Epithelial Notch signaling is a limiting step for pancreatic carcinogenesis

Animals were housed in pathogen-free conditions and maintained in facilities of the University of Michigan Comprehensive Cancer Center and this study was included in our animal use protocol, approved by the University of Michigan University Committee on Use and Care of Animals (UCUCA). P48-Cre (Ptf1a-Cre) [19] (provided by Christopher V. Wright, Vanderbilt University, Nashville, Tennessee, USA) and LSL-Kras ^G12D[20] (provided by David Tuveson, Cambridge Research Institute, Cambridge, United Kingdom) were crossed to generate KC mice, as previously described [4, 20]. ROSA ^DNMAML/+ mice [21, 22] were crossed with KC mice to generate KC;DNMAML mice. Wild type mice and animals having a combination of Kras allele and/or DNMAML, but not the Cre allele were used as controls. Allele-specific PCR on mouse tail DNA, was used to verify the presence of each allele. DNMAML induction was monitored through GFP expression in pancreas epithelium. 5 animals per cohort (≥5 mice/cohort) were aged 15 and 26 weeks before euthanasia. Tissues were collected for histopathological analysis.

Tail DNA was extracted using hot sodium hydroxide and tris (HotSHOT) DNA extraction protocol, as previously described [23]. The primers used for p48-Cre, Kras ^G12D , and ROSA ^DNMAML/+ were as follows: p48-Cre, 5’-catgcttcatcgtcggtcc- 3’ (forward) and 5’-gatcatcagctacaccagag-3’ (reverse); Kras ^G12D , 5’-agctagccaccatgagtaagtctgca-3’ (forward) and 5’-cctttacaagcgccgcagactgtaga-3’ (reverse); Rosa^DNMAML-GFP/+, 5’-aaagtcgctctgagttgttat-3’ (Rosa1), 5’-gcgaagagtttgtcctcaacc-3’ (Rosa2), and 5’-ggagcgggagaaatggatatg-3’ (Rosa3). PCR cycling conditions were as follows: p48-Cre, Kras^G12D, 95°C for 3 min, 95°C for 30 s, 60°C for 30 s, and 72°C for 45 s for 34 cycles, followed by 72°C for 5 min; Rosa^DNMAML-GFP/+, 95°C for 3 min, 95°C for 30 s, 59°C for 30 s, and 72°C for 1 min for 34 cycles, followed by 72°C for 5 min. Amplified PCR products were run on 2% agarose gels with molecular weight markers. PCR products were visualized under Alpha Innotech UV transilluminator.

Animals were administered caerulein (Sigma-Aldrich) by intraperitoneal injections in two series of 8 hourly at a concentration of 75 μg/kg over a 48 hour period, as previously described [24]. Age-matched controls were injected in parallel with experimental mice.

Pancreatic tissues from experimental and control mice were dissected and fixed overnight in 10% neutral-buffered formalin (Fisher Scientific) embedded in paraffin and sectioned (4–5 μm). The University of Michigan Cancer Center Histopathology Core performed all embedding and sectioning. Paraffin-embedded tissue sections were processed using 2 × xylene for 5 min, 2 × 100% ethanol for 5 min, 2 × 95% ethanol for 2 min, and rinsed under running deionized water for 5 min. Antigen retrieval was performed using citrate buffer (BioGeneX) in the microwave and cooled. Blocking of endogenous peroxidase activity was achieved using 3% hydrogen peroxide for 10 min, and then sections were blocked using 5% albumin from bovine serum (BSA; Sigma-Aldrich) for 30 min. Hematoxylin/Eosin (H&E), Periodic Acid Staining (PAS), Gomori trichrome, and immunohistochemistry staining was performed as previously described [5]. For a list of antibodies used, see Additional file 1: Table S1. Images were taken with an Olympus BX-51 microscope, Olympus DP71 digital camera, and DP Controller software.

For immunofluorescence, secondary-antibodies labeled with Alexa Fluor 488 (Life Technologies). Cell nuclei were counterstained with 4’,6-diamidino-2-phenylindole (DAPI; Invitrogen). The immunofluorescent images were acquired using an Olympus IX-71 confocal microscope and FluroView FV500/IX software.

Tissue for RNA extraction was stabilized through overnight incubation in RNAlater-ICE (Ambion) at −20°C, then isolated using RNeasy Mini Kit (QIAGEN) according to manufacturer’s instructions. Reverse transcription reactions were conducted using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Samples for quantitative RT-PCR were prepared with 1× Power SYBR Green PCR Master Mix (Applied Biosystems) and various primers (Additional file 1: Table S2). All primers were optimized for amplification under reaction conditions as follows: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Melting curve analysis was performed for all samples after completion of the amplification protocol. Cyclophilin and Gapdh were used as the reference gene expression controls. Hes1 primers were acquired from Applied Biosystems; Mouse Hes1 (Mm01342805_m1), Mouse Gapdh (Mm99999915_g1). Amplification was preformed using the StepOnePlus System (Applied Biosystems), and experiments were performed in triplicates. The results were calculated following the 2ΔC _p method using the StepOne Software (Applied Biosystems). A two-tailed unpaired t test was used for statistical analysis.

The histopathological analysis was performed as previously described [25]. In brief, 5 randomly selected, non-overlapping high-power images (20× objective) were taken for each slide. A minimum of 50 total acinar or ductal clusters was counted from at least three independent animals for each group. Each cluster counted was classified as acinar, ADM, PanIN1A, PanIN1B, PanIN2, and PanIN3 based on the classification consensus [26]. Prism software was used to perform statistical tests and assess statistical significance. (GraphPad; Mac version 6.0). A P values were calculated according to multiple t test and were considered significant when less than 0.05 and highly significant when less than 0.01.

To block Notch signaling during pancreatic carcinogenesis, we crossed a mouse model of pancreatic cancer, based on pancreas-specific expression of mutant Kras (p48-Cre; LSL-Kras^G12D; hereby referred to as KC) with a transgene expressing a Dominant-Negative form of Mastermind-like1 (DNMAML), which encodes for a truncated form of MAML1 (aa 13–74) fused to GFP, downstream of a LoxP-flanked stop cassette [5, 6, 21] (Figure 1D). The DNMAML-GFP fusion product is a potent inhibitor of Notch 1–4 signaling in vivo and in vitro and interferes with the NICD-CSL/RBPJ-MAML complex formation, which is essential for transcriptional activation of Notch target genes in a cell autonomous manner [18, 27] (Figure 1A). The DNMAML-GFP allows tracking recombination events in individual cells by taking advantage of the GFP fusion protein [21].

To verify the ability of DNMAML to inhibit Notch signaling in the pancreas, we harvested pancreata from 15 week old wild type, KC, and KC;DNMAML mice and analyzed the expression of Notch target genes in whole-tissue mRNA (5 mice/cohort). Expression of Hes1, Hey1, and Hey2, common targets of Notch signaling was assessed by quantitative real time polymerase chain reaction (RT-qPCR) [28‐30] (Figure 1B). We observed elevated levels of Hes1, Hey1, and Hey2 in KC mice compared to WT control samples, in accordance with previously published data showing Notch signaling upregulation in PanINs and PDA [4, 9, 10]. Importantly, the Notch target genes were downregulated in KC;DNMAML mice compared to KC. Thus, DNMAML expression successfully inhibited Notch signaling in the pancreas in vivo.

In order to determine the effects of Notch inhibition in the epithelial compartment during pancreatic carcinogenesis, we harvested tissue from age-matched KC and KC-DNMAML mice at 15 (n = 5) and 26 (n = 28) weeks of age and examined tissue histology for neoplastic progression. In 15 week-old samples we observed a significant reduction in the number of PanINs in KC;DNMAML mice compared to KC mice (Figure 1E). Thus, inhibition of Notch signaling delayed or blocked PanIN formation. Analysis of those PanINs that were present in KC;DNMAML mice did not reveal any significant histological difference compared to PanINs in KC mice. Then, we examined the pancreata of KC;DNMAML and KC mice dissected at 26 weeks of age (Figure 1F). At this time point, there was no significant difference in the number of PanINs between KC and KC;DNMAML mice. There results were consistent with Notch inhibition being insufficient to block PanIN formation, or with loss of Notch inhibition through a negative selection process. We performed histological and molecular analysis of PanINs in KC and KC;DNMAML tissues. PanIN lesions from both genotypes expressed typical markers such as Claudin18 and showed intracellular mucin staining (PAS staining), in addition to collagen deposition (Gomori Trichrome staining). To compare the proliferation index among the two genotypes, we performed Ki67 staining [31]. At 15 weeks and 26 weeks, we observed elevated Ki67 staining in both the tumor epithelium and stromal compartment, independent of genotype (Figure 1E,F). We then analyzed apoptosis, by cleaved Caspase3 staining and did not observe any difference in the two sets of mice (data not shown). Thus, once PanINs had formed, they had the expected proliferation and cell survival rate.

Histopathological analysis of de-identified slides confirmed our initial observations. Namely, KC;DNMAML samples had more acinar clusters compared to KC samples (80% and 30% respectively) and less ADM (20% and 35%, respectively) at 15 weeks. KC pancreata displayed a greater degree of disease progression with a higher proportion of PanIN1A (20%), PanIN1B (8%), and PanIN2 (1%) compared to age-matched KC;DNMAML mice (Figure 1C). At 26 weeks, we observed significantly fewer PanINs and more acini in KC;DNMAML mice compared to KC. Moreover, KC;DNMAML had lower-grade PanINs than KC samples, reflecting the delay in lesion onset (Figure 1C). Thus, PanINs were delayed, but their formation was not blocked, upon inhibition of epithelial Notch signaling. This finding might be explained by progressive loss of Notch-dependency (possibly due to compensation by other signaling pathways) or by loss of Notch inhibition due to competitive disadvantage –thus elimination- of DNMAML expressing cells.

In order to determine whether DNMAML expression had been lost over time, or whether Notch signaling was still inhibited in the epithelial compartment, we analyzed Hes1 and GFP protein localization by immunofluorescence (Figure 2). At 15 weeks, Hes1 was upregulated in KC samples but not in KC-DNMAML tissues as predicted. GFP staining overlapped with DAPI nuclear staining, suggesting that DNMAML was localized in the nuclear compartment (Figure 2A and C). At 26 weeks, Hes1 appeared further upregulated in KC samples, and we also observed Hes1 expression in the PanINs of KC-DNMAML tissues, suggesting increased Notch activity (Figure 2B and C). Moreover, we observed loss of nuclear GFP, and concurrent appearance of cytoplasmic or membrane GFP. While we can’t fully explain the mechanism underlying the latter finding, we nevertheless were able to conclude that the DNMAML-mediated inhibition of Notch signaling was lost over time, either by loss of DNMAML expression or DNMAML-expressing cells, or by inappropriate subcellular localization of the DNMAML-GFP fusion protein. Thus, the loss of phenotype in KC;DNMAML mice over time coincided with re-activation of Notch signaling. These data thus support an essential requirement for epithelial Nocth signaling during PanIN formation.

To further our analysis of KC;DNMAML mice, we investigated whether inhibition of Notch signaling affected the activation of other embryonic signaling pathways that are upregulated in PanIN formation and important for their progression, such as Hedgehog and Wnt [5, 8, 32‐35]. Thus, we harvested tissue from 15 and 26 week old KC;DNMAML and age-matched KC mice and extracted mRNA to analyze Hedgehog and Wnt associated gene expression, such as Patched-1 (Ptch1), a Hedgehog pathway component and target gene, and Wnt3a (Wnt3a), a Wnt ligand [5, 8] (Figure 2D).

In the same set of experiments, we measured Hes1 expression by qPCR, to determine the degree of Notch inhibition in the individual samples (Figure 2D). At 15 weeks, we observed a trend towards a decrease in HES1 in KC-DNMAML samples (n ≥3) when compared to KC samples, as predicted. Our data did not reach statistical significance, possibly due to the heterogeneous nature of whole tissue samples. At 26 weeks, there was no difference in HES1 expression between KC and KC-DNMAML samples, corroborating our immunostaining-based results described above. At neither time point did we observe a difference in Ptch-1 and Wnt3a expression. Thus, inhibiting Notch signaling had no effect on Hedgehog and Wnt activity.

To complement the qPCR-based experiments, we analyzed the expression of downstream effectors of Kras, Hedgehog, and Wnt signaling by immunostaining (Figure 3). To analyze the activity of Kras effector pathways, we performed immunostaining for AKT and MAPK signaling components (ERK). Immunostaining of the activated form of ERK and AKT (phospho-ERK1/2 and phospho-AKT) at 15 and 26 weeks revealed elevated p-ERK and p-AKT in PanINs, independent of genotype. While we did not observe changes in Kras effector pathways, the progressive loss of Notch inhibition in our model does not allow a rigorous epistatic analysis of these pathways.

To examine whether Notch inhibition in the tumor epithelium affected the expression of Hedgehog ligands, we performed immunostaining against Sonic Hedgehog (Shh). Shh is commonly upregulated in PanIN lesions [5, 32, 33, 36]. Studies have shown that upregulation of Hedgehog signaling promotes tumorigenesis [5, 32, 33, 36]. We detected Shh in the PanIN epithelium and not in the surrounding stroma, in both KC and KC-DNMAML; thus the expression of the Hedgehog pathway ligand Shh was not regulated by Notch signaling. To assess Wnt signaling activity in vivo, we performed immunostaining against beta-catenin. We did not observe changes in expression or subcellular localization of beta-catenin in the PanIN epithelium comparing KC and KC-DNMAML mice.

In our previous set of experiments, KC and KC;DNMAML mice developed PanINs spontaneously, over time. Several studies have reported that, in mice, the induction of acute pancreatitis synergizes with oncogenic Kras to induce rapid and extensive PanIN formation [6, 24, 37‐39]. To investigate the potential requirement for active Notch signaling during pancreatitis-induced carcinogenesis, we administered a cholecystokinin (CCK) agonist (caerulein), which induces acute pancreatitis in mice (8 injections/day; 2 day treatment). We collected tissues from age-matched wild type (control), KC, and KC;DNMAML mice 2, 3, and 4 weeks following acute pancreatitis (Figure 4A). Acute pancreatitis leads to acinar damage, mainly represented by ADM, edema of the tissue, and infiltration of inflammatory cells both in WT and KC mice. However, while WT mice rapidly recover, with complete tissue repair usually observed within a week, KC mice are unable to undergo tissue repair (Figure 4B). In contrast, in KC mice, the pancreas becomes progressively fibrotic and the ADM becomes more extensive over the course of the first week after treatment. Then, the pancreas forms tissue-wide PanINs. The WT and KC cohorts in our experiment behaved as expected.. Similarly, KC;DNMAML developed PanINs, although they retained more acinar clusters 2 and 3 weeks after pancreatitis, compared to KC animals (Figure 4B,C). However, by 4 weeks after pancreatitis the two cohorts were indistinguishable. Based on our results from mice aging in absence of pancreatitis, this result might reflect progressive loss of Notch inhibition paralleling PanIN formation.