SUMMARY
Activating mutations in Kras are nearly ubiquitous in human pancreatic cancer and initiate precancerous pancreatic intraepithelial neoplasia (PanINs) when induced in adult murine acinar cells. PanINs normally take months to form, but can be rapidly induced by genetic deletion of acinar cell differentiation factors such as Ptf1a, suggesting that loss of mature cell identity is a rate-limiting step in pancreatic tumor initiation. Using a novel genetic mouse model that allows for independent control of oncogenic Kras and Ptf1a expression, we demonstrate that maintained activity of Ptf1a is sufficient to eliminate Kras-driven tumorigenesis, even in the presence of tumor-promoting inflammation. Furthermore, reintroduction of Ptf1a into established PanINs reverts their phenotype in vivo. Our results suggest that reactivation of an endogenous differentiation program can prevent and reverse oncogenesis in cells harboring tumor driving mutations, thus introducing a novel paradigm for solid tumor prevention and treatment.
INTRODUCTION
Although pancreatic ductal adenocarcinoma (PDAC) is named for its duct-like characteristics, we and others have shown that the signature driver mutation of this cancer, oncogenic Kras, induces tumorigenesis when introduced to exocrine acinar rather than duct cells (De La O et al., 2008; Guerra et al., 2007; Habbe et al., 2008; Ji et al., 2009; Kopp et al., 2012). Importantly, tumor initiation from Kras-mutant acinar cells involves a trans- or de-differentiation process, in which acinar-specific genes are downregulated and duct markers upregulated, that we refer to as “reprogramming” (De La O et al., 2008). Acinar cell-specific gene expression is normally driven by the transcription factor Ptf1a, which itself is downregulated during reprogramming to PDAC precursor lesions known as pancreatic intraepithelial neoplasia (PanIN) (De La O et al., 2008; Krah et al., 2015). Together with several cooperating transcription factors, Ptf1a is essential for maintaining mature acinar identity and restraining Kras-mediated tumorigenesis (Hoang et al., 2016; Krah et al., 2015; Shi et al., 2009; von Figura et al., 2014). How loss of Ptf1a promotes tumor development is not yet known, but could be mediated by changes in the microenvironment: acute deletion of Ptf1a upregulates a pro-inflammatory transcriptional program (Krah et al., 2015), and inflammation is itself known to promote PDAC (Krah and Murtaugh, 2016). Alternatively, the Ptf1a-driven transcriptional program may act cell-autonomously to suppress the effects of oncogenic Kras; in this scenario, the tumor-promoting effects of inflammation may be mediated via downregulation of Ptf1a (Molero et al., 2007). To directly test our hypothesis that enforcing acinar cell differentiation inhibits PDAC, we established an experimental system in which Ptf1a expression can be sustained even in the presence of oncogenic Kras and inflammatory injury. Our results indicate that downregulation of Ptf1a is essential for both basal and inflammation-driven PDAC initiation, and that reintroduction of Ptf1a into established tumor precursors reverts their phenotype in vivo.
RESULTS
A novel mouse model to independently control KrasG12D and Ptf1a expression
We established a mouse model permitting independent regulation of KrasG12D and Ptf1a: by tamoxifen (TM)-dependent excision of floxed STOP cassettes, Ptf1dCreERT induces expression of both KrasG12D and rtTA, the reverse tetracycline transactivator protein (Belteki et al., 2005) (Figure 1A-C). rtTA subsequently activates a tetO-Ptf1a transgene in a doxycycline (DOX)-inducible manner (Figure. 1D) (Willet et al., 2014). Expression of rtTA and tetO-Ptf1a can be monitored by their linked co-expression cassettes, IRES-GFP and IRES-LacZ, respectively. Staining for B-galactosidase (βgal) activity demonstrated acinar-specific activation of tetO-Ptf1a within 24 hours of DOX administration (Figure. 1E-H). Importantly, tetO-Ptf1a expression alone had no detectable effect on pancreas histology (Figure. 1K-L). These results indicate that both the rtTA and transgenic Ptf1a can be rapidly induced, specifically within Pfla-expressing cells, following TM and DOX treatment, respectively.
Sustained Ptf1a expression prevents KrasG12D-mediated pancreatic oncogenesis
To test whether sustained Ptf1a blocks PDAC initiation, we subjected control, KrasG12D and KrasG12D+tetO-Ptf1a mice (Supplementary Table 1) to high-dose TM and an 8-week (8W) chase of continuous DOX (Figure 2A; Supplementary Figure 1). All mice harbored PtfldCreERT and R26LSL-rtTA, and exhibit uniform Cre recombination between groups (Supplementary Figure 2). After 8W, KrasG12D pancreata exhibited large areas of acinar cell loss and precancerous PanIN formation, which were greatly reduced in KrasGI2D+tetO-Ptf1a mice (Figure 2B-D). Regions of PDAC initiation were highlighted by expression of the ductal marker CK19 (Figure 2E-G) and mucin staining with Alcian blue (Figure 2H-J), revealing that tetO-Ptf1a expression dramatically reduced acinar cell transformation (Figure 2N). These results suggest that maintaining Ptf1a expression inhibits acinar cell transformation and prevents initiation of pancreatic tumorigenesis.
As rare PanINs were still generated in KrasG12D + tetO-Ptf1a pancreata, we compared these lesions to PanINs induced by KrasG12D alone. Neither the amount of proliferation (Figure 2K-M, O) nor the number of Ptf1a+ cells within PanINs differed between genotypes (Figure 2P-Q, V), suggesting that PanINs were similar in both groups. We therefore postulated that PanINs in KrasG12D +tetO-Ptf1a mice were “escapers,” recombining KrasG12D to initiate tumorigenesis, but not R26LSL-rtTA, which is required for sustained Ptf1a expression. To test this, we determined the frequency of CK19+ PanINs containing GFP+ cells. Most KrasG12D PanINs co-expressed CK19 and GFP, indicating derivation from acinar cells recombining both the KrasG12D and R26LSL-rtTA loci (Figure 2R, W). In contrast, almost all KrasG12D + tetO-Ptf1a PanINs were broadly GFP-negative (Figure 2S, W), indicating that they failed to recombine R26LSL-rtTA. Thus, incomplete Cre-based recombination explains residual PanIN formation in KrasG12D + tetO-Ptf1apancreata. βgal staining further confirmed that KrasG12D + tetO-Ptf1a PanINs contained only very rare cells expressing tetO-Ptf1a (Figure 2T-U). Taken together, these results indicate that Ptf1a expression is sufficient to prevent PDAC initiation.
Pancreatitis is insufficient to overcome Pfla-mediated tumor suppression
As noted above, deletion of Ptf1a enhances pancreatic inflammation (Krah et al., 2015), which could mediate its pro-tumorigenic consequences. To determine if inflammation could bypass tumor suppression by Ptf1a, we activated tetO-Ptf1a in KrasG12D mice that were also subjected to caerulein-induced pancreatitis, a model of increased PDAC risk (Fig 3A) (Guerra et al., 2007; Lowenfels et al., 1993). Pancreata of control mice were fully recovered at 3 weeks following induction of pancreatitis, while robust PanIN formation was seen in KrasG12D pancreata, as previously reported (Fig. 3B-C) (Guerra et al., 2007; Kopp et al., 2012). In contrast, KrasG12D + tetO-Ptf1a pancreata exhibited reduced PanIN formation and lacked the fibro-inflammatory stroma characteristic of caerulein-treated KrasG12D mice (Fig. 3D). CK19 and Alcian blue analysis confirmed the dramatic reduction in PanIN burden in KrasG12D + tetO-Ptf1a pancreata (Fig. 3E-J, Q). The absence of anti-GFP and anti-Ptf1a staining again indicated that residual PanINs in these mice arose from Ptf1a-negative escaper cells (Fig. 3K-P, R). Thus, Ptf1a is both necessary and sufficient to block PDAC initiation, even in the presence of tumor-promoting inflammation (Krah et al., 2015). These results additionally indicate that downregulation of Ptf1a is a necessary cell-autonomous event for PanIN initiation, as only cell incapable of sustaining Ptf1a expression (i.e. non-rtTA-expressing, GFP-negative cells) give rise to PanINs.
Re-expression of Ptf1a reverses pancreatic transformation in vivo
The therapeutic potential of these findings would be greatest if Ptf1a activation could redirect the cell fate of already-established PDAC precursors, preventing tumor progression and reverting transformed cells to a stable differentiated state. To address this, we allowed PanINs to form in DOX-untreated KrasG12D and KrasG12D + tetO-Ptf1a mice for 8W (DOX-off), followed by DOX-induction of Ptf1a for 3W (DOX-on). Whereas large PanINs were present in KrasG12D pancreata, lesions of KrasG12D+tetO-Ptf1a mice appeared smaller and contained misplaced Ptf1a+ cells (Supplementary Figure 3). To specifically follow the fate of Pfla-reactivating cells within established lesions, we analyzed GFP (rtTA) and βgal (tetO-Ptf1a) production at multiple timepoints before and after DOX (Figure 4). Most PanINs that formed at 8W DOX-off, or those analyzed after 24 hrs DOX-on after 8W DOX-off, contained abundant GFP+ cells (Figure 4A, J), i.e., they expressed rtTA. By contrast, 3W and 6W DOX-on treatment induced a progressive exclusion of GFP+ cells from PanINs (Figure 4B-C, J). At 3W DOX-on, we observed a striking formation of hybrid structures in which clustered GFP+ cells appeared tightly connected to CK19+ lesions (Figure 4B’, K). In contrast to KrasG12D pancreata where amylase+ cells were excluded from PanINs, the emerging GFP+ cells in KrasG12D+ tetO-Ptf1a pancreata were amylase+, suggesting re-differentiation into acini (Figure 4D-E, white arrows). These results were corroborated by the LacZ (tetO-Ptf1a) reporter: while DOX-off mice had no βgal+ pancreatic cells, 24 hr DOX-on induced robust βgal expression in a subset of PanINs, which were then absent from lesions after 3-6W DOX-on (Figure 4F-I, L). Together, these data suggest that Ptf1a re-expression reversed transformation of pre-cancerous cells in vivo.
DISCUSSION
Pancreatic cancer evolves through premalignant stages lasting a decade or more, providing a window for prevention and reversion of this deadly disease with the development of improved tools for detection and management (Hruban et al., 2000; Yachida et al., 2010). Several recent mouse modeling studies have shown the importance of acinar cell differentiation factors, including Ptf1a, in restraining PDAC initiation (Flandez et al., 2014; Krah et al., 2015; Roy et al., 2016; Shi et al., 2009; von Figura et al., 2014). Consistent with these findings, human genome-wide association studies have identified polymorphisms in NR5A2 and PDX1,transcriptional partners and regulators of Ptf1a, which increase the risk of PDAC (Petersen et al., 2010; Wolpin et al., 2014). Ptf1a itself is downregulated in human PanINs, and required in mice to maintain homeostasis, restrain pancreatic inflammation, and suppress RAS-related gene signatures (Hoang et al., 2016; Krah et al., 2015). Together, these studies suggest that maintaining acinar cell differentiation could be an attractive therapeutic approach to limit pancreatic tumor initiation and progression (Murtaugh, 2014; Rooman and Real, 2012).
Such an approach is supported by the findings reported here, that sustained Ptf1a expression can prevent and reverse the early stages of pancreatic tumorigenesis. Our results indicate that downregulation of Ptf1a is essential for PanIN formation and maintenance, as only Ptf1a-negative “escaper” cells contribute to PanINs in KrasG12D+tetO-Ptf1a pancreata, in two different experimental schemas of PDAC initiation (Figures 2 and 3). Additionally, transgenic reintroduction of Ptf1a into established lesions reverts their phenotype to amylase-positive acinar cells, providing the first functional evidence that differentiation can override KrasG12D-mediated oncogenesis (Figure 4). These findings are consistent with a previous study, which suggested that PanIN cells could revert to normal acini following the silencing of KrasGl21D (Collins et al., 2012). As previous work from our labs suggests that Ptf1a suppresses RAS-dependency genes (Krah et al., 2015), we speculate that activation of Ptf1a indirectly inhibits transformation by inhibiting expression of genes required for oncogenic Kras function, while also promoting tissue reorganization, maintenance, and homeostasis (Hoang et al., 2016). Importantly, while Kras activation is an effectively irreversible genetic alteration in PDAC, downregulation of Ptf1a occurs at the transcriptional level, and is in principle reversible. Strategies aimed at restoring Ptf1a expression should complement efforts to develop inhibitors of KRAS itself and its downstream effectors (Collins and Pasca di Magliano, 2013; McCormick, 2015).
Chemoprevention strategies are under active investigation in PDAC (Miller et al., 2016), and our findings suggest that maintenance of Ptf1a and other acinar cell differentiation factors will be pivotal to their success, and could enable PDAC prevention even in high-risk individuals with chronic inflammation. Our results highlight the capacity of an epigenetic differentiation program to overcome the genetic alterations driving tumorigenesis, which may apply beyond PDAC to other solid tumors in which initiation involves alterations of differentiation state (Krah and Murtaugh, 2016; Roy and Hebrok, 2015).
AUTHOR CONTRIBUTIONS
N.M.K., R.M.J. and L.C.M. designed the research study. N.M.K, D.Y., J.S, and L.C.M acquired and analyzed the data. C.V.E.W. provided pivotal reagents. The manuscript was written by N.M.K and L.C.M. with input from R.M.J. and C.V.E.W.
ACKNOWLEDGMENTS
We are grateful to members of our laboratories as well as Howard Crawford, Gabrielle Kardon and Matthew Firpo for helpful input. This work was supported by the National Institutes of Health, through the following grants: F30-CA192819 (N.M.K.), R01-DK061220 and R01-CA194941 (L.C.M. and R.J.M.); U01-DK089570 (C.V.E.W.). We declare no conflicts of interest.
METHODS
Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author, L. Charles Murtaugh (murtaugh{at}genetics.utah.edu).
EXPERIMENTAL MODEL AND SUBJECT DETAILS
Mice
Experimental mouse alleles have been utilized in previous publications within the pancreatic research community: Ptf1aCreERT (Ptf1atm2(CreER1)CVW) (Kopinke et al., 2012; Kopp et al., 2012; Krah et al., 2015), KrasLSL-GI2D (Krastm4ty]) (Hingorani et al., 2003), Rosa26rtTA-IRES-EGFP (Belteki et al., 2005; Collins et al., 2012), and tetO-Ptf1a (Willet et al., 2014). To induce Cre-mediated recombination, tamoxifen (Cayman Chemical, Ann Arbor, MI) in corn oil was administered via oral gavage at 0.25 mg/g body weight on three consecutive days. To induce tetO-Ptf1a expression, all experimental mice were administered 1 mg/mL of Doxycycline (Research Products International, Mt. Prospect, IL) with 1% D-sucrose (Fisher Scientific) in their drinking water. Water bottles containing Doxycycline were protected from light and refreshed two times per week. All mice utilized in this study were between 6 and 12 weeks of age at the beginning of experiments. No mice with health status reports, as determined by an in-house veterinarian, were utilized in any analysis. For each experiment and experimental group, we utilized approximately identical number of male and female mice. Previous studies have not identified a gender-dependent PanIN burden in any mouse model of PDAC initiation or progression. Furthermore, human PDAC affects men and women with similar incidence, leading to ~7% of cancer related deaths in both genders (Siegel et al., 2016).
All experiments involving mice were performed according to institutional IACUC and NIH guidelines.
Tissue processing and histology
Pancreata were dissected into ice cold PBS, separated into multiple parts and processed for frozen and paraffin sections, as previously described (De La O et al., 2008; Krah et al., 2015). For paraffin sectioning, tissues were fixed in zinc-buffered formalin (Z-fix; Anatech, Battle Creek, MI) at room temperature overnight, followed by processing (dehydration in ethanol washes) into Paraplast-Plus (McCormick Scientific). Frozen specimens were fixed for 1-2 hr in 4% paraformaldehyde in 1x PBS on ice, followed by processing into Tissue-Tek O.C.T. compound (Fisher Healthcare). Paraffin and frozen sections were 6-8-microns with ≥100 μm spacing between individual sections, all placed on a single slide.
IHC and immunofluorescence followed established protocols (De La O et al., 2008; Krah et al., 2015) and included high temperature antigen retrieval (Vector Unmasking Solution; Vector Laboratories, Burlingame, CA), prior to staining all paraffin sections. Primary antibodies are listed in the Key Resource Table. Secondary antibodies, raised in donkey (Jackson Immunoresearch, West Grove, PA) were diluted 1:250 in blocking solution. Vectastain reagents and diaminobenzidine (DAB) substrate (Vector Laboratories) were used for all IHC experiments (See Key Resource Table). Immunofluorescence sections were counterstained with DAPI and mounted in Fluoromount-G (Southern Biotech), and photographed on an Olympus IX71 microscope, using MicroSuite software (Olympus America, Waltham, MA). Images were processed in Adobe Photoshop, with exposure times and adjustments identical between genotypes and treatment groups.
For Alcian blue staining, paraffin sections were pre-incubated 15 min in 3% acetic acid, stained 10 min in 1% Alcian blue in 3% acetic acid), and washed extensively in 3% acetic acid and dH2O. Following staining, all slides were washed in 0.5% acetic acid, dehydrated and equilibrated into xylene, and mounted with Permount.
METHOD DETAILS
Quantifications
PanIN scoring
To measure the number of PanINs per pancreas, the entire surface area of each Alcian blue/eosin-stained section was photographed at 4x original magnification, followed by photo-merging in Adobe Photoshop. The surface area was measured using ImageJ software (NIH). Alcian blue+ PanINs were counted manually under the microscope and marked on composite images in Adobe Photoshop. PanIN burden was calculated as total number of Alcian blue+ lesions per cm2. As previously described, metaplastic lesions that did not stain with Alcian blue were not counted (Krah et al., 2015). To avoid double-counting tortuous lesions that could occupy multiple regions in 3-D space, no more than one lesion was scored within an anatomically distinct pancreatic lobule (De La O et al., 2008; Krah et al., 2015).
Quantification of immunofluorescence images
To quantify the R26rtTA recombination frequency, we imaged 10-12 randomly selected 20x fields per specimen (across multiple sections). Using ImageJ (NIH), cell co-expressing GFP with the acinar differentiation marker, Amylase, were detected by additive image overlay with DAPI and anti-GFP, and counted using the Analyze Particles function as described previously (Keefe et al., 2012; Krah et al., 2015). To ensure counting accuracy, random images were manually spot-checked, using Adobe Photoshop. All calculations were performed in Microsoft Excel and results graphed as individuals with error bars representing the standard deviation. The p-values were determined by two-tailed, unpaired t-test in Graphpad Prism 7.
To quantify the number of GFP-positive PanINs, 10 randomly selected fields were imaged per mouse. Each PanIN was manually scored according to the number of GFP+/CK19+ cells present. If more than two cells (or regions) co-expressed CK19 and GFP, the PanIN was considered GFP-positive.
Quantification of histological images
To quantify cell proliferation in PanINs, each lesion was scored according to its number of Ki67+ nuclei. Each lesion was categorized as low-(0-1 Ki67+ nuclei), mid-(<50% Ki67+ nuclei), or high-grade (>50% Ki67+ nuclei). At least 10 fields with PanINs were counted per animal (n=3 mice per genotype). To quantify the number of Ptf1a-negative vs. positive PanINs, as many lesions as possible were imaged from Ptf1a-stained KrasG12D and KrasG12D + tetO-Ptf1a pancreata (n=5 mice per genotype). If two or more cells contained Ptf1a+ nuclei, the PanIN was considered Ptf1a-positive. The number of βgal+ PanINs was determined by counting βgal+ cells per lesion; if the lesion contained 2 or more βgal (tetO-Ptf1a)-positive cells, the lesion was considered positive.
Caerulein treatment
Acute pancreatitis was induced by i.p. injection of caerulein (Bachem, Torrance, CA), 0.1 μg/g in filter sterilized saline, six times (every 60 minutes) daily over two consecutive days, as previously described (Keefe et al., 2012; Kopp et al., 2012; Krah et al., 2015). Controls were injected with an equal volume of sterile saline. Pancreata from all caerulein-treated mice were harvested three weeks following final injection and processed as described above.
QUANTIFICATION AND STATISTICAL ANALYSIS
Statistics
All statistics were performed in Graphpad Prism 7. A two-tailed t-test was used to calculate p-values for PanIN burden and percentage of Ptf1a-negative PanINs (Figure 2). A Fisher exact test was used to calculate p-values for nominal data, such as relative frequencies of GFP/CK19 PanINs (Figure 2 and Figure 3). Where multiple groups were compared against one another (Figure 4), p-values were determined by ANOVA. All graphs, regardless of the statistical test used, show the mean ± the standard deviation. Statistical significance was defined as a P-value of <0.05 for the indicated analysis, as determined by Graphpad Prism 7 software.