Loss of Jag1 cooperates with oncogenic Kras to induce pancreatic cystic neoplasms

Jag1 functions as a tumor suppressor in delaying pancreatic cancer precursor, while at the same time, promotes a phenotypic switch from benign cystic lesions to invasive carcinoma of the pancreas.


Introduction
Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal malignancies. Investigation of the PDAC precursors, their cellular origins, initiation and progression is of great importance for early detection and intervention for the disease. Three distinct precursor lesions of PDAC have been identified, including pancreatic intraepithelial neoplasm (PanIN), intraductal papillary mucinous neoplasm (IPMN), and mucinous cystic neoplasm (MCN). PanINs are the most important and common precursors of PDAC. They are microscopic pancreatic lesions associated with ducts, but interestingly, they can originate from acinar cells (1). IPMNs belong to the heterogeneous group of pancreatic cystic neoplasms. Human IPMNs display distinctive intraductal growth, and evidence from mouse models suggests that IPMNs most likely arise from the progenitor niche of the pancreatic ductal epithelium (1). MCNs are the most infrequent precursor lesions of PDAC and are characterized by the presence of progesterone receptor (PR)-and estrogen receptor (ER)positive ovarian type stroma. A wide variety of genetic changes have been identified in precursor lesions with increased frequency in advanced cases, including mutations of KRAS, p16INK4A/CDKN2A, SMAD4, and TP53 genes, which are largely overlapping in PanINs and IPMNs, and detected less frequently in MCNs (2). The mechanisms underlying the differentiation of different precursor lesions remain to be elucidated.
Notch signaling plays important yet complex roles in PDAC. Whereas Notch2 was shown to be required for progression of PanIN to PDAC (3), loss-and gain-of-functions of Notch1 both rendered acinar cells more susceptible to Kras-induced PanIN formation and progression (4,5,6). We recently discovered a PDAC-suppressive function for Lfng, a glycosyltransferase that modifies Notch receptors to enhance Delta ligand-mediated Notch activation and inhibit Jagged ligand-mediated Notch signaling, in the Kras LSL-G12D/+ ;Pdx1-Cre mouse model, in which Kras activation started in the embryonic pancreas (7). Deletion of Lfng caused sustained Notch3 activation (7), and Notch3 was associated with Jag1 expression in human PDACs (8), suggesting that Lfng may inhibit Jag1-Notch3 signaling to suppress PDAC development.
Here we deleted Jag1 in Kras LSL-G12D/+ ;Pdx1-Cre to determine its role in Kras-driven pancreatic tumorigenesis. Deletion of Jag1 accelerated early stage lesions including acinar-to-ductal metaplasia (ADM) and low-grade PanINs. Strikingly, these precursor lesions were diverted along a differentiation pathway towards cystic neoplasms instead of PDAC. Specifically, loss of Jag1 results in the formation of non-mucinous cysts reminiscent of serous cystic neoplasms (SCNs) in humans, and less frequently, IPMN. This study revealed a Jag1-controlled phenotypic switch between PDAC and largely benign cystic neoplasms, which appeared to be associated with differential expression of Lkb1 and Sox9 in pancreatic ductal cells.

Deletion of Jag1 accelerates Kras-driven ADM and PanIN formation
We deleted Jag1 in the mouse pancreas starting from embryonic stage using Pdx1-Cre. Quantitative RT-PCR showed a robust decrease in the level of Jag1 mRNA in the Jag1 flox/flox ;Pdx1-Cre (hereafter referred to as Jag1 KO ) pancreas, which was further reflected by decreased expression of the Notch target genes Hes1 and Hey1 (Fig 1A). Western blot analysis confirmed lower level of Jag1 protein in the pancreas of Jag1 KO mice compared with the Jag1 flox/flox mice ( Fig 1B). Similar to a previous report (9), Jag1 ablation in the pancreatic lineage caused gradual loss of acini and replacement with adipocytes, accompanied by dilation of ducts and ductal desmoplasia ( Fig 1C). Interestingly, we also found large cysts in 3 out of 24 Jag1 KO mice older than 6 mo. Next, we generated Kras LSL-G12D/+ ;Jag1 flox/flox ;Pdx1-Cre mice (hereafter referred to as KJC) to determine the impact of Jag1 loss on Kras G12D -induced pancreatic cancer initiation and progression. KJC mice developed precancerous lesions shortly after birth, without the replacement of acinar cells by adipocytes. At 1 mo of age, the pancreas in KJC mice had in average about 67% of the acinar compartment replaced by abnormal ductal structures, whereas Kras LSL-G12D/+ ;Pdx1-Cre (hereafter referred to as KC) mice showed no or very few lesions at this age (Fig 1D and E). The lesions in KJC mice were positive for cytokeratin 19 (CK19), resembling ADMs or low-grade PanINs ( Fig 1F). In addition, KJC pancreas displayed extensive desmoplasia, associating with robust expression of vimentin and smooth muscle actin α (SMA), which were negative in the acinar compartment of KC mice ( Fig 1F). It has been shown that clinical specimens of pancreatic cancer express elevated levels of connective tissue growth factor (CTGF), and this correlates with the extent and intensity of desmoplasia (10). Indeed, CTGF mRNA level in KJC pancreas is five to six fold higher than that in KC pancreas ( Fig 1G).
ADM occurs in response to pancreatic tissue injury, and similar metaplastic changes occur when exocrine cells are isolated and cultured, accompanied by up-regulation of Notch pathway genes (11). Stimulation of Notch by Jag1 diminished the proliferation of cultured metaplastic exocrine cells, whereas inhibition of Notch signaling had an opposite effect. This effect seemed to be Hes1-independent and mainly coincided with Hey1 and Hey2 expression (11). These observations are in agreement with our result showing that deletion of Jag1 accelerated ADM formation in vivo. Coincidently, deletion of Jag1 also caused decreased expression of Hey1 and Hey2, but not that of Hes1 ( Fig 1G). Thus, Jag1-mediated Notch signaling appears to suppress ADM and PanIN formation caused by Kras G12D expression in the pancreas starting from developmental stage.
A phenotypic switch from ductal adenocarcinoma to cystic neoplasms by Jag1 deletion We examined proliferation and apoptosis of pancreatic cells by immunostaining for Ki67 and cleaved caspase 3, respectively. At 1 mo of age, KJC mice showed increased proliferation of ductal cells compared to KC mice. Interestingly, the ducts in KJC mice also contained more apoptotic cells than KC mice (Fig 2A and B). We performed histological examination throughout the pathological course of pancreatic tumorigenesis in KJC mice. These mice developed PanIN-1A and PanIN-1B lesions as early as postnatal day 18 (Fig 2C-F). Dilation of the ductal structures and development of desmoplasia also started at this time ( Fig 2E). Notably, the overall morphology of the main duct remained normal, and the vast majority of PanINs were not associated with large ducts (Fig 2F), suggesting that PanINs originated from the acinar compartment, as reported in other Kras G12D -driven mouse models (12,13). These lesions continued to grow, becoming cystic neoplasms by 2-3 mo of age ( Fig 2M). Characterization of cystic lesions in KJC mice (see below) identified two histological subtypes: type I, reminiscent of SCN ( Fig 2G and H), and type II, resembling IPMN (Fig 2I and J). At 3 mo and older, 34.5% KJC mice (n = 29) and 10.5% Kras LSL-G12D/+ ; Jag1 flox/+ ;Pdx1-Cre mice (n = 19) showed distended abdomen because of large-size pancreatic cysts containing clear to strawcolored fluid (Fig 2N), highlighting the importance of the gene dosage in this SCN/IPMN phenotype. Histological examination of the pancreas in 28 KJC mice (≥3 mo) found nine cases of small cystic neoplasm (coexisted with PanIN), 14 cases (50%) of SCN or IPMN lesions, and only one case (3.6%) of PDAC (Fig 2K and L). For comparison, there were two cases (6.0%) of IPMN-like lesion (no SCN-like lesion) and up to 11 cases (33%) of PDAC in 33 KC mice older than 3 mo ( Fig 2O and Table 1). Thus, deletion of Jag1 was associated with significantly increased incidence of cystic neoplasms and decreased incidence of PDAC (χ 2 test, P < 0.0001), suggesting a switch from ductal adenocarcinoma to cystic neoplasms.
The vast majority of KJC mice did not die from pancreatic cystic neoplasms. However, almost all of them developed ulcerating facial skin lesions necessitating euthanasia (Fig S1), which precluded analysis of pancreas-related death. A few KJC mice also developed tumors near the anus. Jag1 KO mice did not show any skin lesions, whereas a few KC mice showed skin lesions at variable sites (data not shown). It has been shown that Pdx1 is physiologically expressed in the adult mouse epidermis, and in vitro analysis revealed differentiation-dependent expression of Pdx1 in terminally differentiated keratinocytes (14). Although we cannot rule out the possibility of the skin phenotype being related to a diabetic syndrome, the fully penetrated skin phenotype in KJC mice suggests that loss of Jag1-mediated Notch signaling may cooperate with oncogenic Kras to induce skin carcinogenesis. Interestingly, deletion of Notch1 also increased susceptibility to Kras G12D -induced skin carcinogenesis with Pdx1-Cre (14).

Characterization of cystic neoplasms in KJC mice
We characterized two histological subtypes of cystic neoplasm in KJC mice by immunohistochemistry. As expected, duct-lining epithelial cells in both types stained positive for CK19 (Fig 3A and D). Alcian blue staining as well as Muc5AC immunostaining revealed abundant apical mucin in cystic epithelial cells of type II but not type I lesions (Fig 3B, C, E, and F). IPMN and MCN are two subtypes of human pancreatic cystic neoplasms with mucinous differentiation (15). MCNs are distinct from IPMNs by an underlying ovarian-like stroma, which often shows nuclear staining of PR and ER, as well as expression of SMA and desmin (16). Although type I lesions lack the columnar mucin-producing epithelium typically found in MCN, their ovarian-like stroma show weak to modest ER expression and is SMA-positive ( Fig 3G-I), similar to the MCN-like lesions in female mice overexpressing Wnt1 and Kras G12D , which also lacked the typical mucinous epithelium (17). The serous cysts lined by cuboidal or flattened non-mucinous epithelial cells in the type I lesion are reminiscent of SCN rather than MCN. Human MCNs are presented almost exclusively in middle-aged women (96.5% patients are females) (18), whereas SCNs in humans exhibit female predilection   Alcian blue staining in pre-weaning KJC pups showed mucinous ductal structures resembling low-grade PanINs or small-size IPMNs, as well as non-mucinous ductal lesions of relatively large size, which could be the precursor of SCN-like lesions (Fig 3M-O). To explore this phenotype further, we crossed Rosa LSL-YFP reporter into Jag1 KO and KJC mice. Whereas anti-YFP immunostaining in the pancreas of Rosa LSL-YFP ; Jag1 flox/flox mice was completely negative (Fig S2A), Rosa LSL-YFP ; Jag1 flox/flox ;Pdx1-Cre pancreas showed YFP-positive cells in islets, but not in the adipocytes that have replaced acinar cells (Fig S2B and E), indicating these adipocytes are not derived from the pancreatic lineage. In Rosa LSL-YFP ;Kras LSL-G12D/+ ;Jag1 flox/flox ;Pdx1-Cre pancreas, all ductal lesions were YFP-positive, indicating that they have originated from Pdx1expressing progenitors or their descendants (Fig S2C). An isotype control of the immunostaining in the same tissue was negative ( Fig S2D). Interestingly, some stromal cells were YFP-positive, suggesting these cells may have undergone epithelial-mesenchymal transition ( Fig S2F).

Jag1 expression is lost in SCN but retained in PDAC
It has been shown that Jag1 is expressed throughout the pancreatic epithelium at embryonic day 12.5 and gradually restricted to islet clusters and ducts thereafter. By postnatal day 3, Jag1 expression is confined to ductal cells (9,20). We performed Jag1 immunohistochemistry in KJC mice at various ages with different types of lesions. At 1 mo of age, KJC mice had developed extensive ADM and PanIN lesions. Although Jag1 was undetectable by immunohistochemistry in the ductal cells in wild type mice at this age (Fig 4A), Jag1 expression was up-regulated in subsets of ADM lesions ( Fig 4B) and abnormal ducts ( Fig 4C) in KJC mice, indicating that pancreatic deletion of Jag1 was incomplete in these mice. By 2-3 mo of age, many KJC mice had developed SCN-like lesions, where ductal cells lining the SCN were completely Jag1-negative, whereas blood vessels in the same section stained positive for Jag1 (Fig 4D and G). A few KJC mice formed IPMNlike lesions at this age, which exhibited no or a very low level of Jag1 expression (data not shown). Interestingly, high level Jag1 expression was observed in ductal cells of a rarely formed PDAC from an adult KJC mouse (Fig 4E and H). Thus, Jag1 expression was lost in cystic neoplasms but retained in the invasive carcinoma in KJC mice. This could be explained by the mosaic deletion of Jag1 in these animals, as the Pdx1-Cre strain is known to display mosaic expression of Cre recombinase throughout the pancreas epithelium (21). To confirm the functionality of Jag1 expression, we performed immunostaining for Jag1 and Notch1 (or Notch2) on consecutive sections of KJC pancreas. There was no overlapping between Jag1 expression and presence of cleaved Notch1 (data not shown). However, cytoplasmic staining of Jag1 and nuclear staining of Notch2 were observed at the same location of consecutive sections (Fig 4F and I), suggesting that Jag1 may activate Notch2 during Kras-initiated pancreatic cancer development. These findings are in corroboration with a previous report showing that Notch2 functions in ductal cells and PanIN lesions and are required for PanIN progression and malignant transformation, and that deletion of Notch2 combined with Kras G12D expression resulted in MCN-like cystic lesions in a subset of mice (3).
We analyzed JAG1 expression in two published human gene expression data of matching pairs of PDAC and adjacent non-tumor tissue (22,23). In both data sets, JAG1 expression is significantly higher in tumors than in non-tumor tissues ( Fig 4J). Moreover, analysis using two of the human data sets containing survival information found significant association between high JAG1 expression and poor overall survival among pancreatic cancer patients (Fig 4K). We performed PathwayMapper (24) analysis using The Cancer Genome Atlas (TCGA) pancreatic adenocarcinoma data set hosted in the cBioPortal database. Six of the Notch signaling pathway genes, including JAG1, JAG2, NOTCH3, MAML1, MAML2, and MAML3, were upregulated in this data set (Fig 4L). Expression heat map of these genes showed an active Jag1-Notch signature in the JAG1-high specimens (Fig 4M). Previous immunohistochemical study found that JAG1 expression was significantly higher in intraductal papillary mucinous carcinoma than in intraductal papillary mucinous adenoma and was significantly related to recurrence, suggesting that JAG1 levels also reflect IPMN aggressiveness (25). Thus, human patient data corroborate our findings in mice, in which we found that Jag1 expression is required for the progression of precancerous lesions to PDAC, and Jag1 levels determine the aggressiveness of those lesions.

Loss of Sox9 expression in cystic neoplasms in KJC mice
Sox9 is a ductal fate determinant and a recent study suggests that Sox9 may prevent dedifferentiation of pancreatic ductal cells and   IPMN formation (26). Interestingly, Sox9 is a direct target of Jag1mediated Notch signaling in the biliary system (27,28), and is regulated by Hes1 in the pancreas during pancreatic tumorigenesis (29). It is conceivable, therefore, that Sox9 expression may be altered in the Jag1-deficient pancreas. Indeed, immunostaining in KJC mice found that ductal cells lining SCN-like lesions were completely negative for Sox9, and IPMN-like lesions showed no or very weak Sox9 staining, whereas a rarely formed PDAC showed clear nuclear staining of Sox9 in ductal tumor cells (Fig 5A-C). Likewise, Notch2 nuclear staining was readily detected in PDAC cells, but not in the ductal cells lining the SCN-like or IPMN-like lesions (Fig 5D-F). These data suggest that cystic neoplasms in KJC mice have lost Sox9 expression, associated with loss of Notch2 activation. Western blot analysis showed significantly lower level of Sox9 protein in KJC mice with large cystic lesions than in KC mice at 4-6 mo of age (Fig 5G), suggesting that deletion of Jag1 is effective at down-regulating Sox9 expression during Kras G12D -driven pancreatic tumorigenesis. Interestingly, analysis in TCGA pancreatic adenocarcinoma data set found a positive correlation between JAG1 and SOX9 expressions (Fig 5H), and high SOX9 mRNA level is associated with poor overall survival, whereas low SOX9 expression is associated with significantly longer survival (Fig 5I and J).

Jag1 regulates Lkb1 expression in pancreatic ductal cells
LKB1 gene inactivation has been involved in human IPMNs (30), and it is more common in IPMNs than in PDACs (31). Deletion of Lkb1 in the mouse pancreas resulted in progressive acinar cell degeneration, ADM, and development of serous cystadenomas (32). Mice with an inhibitory knock-in allele of Lkb1 also showed cystic structures in the pancreas (33), and postnatal duct-specific deletion of Lkb1 caused ductal dilation and ADM (34). Lkb1 haploinsufficiency cooperated with Kras G12D to promote pancreatic cancer, some of which exhibits a cystic morphology (35). Moreover, inactivation of Lkb1 and expression of Kras G12D in pancreatic ducts synergized to induce IPMN (36). Given the similarities between these mice and KJC mice, we wondered whether Jag1 could regulate Lkb1 expression under Kras G12D -driven pancreatic tumorigenesis conditions. Immunohistochemistry showed Lkb1 expression in a subset of PanIN ductal cells in KC mice (Fig 6A), whereas in KJC mice, Lkb1 staining is positive in very few cells of IPMN-like lesions and completely negative in SCN-like lesions (Fig 6B and C). Indeed, quantitation of Lkb1 + epithelial cells showed a significant decrease in KJC mice with SCN-like lesions compared to KC mice as well as KJC mice harboring IPMN-like lesions (Fig 6D). Although statistically insignificant, quantitative RT-PCR found decreased Lkb1 mRNA expression in the pancreas of KJC mice compared with KC mice at 2-3 mo of age (Fig 6E). We isolated primary cells from KJC mice to directly test whether Jag1 regulates Lkb1 expression in pancreatic cells. Treatment of these cells with exogenous Jag1 caused up-regulation of the Notch target genes Hes1 and Hey1 as well as Lkb1 (Fig 6F). Treatment with the gamma secretase inhibitor (GSI) that inhibits Notch signaling had no effect on Lkb1 expression despite down-regulation of Hes1 and Hey1 (Fig 6G). Because Jag1 was deleted in the vast majority of the pancreatic cells in KJC mice with SCN-like lesions, Notch signaling decreased by GSI in these cells is most likely Jag1-independent. The fact that GSI had no effect on Lkb1 expression, whereas treatment with Jag1 increased Lkb1 in these cells suggests that Lkb1 is regulated by Jag1-dependent Notch signaling. To test this hypothesis, primary cells from KJC mice were treated with Jag1 together with GSI (or DMSO as control). Indeed, the presence of GSI resulted in decreased mRNA levels of Hes1 and Hey1, as well as a modest but significant decrease in Lkb1 mRNA level ( Fig 6H). Next, we tested the effect of Jag1 or GSI in human PDAC cell lines. Treatment with exogenous Jag1 caused a drastic increase of LKB1 mRNA in Miapaca2, and less significantly in Panc1 cells (Fig 6I). Conversely, incubation with GSI caused a decrease of LKB1 mRNA in Panc1, but no effect in Miapaca2 (Fig 6J). JAG1 expression is lower in Miapaca2 than Panc1 (https://portals.broadinstitute.org/ccle/page?gene=JAG1). This may explain why adding Jag1 caused a more significant increase of LKB1 expression in Miapaca2 than in Panc1, whereas treatment with GSI down-regulated LKB1 in Panc1 but not in Miapaca2. Collectively, these data suggest that Jag1 may influence pancreatic tumorigenesis in part through the regulation of Lkb1.
Ductal cell-specific deletion of Jag1 in the adult pancreas does not lead to cystic neoplasms Pancreatic ductal cells have been identified as cell-of-origin of IPMN (26,37,38). We tested whether deletion of Jag1 in conjunction with Kras G12D expression in the ductal cells of adult pancreas would cause IPMN. For this, we crossed Jag1 flox/flox or Kras LSL-G12D/+ ;Jag1 flox/flox with Sox9-CreER, which express a tamoxifen-inducible Cre recombinase in the ductal lineage. For simplicity, these mice will be referred hereafter as Jag1 KO-Duct and KJC Duct mice, respectively. We treated these mice with tamoxifen at 1 mo of age and monitored for tumor development up to 4 mo posttamoxifen administration. Histological examination of the pancreas in five KJC Duct mice failed to show any IPMN or SCN lesions, and only one mouse showed a single PanIN lesion (Fig S3C and D). Noteworthily, ductal cell-specific deletion of Jag1 alone in adult mice (Jag1 KO-Duct ) caused no morphological abnormality in the pancreas (Fig S3A and B), suggesting that loss of acini, dilation of ducts, and ductal desmoplasia seen in Jag1 KO mice ( Fig 1C) were likely due to deletion of Jag1 in pancreatic progenitor cells during embryonic stage. Collectively, these data reveal that deletion of Jag1 affects pancreas homeostasis and cooperates with Kras G12D to induce SCN or IMPN when this genetic alteration occurs during pancreatic development. Interestingly, it has been reported that epithelial cells in SCNs are ultra-structurally similar to centroacinar cells (19) and that Jag1 expression is required for the maintenance of centroacinar cells as an environmental niche in the developing pancreas (39). These findings together with our genetic studies using mouse models suggest that primitive cells in the developing pancreas may serve as cell-of-origin of cystic neoplasms, including SCNs and IPMNs. SCN-like lesions may have arisen from unidentified progenitor cells with complete loss of Jag1 and concurrent Kras activation.

Discussion
This study showed that deletion of Jag1 accelerated Kras G12D -mediated ADM and PanIN development, and that expression level of Jag1 dictates the progression of preneoplastic lesions in the pancreas: complete and partial loss of Jag1 resulted in SCN-like and IPMN-like cystic neoplasms, respectively, whereas retained Jag1 expression led to the progression into PDAC. From a mechanistic perspective, Jag1 may regulate the expression of Lkb1 and Sox9, both of which play instrumental roles in Kras G21D -driven pancreatic tumorigenesis. Jag1 expression was up-regulated in ADM/PanIN lesions, and deletion of Jag1 accelerated ADM/PanIN formation in KJC mice, raising the possibility that Jag1 may suppress ADM initiation and progression to PanIN in the acinar compartment. On the other hand, abnormal ducts and PDAC displayed high level Jag1 expression, and deletion of Jag1 resulted in very low incidence of PDAC, suggesting that Jag1 may be required for advancement from PanIN to PDAC. The apparent dual role of Jag1 in pancreatic cancer may be dependent on the stage of cancer initiation and progression, or the cell-of-origin of the lesion. One of the limitations of this study is the use of Pdx1-Cre in the modeling of pancreatic cancer. Pdx1-Cre mediates expression of Kras G12D in all lineages of the pancreas starting from embryonic stage. Activation of oncogenic Kras at this stage does not mimic the real situation paralleling PDAC, an elderly disease in humans. In addition, deletion of Jag1 during organogenesis may have caused developmental defects of the pancreas (40), thereby setting up a precondition for Krasinduced pancreatic cancer initiation and progression. Future studies using inducible CreER systems in adult mice will be required for the delineation of Jag1 functions in the pathogenesis of acinaror ductal-originated pancreatic cancer.
Lkb1 is mainly described as a tumor suppressor, whereas overall Jag1 is shown to be pro-tumorigenic here. Paradoxically, we showed that Jag1 positively regulates Lkb1 expression in pancreatic cells. As discussed above, deletion of Jag1 accelerated ADM/PanIN in KJC mice, suggesting a tumor-suppressive role of Jag1 in the early stage of tumorigenesis. Loss of Jag1 led to the development of cystic lesions, which appears to involve down-regulation of Lkb1. Thus, both Jag1 and Lkb1 may function as a tumor suppressor in the pathogenesis of pancreatic cystic neoplasm. Germline LKB1 loss-of-function mutations are responsible for Peutz-Jeghers syndrome, a disease characterized by a predisposition to gastrointestinal neoplasms. Lkb1 deletion in the mouse pancreas caused serous cystadenomas, a tumor type associated with Peutz-Jeghers syndrome (32). Interestingly, Lkb1-deficient pancreas exhibits similar phenotype as the Jag1-deficient pancreas, except that the latter requires Kras G12D to induce cystic lesions. Moreover, deletion of Lkb1 in the liver resulted in bile duct paucity leading to cholestasis (41), similar to Alagille syndrome, an autosomal dominant disorder caused predominantly by mutations in JAG1. There has been mutual regulation of Lkb1 and Notch signaling in the biliary system (41). Thus, crosstalk between Jag1-mediated Notch signaling and Lkb1 appears to function in both pancreatic and biliary tracts.
IPMN lesions are thought to arise from the progenitor niche of the ductal epithelium. In this study, we used Sox9-CreER, which has been shown to label about 70% of pancreatic ductal cells in adult mice (42), to induce the deletion of Jag1 and expression of Kras G12D in adult pancreatic ducts. With the same Sox9-CreER, concurrent Lkb1 deletion and Kras activation in the ductal epithelium resulted in IPMN in adult mice (36), whereas deletion of Jag1 combined with Kras G12D expression in the ductal epithelium failed to induce cystic neoplasm. The negative result suggests that the phenotype seen with the Pdx1-Cre is likely due to Jag1 being deleted during development in the progenitor cells of the pancreas. However, we could not rule out the possibility of Jag1 playing a role in the cells of the ductal lineage. Additional insults including duct obstruction or mutations in other genes such as p53 may be required to initiate neoplasia from ductal cells. Future studies using Sox9-CreER-mediated Jag1 deletion/ Kras G12D expression in conjunction with pancreatic duct ligation or p53 deletion may determine whether Jag1 plays a role in ductaloriginated pancreatic cancer. Finally, the results from this study suggest that SCN may have arisen from a progenitor in the acinar compartment, which should be tested through the deletion of Jag1 with lineage-specific inducible CreER system.

Quantitative RT-PCR
Total RNA was extracted with RNeasy Mini Kit (QIAGEN) and reversetranscribed by iScript cDNA synthesis kit (Bio-Rad). PCR was performed using QuantiTect SYBR Green PCR Kits (QIAGEN) with BioRad CFX96 qPCR System. The experiment was performed in triplicate, and results were normalized with the expression level of Gapdh and compared with the rest. (J) Overall survival in pancreatic cancer patients with SOX9 expression lower than 1× SD below the mean, in comparison with the rest. ****P < 0.0001 (t test). Scale bars: 12.5 μm. Source data are available for this figure. presented as mean ± SEM. Primer sequences for mouse Lkb1, Ctgf, Jag1, Hes1, Hes5, Hey1, and Hey2 have been described previously (7,9,44,45).

Cell culture
Miapaca2 and Panc1 were purchased from ATCC. Primary pancreatic cells were isolated from KJC mice with SCN-like lesions. Briefly, pancreatic tissue was rapidly collected, minced with blades, and plated on a 100-mm plate with DMEM and 10% FBS for 24 h. The supernatant was discarded and attached cells were allowed to grow until confluency. Cells were seeded into a six-well plate precoated with recombinant Jagged1-Fc chimera protein or recombinant IgG1 Fc protein as control (R&D Systems) at 10 μg/well in 100 μl PBS, or treated with GSI (MK-0752, 20 μM) or vehicle control (DMSO) for 48 h.

Gene expression analysis of human data sets
Data sets used for JAG1 expression analysis were downloaded from the International Cancer Genome Consortium (PDAC-AU, n = 188) and GEO (GSE28735, n = 90, and GSE16515, n = 32). Clinical follow-up data with overall survival information were only available for GSE28735(n = 42) and International Cancer Genome Consortium (PDAC-AU; n = 188), which were used to complete survival analysis. Expression data were normalized using Robust Multi-Array Average measure from R Bioconductor package "affy." Differential expression of JAG1 in normal and tumor tissue was analyzed by one-way ANOVA. For the survival analysis, Kaplan-Meier survival curves were generated with the upper quartile as cutoff. The pancreatic adenocarcinoma data set (TCGA, Firehose Legacy; n = 186) hosted in cBioPortal (https://www.cbioportal.org) was used for the PathwayMapper analysis (24) and gene expression heat map with its online tool. Correlation and survival analysis related to SOX9 expression was performed using the same data set and online tools in cBioPortal.

Statistics
Statistical analyses were performed using Prism version 8.3.1 (GraphPad Software). The data are presented as the mean with SEM. Survival was analyzed using the Kaplan-Meier method and compared by the logrank test. Correlations analysis was performed using chi-square test. Two-group comparisons were analyzed using two-tailed t test. P-value of 0.05 or less was considered statistically significant.