosity, leads to a more severe inflammatory response and impaired recovery following caerulein-induced pancreatitis (Figure 4 in [5]). In contrast to conditional inactivation, constitutive inactivation leads to Nr5a2 heterozygosity in additional cells including immune and stromal cells. Thus, dosage of NR5A2 in these additional cell types may modulate inflammatory responses in the pancreas. Indeed, high NR5A2 expression has been reported in infiltrating lymphocytes and neutrophils in PDAC [7]. There is evidence that NR5A2 may be involved in the maturation and normal function of T cells [8], and pharmacologic inhibition of NR5A2 has been shown to abrogate T-cell effector function and subsequently ameliorate a mouse model of hepatitis [9]. Moreover, a recent single-cell transcriptomic analysis of mouse pancreas demonstrated that induction of mutant Kras in acinar cells orchestrates T-cell infiltration and dramatic signaling changes in immune and stromal cells [10]. Put together, the responses of infiltrating immune cells and other cell types in the pancreas may contribute to the earlier inflammation, high-grade PanINs, and formation of large cystic lesions observed by Cobo et al, and these responses may require normal NR5A2 function.

If this is the case, why then are large cystic lesions not observed when Nr5a2 heterozygous inactivation is driven by a Pdx1^late:Cre allele [3], which, like the Ptf1a:Cre allele used by Cobo et al, preserves NR5A2 function in non-pancreas cells? A simple explanation may be that the Pdx1^late:Cre allele compared to the Ptf1a:Cre allele activates mutant Kras later (E12.5 versus E10.5) and in a smaller subset of pancreas cells, which may decrease levels of inflammation. Perhaps more importantly, the Ptf1a:Cre allele results in hemizygosity of Ptf1a. Reduced levels of PTF1A have been shown to accelerate Kras-driven PanIN and PDAC formation [11]. Indeed, in the context of a Kras-mutant allele, Cobo et al observe high-grade PanIN lesions in ~50% of mice at 4–7 weeks of age following Ptf1a:Cre-driven heterozygous Nr5a2 inactivation, whereas no mice with Pdx1^late:Cre-driven heterozygous inactivation of Nr5a2 have high-grade PanIN lesions (Supplementary Figure 5 in [3]). PTF1A and NR5A2 are transcription factors that collaborate in establishing and maintaining normal acinar cell identity [12]. Thus, accelerated ADM and the formation of large cystic lesions observed by Cobo et al may ensue from reduced levels of PTF1A that destabilize this transcription factor network and acinar cell identity.

Though GWAS have implicated NR5A2 in modulating the risk for PDAC, the cell types that mediate this association are unknown. While a cell-autonomous role for NR5A2 in acinar cells has been shown to be critical, the findings of Cobo et al suggest that levels of NR5A2 in cells other than pancreas epithelium may also contribute to the pathogenesis of PDAC and its precursor lesions. Future studies evaluating the effect of risk alleles on NR5A2 activity and regulation in these diverse cell populations will elaborate our understanding of PDAC pathogenesis and inform whether targeting these pathways might serve as a preventive strategy for this intractable disease.

A new model of pancreatic cystic lesions

Several precursor lesions have been associated with PDAC, including microscopic PanINs, larger and radiographically visible intrapancreatic mucinous neoplasia (IPMN), and the relatively rare mucinous cystic neoplasia (MCN). Given the need for early PDAC detection to improve survival, it is important to explore the biology and clinical significance of precursor lesions, especially those that could be detected radiographically. The mice observed by Cobo et al had large cystic lesions at 20 weeks of age with marker staining that resembled pancreato-biliary or gastric-type IPMN, and lacked the ovarian-like stroma and markers for MCN. How faithfully these cystic lesions resemble human precursors remains to be determined and would be supported by identifying mutations associated with IPMNs such as GNAS activation or RNF43 inactivation [13].

The findings by Cobo et al thus add to mouse models of pancreatic cystic lesions and raise a number of important new questions. Do these cystic lesions arise from acinar cells consistent with an ADM–PanIN–PDAC progression, or do they arise from ductal cells consistent with an IPMN–PDAC progression, as is believed from existing models? The Ptf1a:Cre allele is expressed in both epithelial cell lineages. How do these different origins and trajectories shape cancer behavior and vulnerabilities? Can one precursor lesion promote alterations and neoplasia in surrounding cells, resulting in a field cancerization effect? And how can we translate our understanding of the underlying pathogenic pathways including NR5A2 dysregulation into early detection, chemoprevention, and treatment strategies? The findings by Cobo et al add an important tool to help advance our understanding and management of this deadly disease.