Inflammation-Induced NFATc1–STAT3 Transcription Complex Promotes Pancreatic Cancer Initiation by Kras^G12D

Commonly diagnosed at advanced and incurable stages, pancreatic ductal adenocarcinoma (PDA) represents the fourth leading cause of cancer-related death in Western countries, rendering it one of the most lethal human cancers (1, 2). PDA evolves through a series of histopathologic changes referred to as acinar-to-ductal metaplasia and progressive pancreatic intraepithelial neoplasia (PanIN), which are accompanied by a recurrent pattern of genetic alterations; the earliest and most prevalent of which is oncogenic activation of KRAS (3). The relevance of the Kras^G12D mutation for pancreatic carcinogenesis has been elegantly demonstrated in genetically engineered mouse models (GEM) with conditional activation of this oncogene in the embryonic pancreas. Of note, as originally described by Hingorani and colleagues (4), Kras^G12D activation in pancreatic epithelial cells induces the development of PanIN precursor lesions, which eventually progress to invasive PDA after a long latency. Collectively, these studies in mice and humans suggest that PDA originates from Kras^G12D-initiated cells, which need long-time exposure to either cell-autonomous or environmental clues that act as tumor promoters. Importantly, pancreatic cancer cells are surrounded by a pronounced proinflammatory microenvironment that is driven by the secretion of tumor-derived proinflammatory cytokines (5, 6). Furthermore, recent findings unraveled that inflammatory cytokines, such as tumor-derived granulocyte macrophage colony-stimulating factor (GM-CSF), can exert cancer-promoting effects in vivo by directly modifying gene expression networks in pancreatic epithelial cells, rather than exclusively turning on and off these pathways in inflammatory cell populations from the tumor microenvironment (5–7).

Moreover, chronic pancreatitis is regarded as a major risk factor for the development of pancreatic cancer, further highlighting the key role of inflammation in the pathophysiology of pancreatic cancer development (8, 9). To this end, Guerra and colleagues (10–13) recently established a new experimental GEM, whereby induction of a mild form of pancreas inflammation synergizes with Kras^G12D to initiate early PanIN lesions and promote their rapid progression toward invasive PDA. This model highlighted the crucial role of inflammation in the process of malignant transformation in the pancreas.

However, the mechanisms linking inflammation and malignant transformation and progression in pancreatic epithelial cells are still poorly understood. As oncogenic activation of the Kras^G12D signaling pathways is still deemed undruggable, interaction partners that promote and cooperate with Kras^G12D-driven carcinogenesis may open new avenues for novel drugs in prevention and therapy (4, 14, 15). Here, we demonstrate that NFATc1, a transcription factor originally discovered in T lymphocytes (16), is strongly induced upon inflammatory stimuli and dramatically accelerates malignant transformation in the pancreas when concomitant Kras^G12D mutation is present. We also find that NFATc1 forms chromatin-bound complexes with STAT3 in epithelial cells, another well-characterized and inflammation-induced transcription factor. The generation of genome-wide chromatin immunoprecipitation sequencing (ChIP-seq) and expression profiling datasets reveal that the NFATc1–STAT3 cooperativity regulates genome areas involved in the transcriptional activation of cancer-associated gene networks. Combined, these data provide robust evidence for the existence of a novel interaction between two important transcription factors (the NFATc1–STAT3 complex) in pancreatic epithelial cells. More importantly, these transcriptional pathways, which exert distinct functions in inflammatory cells, act in concert in pancreatic epithelial cells to mediate growth-promoting effects upon inflammation in the setting of Kras mutations. The relevance of these findings is underscored by the fact that small molecules that target these pathways are being tested in early clinical trials. Consequently, our findings not only advance our understanding of how inflammation drives the progression of pancreatic cancer but may also open new avenues for the rational design of future combinatorial therapies for patients with chronic inflammatory conditions that are at risk to develop malignancies.

This work was prompted by recent observations suggesting that activation of transcription factor pathways in pancreatic epithelial cells through environmental inflammatory conditions can promote carcinogenesis, specifically in the presence of oncogenic Kras mutations (17). We initially focused our attention on the transcription factor NFATc1, which, though absent in healthy human and murine pancreas, becomes highly induced in Kras^G12D-expressing neoplastic pancreatic cells when these mice were treated with daily doses of cerulein to induce inflammation (Supplementary Fig. S1A–S1C), strengthening our own previous observations of nuclear NFATc1 activation in pancreatitis-associated human PDA (18). Thus, we tested whether recapitulating the induction of NFATc1 in pancreatic epithelial cells exerts oncogenic functions in cooperation with Kras^G12D. For this purpose, we first generated an inducible transgenic Nfatc1 mouse model (Fig. 1A) by introducing a loxP–STOP–loxP hemagglutinin (HA)-tagged c.n.NFATc1 cDNA into the mouse ROSA26 locus by homologous recombination and crossed it with p48-Cre and Pdx1-Cre mice to obtain animals that express a form of nuclear-localized NFATc1 that is transcriptionally active (Supplementary Fig. S2A and S2B). Nfatc1 mice were born at the expected Mendelian ratio and did not display gross abnormalities in the pancreas. Increased cell proliferation was observed in pancreata of young mice, but despite the early proliferative effect on pancreatic cells, Nfatc1 mice failed to develop advanced PanIN lesions within a 1-year observation span (data not shown). Therefore, though NFATc1 activation promotes cell growth, it does not cause cancer by itself, and instead may synergize with oncogenic Kras^G12D to promote neoplastic cell growth in response to inflammation. This intriguing hypothesis is particularly attractive as the majority of human PDAs are characterized by the combined expression of both proteins (18). Consequently, we mimicked this situation by generating Kras^G12D;Nfatc1 mice carrying transgenic expression of both proteins (Fig. 1A and Supplementary Fig. S2A and S2B), a genetic manipulation that dramatically shortened animal survival (140 of 161 days; P < 0.0001) when compared with littermates expressing Kras^G12D or Nfatc1 alone (Fig. 1B). Kras^G12D;Nfatc1 mice developed severe cachexia and abdominal distension caused by the accumulation of sanguineous ascites and bile duct obstruction highly resembling clinical features of human PDA (Fig. 1C). At necropsy, the pancreata from Kras^G12D;Nfatc1 mice were enlarged by tumor masses, which invariably contained both solid and cystic regions (Fig. 1C). At the histologic level, the pancreas of Kras^G12D;Nfatc1 animals at 4 weeks of age displayed substantial replacement of acinar cell areas by numerous acinar-to-ductular metaplasia (ADM; Fig. 1D, 1I), PanIN precursors (Fig. 1D, 1II–III), and atypical flat lesions (AFL; Fig. 1D, 2I). At 8 weeks, the full spectrum of preinvasive lesions ranging from ADM to early- and late-stage PanIN1-3 lesions (Fig. 1D, 2II–III and Supplementary Fig. S2C) was observed, and by 36 weeks of age, all animals showed invasive and metastatic cancers (Figs. 1D, 3I–III and 2A) ranging from well-differentiated (G1) and moderately differentiated (G2) PDA to poorly differentiated G3 tumors with anaplastic and/or adenosquamous components and low levels of cytokeratin-19 expression (Fig. 2B and Supplementary Fig. S2D). Notably, equivalent to human PDA, pancreata from Kras^G12D;Nfatc1 mice showed robust nuclear NFATc1 expression throughout carcinogenesis (Fig. 2C). Tumor progression in Kras^G12D;Nfatc1 mice was further characterized by an increased proliferative index in epithelial cells as assessed by Ki67 quantification (5% vs. 17%; P < 0.05; Fig. 2D and Supplementary Fig. S2E), which positively correlated with the upregulation of cell cycle–promoting genes (e.g., Cdk4), and silencing of the p16^INK4a tumor suppressor (Fig. 2E and Supplementary Fig. S2F). Together, the Kras^G12D;Nfatc1 model not only recapitulates key features of human PDA but also demonstrates profound pro-oncogenic properties of NFATc1. Moreover, this observation suggests that stimuli mediating the transition of pancreatitis to pancreatic cancer in the background of Kras^G12D may proceed, at least in part, via this transcription factor pathway.

Here, we sought to elucidate the mechanism of NFATc1 to accelerate pancreatic carcinogenesis. First, we generated primary cell lines derived from Kras^G12D;Nfatc1 tumors (hereafter referred to as KNC 1–6 cell lines), and used microarray-based expression profiling analyses as genome-wide reporter assays for determining the effects of inactivating NFATc1 in KNC cells by RNAi. The data of these experiments were subjected to gene set enrichment analysis (GSEA) for the identification of NFATc1-dependent gene signatures (19). GSEA pathway analysis revealed enrichment of NFATc1 signatures including target genes implicated in transformation, growth, and inflammation, such as Cyclin D1 and D3 and CDK1 and CDK4 (Fig. 3A and B). Most notably, we identified an enrichment of STAT3 and related inflammatory pathways in NFATc1-expressing cells, and, consequently, NFATc1 depletion was accompanied by a massive loss of STAT3 expression (Fig. 3A–C and Supplementary Fig. S3A–S3C). This observation is important, as earlier studies had suggested that high STAT3 expression and activity levels associate with the development of PDA in an inflammatory setting in both humans and Kras^G12D mice (17, 20). Therefore, we hypothesized the existence of a pro-oncogenic NFATc1 and STAT3 cooperativity during PDA development. In line with this, we found high levels of STAT3 expression and activation (indicated by Y705 phosphorylation) in Kras^G12D;Nfatc1 tumors compared with Kras^G12D littermates, and in human and murine PDA cells with concomitant high NFATc1 expression levels and Kras^G12D mutation (Fig. 3D–F and Supplementary Fig. S3D). Furthermore, immunohistochemistry staining of PDA samples from 217 patients identified nuclear NFATc1 expression in 70.2% (151 of 217) and, most importantly, the vast majority (86.7%) of NFATc1-positive PDA showed coexpression of nuclear phosphorylated (p) STAT3 (Y705; Fig. 3G and H; further details are provided in Supplementary Data). Consistent with the observed positive correlation of nuclear NFATc1 and STAT3 activation levels, immunofluorescence microscopy in PDA cells revealed accumulation of NFATc1 and pSTAT3 (Y705) in euchromatic regions of tumor cell nuclei, suggestive of a functional cooperation of both transcription factors in gene activation (Fig. 3I). Correspondingly, coimmunoprecipitation identified endogenous NFATc1–STAT3 complexes in human and murine PDA cells (Fig. 3J) and demonstrated that successful complex formation requires STAT3 activation at Y705 [pSTAT3 (Y705); Fig. 3K and L]. In fact, mutational disruption of the Y705 activation site or treatment with the STAT3 inhibitor WP1066 disrupted complex formation with NFATc1 in cancer cells (Fig. 3K and L). Thus, our combined cell biologic, biochemical, and molecular datasets derived from studying both mice and humans support the notion of an NFATc1–STAT3 interplay in the nucleus of pancreatic cancer cells that functionally promotes Kras^G12D-induced carcinogenesis.

To investigate the function of the NFATc1–STAT3 interaction in gene regulation and carcinogenesis, we generated KNC cells with stable STAT3 knockdown (KNC–shSTAT3; Supplementary Fig. S4A) and performed ChIP to enrich DNA fragments bound by NFATc1, followed by direct high-throughput sequencing (ChIP-seq). Data analysis using the PeakRanger algorithm [with negative binomial P < 10⁻⁴ at a false discovery rate (FDR) < 5 × 10⁻²] identified 1,798 NFATc1-binding genomic regions. Multiple EM for Motif Elicitation (MEME)-ChIP de novo identification of motifs overrepresented within peak regions (21) disclosed highest enrichment of the previously established NFAT consensus motif GGAAA (Fig. 4A). Furthermore, the MEME algorithm identified a significant accumulation of the restricted STAT3 consensus site (GGAA for monomeric STAT3 vs. TTCN₃GAA for all STAT dimers), which was centered in the NFATc1 peak summit (P = 2 × 10⁻¹²; Supplementary Fig. S4B), suggesting a heterodimeric binding of both factors on sites of NFATc1 enrichment. Interestingly, we found a striking accumulation of NFATc1 binding sites distant to transcription start sites (TSS; within 50–500 kb upstream and downstream; Fig. 4A; Supplementary Fig. S4C), arguing that NFATc1 may preferentially operate through long-range chromatin interactions to control target gene expression in pancreatic cancer cells. In fact, specific histone modifications, for example, H3K4me1 and H3K27ac (22), have been shown to mark features of active enhancer regions. To determine whether predicted NFATc1 enrichment sites overlap with active enhancer regulatory regions, we used ENCODE consortium datasets (23). Congruently, we found that 1,155 of 1,789 NFATc1 binding peaks fall within genomic regions typically marked by H3K4me1 (5.18-fold enriched over control regions, empirical P after 1,000 simulations < 5 × 10⁻³²⁴) and H3K27ac (Fig. 4B). In contrast, H3K4me3, a landmark for TSS binding, was not enriched in NFATc1 binding sites (76 of 1,789; Fig. 4B). In line with these findings, ChIP experiments at randomly selected regions indeed confirmed highly enriched NFATc1 binding at sites of high H3K4me1 and H3K27ac levels, indicative of active enhancers (Supplementary Fig. S4E and S4F). Importantly, more than two thirds of the identified NFATc1 peaks were explicitly regulated by STAT3, as inferred from a pairwise comparison between average binding levels across all 1,798 peaks in KNC–shControl versus KNC–shSTAT3 cells (Fig. 4C–E). Single ChIP experiments confirmed that STAT3 plays a critical role in NFATc1 recruitment and, therefore, RNAi-mediated depletion of this transcription factor diminished NFATc1 enrichment at active enhancer regions (Fig. 4F). Taken together, these data provide evidence for the existence of nuclear NFATc1–STAT3 transcription complexes that exert a mutually dependent binding to regulatory regions within the genome in pancreatic cells. To better understand the potential impact of this new transcriptional complex on PDA progression, these findings led us to subsequently define the gene networks that are regulated by these factors using genome-wide expression profiles.

To gain insight into the gene-regulatory functions of the NFATc1–STAT3 interplay, we first used the Genomic Regions Enrichment of Annotations Tool (GREAT; ref. 24) and analyzed the genome-wide ChIP-seq data as these relate to pathway affiliations and disease relevance. Numerous NFATc1 peaks coincided with gene signatures with functional implications for cell motility, cell migration, and extracellular matrix regulation (Supplementary Fig. S5A). We then matched ChIP-seq data with results from expression profiling to identify direct target genes of NFATc1–STAT3 complexes. Comparison of datasets identified distinct target gene subsets, of which selected candidates were chosen for further validation. Among these targets was EGFR, a receptor tyrosine kinase that has been reported to be a key player in inflammation-associated carcinogenesis with implications for transformation, tumor cell growth, and metastasis (25, 26). Furthermore, our target gene analyses revealed a prominent member of the cyclin protein family (Cyclin D3) that exerts essential growth-stimulating functions in pancreatic cancer (27, 28) as a target. Of note, gene signatures that associate to activate EGFR signaling as well as Cyclin D3 were also among the most significantly regulated NFATc1 downstream targets in KNC tumor cells identified by GSEA (Fig. 3A). Other subsets of direct target genes encompass the NFAT pathway regulator Rcan1, matrix metalloproteinase 13 (Mmp13), and the Wnt family members Wnt1 and Wnt10a (Fig. 5A). Consistent with ChIP-seq data, NFATc1 specifically regulates these direct target genes through interaction with nearby enhancer regions, as indicated by NFATc1 binding to sites of enriched H3K4me1 and H3K27ac modifications (Fig. 5A and B), low promoter-occupancy levels, increased DNAse hypersensitivity, and site-specific recruitment of histone acetyltransferase p300 (Supplementary Fig. S5B–S5D). In addition, we confirmed that NFATc1 recruitment to the enhancers of these genes required STAT3 interaction, as its genetic depletion or mutational inactivation [STAT3 (Y705)] diminished NFATc1 recruitment and target gene transcription (Fig. 5C and D and Supplementary Fig. S5E and S5F). Conversely, increased NFATc1 recruitment at specific enhancer sites was observed when STAT3 was activated upon IL6 treatment of pancreatic cancer cells, and this was accompanied by recruitment of RNA polymerase II at the corresponding promoter (Supplementary Fig. S5G and S5H). Interestingly, STAT3 is preferentially recruited to corresponding promoter sites rather than to enhancers upon activation (Supplementary Fig. S5I), in a manner that requires the presence of NFATc1 binding to nearby enhancers. Hence, increasing target promoter transactivation upon STAT3 titration occurred only in the presence of an intact NFAT consensus site within the corresponding enhancer (Fig. 5E). Thus, these data suggest that both proteins support the type of chromatin looping that is characteristic of enhancer–promoter communications at regulatory regions that mediate robust gene activation (29, 30).

Finally, the relevance of this novel pathway was confirmed by expression studies, showing strong induction of EGFR during pancreatic carcinogenesis in both human and mouse PDA cells, in which treatment with NFAT inhibitors (cyclosporin A) or genetic Nfatc1 suppression diminished expression of the identified target genes (Fig. 5F–J). Together, these data demonstrate for the first time that nuclear interactions between the transcription factors NFATc1 and STAT3 regulate genes known to promote cancer initiation and progression in vivo and identify enhancer-to-promoter communication as one of the putative mechanisms by which these proteins achieve their functions.

The studies described above have thus far revealed that NFATc1, after being induced by inflammatory conditions, accelerates Kras^G12D-mediated initiation of pancreatic carcinogenesis by forming a complex with STAT3, and identified their direct target genes using two complementary genome-wide methods (ChIP-seq and expression profiling). Thus, to determine the significance of this novel pathway in the original context of inflammation-induced cancer promotion as well as to evaluate the chemopreventive potential of its targeting, we treated Kras^G12D mice with cerulein daily for 4 weeks to induce a mild and persistent inflammation, as described recently (10). Consistent with previous reports, inflammation accelerated Kras^G12D-driven carcinogenesis and caused rapid formation of ADM with subsequent progression to high-grade PanIN2 and PanIN3 lesions (Fig. 6A). The promotion of these Kras^G12D-initiated neoplastic lesions was characterized by a strong induction of NFATc1 and pSTAT3 (Y705) in neoplastic epithelial cells and subsequent induction of oncogenes that function as a downstream target of this cooperativity, as demonstrated for EGFR tyrosine kinase and Wnt10a (Fig. 6A and B). The results are congruent with our recently proposed model for the role of self-reinforcing loops in the pathobiology of pancreatic cancer that occurs in the presence of persistent inflammation (31), as they indicate that the NFATc1–STAT3 complex transduces signals activated by cerulein from the cell membrane to the nucleus to turn on genes encoding proteins that reinforce cell growth stimulation.

Finally, to determine the extent to which the induction of the NFATc1–STAT3 cooperativity contributes to inflammatory-driven cancer promotion by KRAS^G12D, we used both pharmacologic and genetic strategies to inactivate this complex. Consequently, we specifically inactivated NFATc1 in pancreatic epithelial cells by interbreeding Kras^G12D mice with NFATc1^fl/fl;Pdx1-Cre animals (Kras^G12D;Nfatc1^Δ/Δ mice; Supplementary Fig. S6A–S6C). Here, cerulein treatment failed to induce STAT3 activation and subsequent target gene expression (Fig. 6A and B). Although genetic depletion of Nfatc1 did not affect ADM and PanIN formation in untreated 3-month-old Kras^G12D mice (data not shown), it significantly antagonized the cerulein-induced proliferation rates in Kras^G12D epithelial cells (Fig. 6C and Supplementary Fig. S6D) and significantly blocked ADM, as evidenced by restored normal duct levels in Nfatc1-null tissues following cerulein challenge (Fig. 6D). This supports the hypothesis that disruption of NFATc1–STAT3 cooperativity rather than Nfatc1 ablation in progenitor cells itself protects from pancreatic cancer initiation and progression.

On the basis of these results, we hypothesized that drugs that are currently used in the clinical setting to inhibit NFAT might have a similar effect on inflammation-associated PanIN formation. To test this hypothesis, we suppressed NFATc1 activity in Kras^G12D mice in vivo by daily treatment with cyclosporin A along with cerulein for 3 months and examined the biochemical and pathobiologic effect of this intervention on preneoplastic epithelial cells. Congruent with the genetic data described above, cyclosporin A treatment blocked STAT3 activation, significantly reduced cell proliferation, and prevented ADM in 12-week-old mice (Fig. 6A–D). Noteworthy, the disruption of the NFATc1–STAT3 interaction (by either genetic or pharmacologic approaches) was paralleled by the lack of EGFR induction in both genetically modified mice (Fig. 6A) and acinar cell explants (Fig. 6E). Similar findings were obtained for other cancer-related NFATc1–STAT3-regulated oncogenic targets identified by our genome-wide approaches (Fig. 6A and B and Supplementary Fig. S6E).

Persistent inflammation is a hallmark feature of PDA that promotes the transition of this disease from its preneoplastic state to frank PDA in the context of KRAS mutations. The goal of the current study has been to provide insight into mechanisms of cancer progression driven by inflammation. Our guiding hypothesis has been that transcription factors, which were originally discovered in inflammatory cells and thought to act in the tumor microenvironment, are activated in preneoplastic and neoplastic cells through inflammatory stimuli, thus executing their function within the epithelial compartment to promote inflammation-associated carcinogenesis in the presence of KRAS mutations. Our study led to the following findings: (i) the pro-oncogenic transcription factors NFATc1 and STAT3 are activated by inflammatory stimuli and subsequently cooperate to govern a defined oncogene expression profile in neoplastic epithelial cells; (ii) activation of the NFATc1–STAT3 cooperativity in GEMs promotes Kras^G12D-driven carcinogenesis, whereas their inactivation has the opposite effect; (iii) mechanistically, NFATc1–STAT3 complexes control gene expression through enhancer-to-promoter communication, a powerful epigenetic regulatory mechanism in the field of gene expression (29, 30); (iv) identified NFATc1–STAT3-regulated genes, for example, those encoding EGFR and Wnt family members, which are targets of novel drugs being tested in the setting of experimental therapeutics; (v) pharmacologic disruption of the NFATc1–STAT3 complex hampers its tumor-promoting effects; and (vi) ectopic coexpression of NFATc1 and STAT3 is observed in human pancreatic cancer tissues, suggesting a possibility of immediate translation of our findings.

Our findings elucidate fundamental biochemical properties displayed by transcription factors under inflammatory conditions to achieve their tumor-promoting effect. For instance, we find that at the transcriptional level, NFATc1 binds to GGAAA consensus sequences on DNA, albeit with weak affinity (32). Efficient NFATc1 DNA binding, however, can be mediated and maintained through interactions with other partner proteins (32–34). Therefore, we identified that other inflammatory transcription factors, such as STAT3, partner with NFATc1 in the nucleus of pancreatic epithelial cells. Like NFATc1, STAT3 activation translates inflammatory signals from the tumor microenvironment into the expression of specific gene networks involved in carcinogenesis (35). Aberrant expression and activation of STAT3 is frequently observed in human pancreatic carcinoma and can favor the progression of PanIN lesions in the transgenic Kras^G12D model (36). Activating STAT3 mutations are not observed in PDA (17, 20, 36, 37); however, the mechanisms of enhanced STAT3 expression are poorly understood. One proposed mechanism is a feed-forward loop that maintains STAT3 expression through (IL6-mediated) elevated activation levels of pSTAT3 (Y705; 37). It is worth underscoring that our studies demonstrate that NFATc1 stimulates STAT3 expression in primary tumor cells, whereas genetic loss or pharmacologic inhibition of NFATc1 by cyclosporin A diminishes expression of STAT3. Likewise, extensive analyses of human PDA samples found a significant positive correlation between nuclear STAT3 and NFATc1 expression levels, although we cannot fully exclude correlations between negative staining for NFATc1 and STAT3 that can be driven artificially in a few samples in which inadequate antigen retrieval or autolysis has occurred. However, it remains to be elucidated whether NFATc1 directly influences STAT3 expression or whether it stimulates STAT3 expression by maintaining STAT3 activation, as we and others have observed a regulatory impact on IL6 expression and STAT3 pathway activation by NFATc1 (38–40; data not shown). Even more striking, we find that this new NFATc1 pathway is not limited to mediating regulation of STAT3 expression, but also leads to the formation of NFATc1–STAT3 nucleoprotein complexes, which are essential for the transcription of gene networks that account, at least in part, for tumor progression in PDA. Interestingly, ChIP-seq analyses for genome-wide identification of genes regulated by the NFATc1–STAT3 complex revealed more than 1,100 putative NFATc1 target sites, whose binding intensity was significantly regulated by nuclear STAT3. Although NFATc1–STAT3 interactions on target gene promoters that influence cell migration and proliferation have been identified (40), our further analyses, in contrast, demonstrated that our identified binding sites are mostly located at putative enhancer regions located upstream, downstream, within, or even several thousand bases away from their corresponding target genes (41). These enhancer regions have been described as specialized areas in the nucleus where protein–DNA complexes are responsive to signal-regulated transcription factors and translate environmental stimuli into the regulation of gene expression networks, thus constituting at least one type of regulatory modules within genomes that support environmental–gene interactions. Enhancer activation defines time point, duration, and intensity of gene expression via complex mechanisms of chromatin regulation (42). Enhancers may spread stimulating signals as a result of acetylation and rearrangement processes of nucleosomes along the chromatin or induce the transcription of downstream genes through loop formation and promoter communications (41, 43). Our data support a model in which NFATc1–STAT3 complexes regulate target gene transcription through highly specific enhancer–promoter interactions, presumably via formations of chromatin loops. Our comparative analyses of genome-wide ChIP-seq studies and expression analysis confirmed this model and, moreover, provided evidence for the existence of distinct gene expression networks that are regulated via NFATc1–STAT3 complexes. We find that most of the NFATc1–STAT3 targets identified hereby exert functions in oncogenic processes such as cell-cycle propagation, migration, and invasion, as well as remodeling of the extracellular matrix of the pancreas. Some relevant examples include the EGFR, an oncogenic tyrosine kinase with critical implications for pancreatitis-promoted pancreatic carcinogenesis in mice and humans, as well as proproliferative cyclin D3 and MMP-13 (44), a central component of the MMP activation cascade and mediator of tumor cell invasion. The identification of Wnt1 and Wnt10a, ligands of the classical Wnt-β–catenin pathway and important regulators of growth, stemness, and differentiation, extends the scope of these investigations and suggests important cross-talk interactions between the NFATc1–STAT3 network and the Wnt pathway in pancreatic cancer progression. Interestingly, these data are in agreement with the recently proposed model whereby pancreatic cancer proceeds by the establishment of positive feedback loops that are self-reinforcing (31).

In conclusion, our results support the notion that transcription factors, previously known to regulate the function of immune cells, are activated by inflammatory stimuli and operate as nucleoprotein complexes within epithelial cells to promote Kras-driven carcinogenesis. Moreover, we unravel detailed mechanisms as to how these novel transcriptional complexes form and execute genome-wide instruction by binding and activating cancer-associated gene expression networks. These investigations led to the discovery of several proteins (e.g., EGFR) that, together with their regulators, such as Wnt family members or STAT3, are currently being evaluated as drug targets in early clinical trials. For instance, the oral small-molecule Wnt signaling inhibitor LGK974 is currently being tested in a phase I, open-label, dose-escalation study in several solid malignancies, including pancreatic cancer (NCT01351103). Furthermore, several STAT3 inhibitors (OPB-31121 and OPB-51602) are clinically tested in solid malignancies (NCT00955812, NCT01423903, and NCT01867073). Thus, the new knowledge provided by the current study helps to build the rationale for the future design of combinatorial therapies that should be more efficient for controlling inflammation-associated cancer progression. Finally, the finding that modulation of transcriptional networks that work via epigenetic mechanisms (e.g., NFATc1–STAT3 enhancer–promoter communications) can modulate oncogenic KRAS function in inflammation-promoted carcinogenesis expands our mechanistic understanding of this disease beyond the genetic-centric model that dominated the last two decades of research in this field. In light of the failure to translate these findings into successful therapies, for example through gene therapy, our data also highlight the possibility that therapeutic strategies that target transcriptional and epigenetic changes may be more beneficial for the management of inflammatory-driven preneoplastic diseases, such as in patients with chronic pancreatitis who are known to have an increased risk of developing pancreatic cancer.

Panc-1 and PaTu8988t cells were obtained from the European Collection of Animal Cell Cultures (ECACC) and HP Elsaesser (Philipps University, Marburg, Germany); L3.6 cells were a gift from Dan Billadeau (Mayo Clinic, Rochester, MN). Murine TD-2 cells were described previously (18). Testing and authentication of human cell lines were not performed by the authors. Primary pancreatic cancer cell lines were derived from murine Kras^G12D;Trp53^−/− and p48-Cre;Kras^G12D;Nfatc1 pancreatic tumors. A detailed description can be found in the Supplementary Methods.

P48-Cre, Pdx1-Cre, and LSL-Kras^G12D mice have been described previously (45–47). Nfatc1^fl/fl mice were kindly provided by Laurie Glimcher (48). The c.n.Nfatc1 knockin strain (C57BL/6 background) was generated by cloning an N-terminal HA-tagged constitutively active version of NFATc1 containing serine to alanine substitutions in the conserved serine-rich domain and all three serine–proline repeats into the ROSA26 promoter locus (Artemis Pharmaceuticals). The strains were interbred to generate Pdx1/p48-Cre;c.n.Nfatc1, Pdx1/p48-Cre;Kras^G12D, Pdx1/p48-Cre;c.n.Nfatc1;Kras^G12D, Pdx1-Cre;Nfatc1^Δ/Δ, and Pdx1-Cre;Nfatc1^Δ/Δ;Kras^G12D cohorts. Mutant mouse strains were genotyped by PCR as previously described by the laboratories that generated them. The following primers were used to genotype Nfatc1: 5′-catgtctgggagatggaagc-3′. Chronic pancreatitis was induced by single daily intraperitoneal injections of cerulein (0.2 mg/kg body weight; Sigma-Aldrich) 3 days/week for a period of 4 weeks (10). Mice were sacrificed after 12 weeks of treatment. All procedures were conducted in accordance with the regulatory standards of and were approved by the Regierungspräsidium Gießen.

ChIP-seq analysis was done as previously described (49). ChIP DNA was end-repaired and A-tailed. Illumina adaptors were ligated to the ChIP DNA fragments. Fractions (175-bp to 225-bp size) were cut out from a gel, eluted by Qiagen gel extraction kit, and enriched by 20 cycles of PCR amplification. The library size was controlled with the Experion-system (Bio-Rad) and subsequently quantified by PicoGreen assay and subjected to Illumina GAIIx sequencing according to the manufacturer's instructions. Only high-quality reads passing the internal Illumina-Raw data-filter (PF-cluster) were considered.

All studies carried out on human specimens were approved by the Mayo Clinic Institutional Review Board. Ten adenocarcinoma tissue microarrays (TMA) containing samples from 217 patients were analyzed and stained for NFATc1 and pSTAT3 expression in the Pathology Research Core. More details about staining procedures and data analysis can be found in the Supplementary Methods.

Data are presented as averages ± standard deviations (SD) or standard errors (SE) as noted and were analyzed by built-in t test using Microsoft Excel. P < 0.05 was considered significant. Tumor incidences and survivals were calculated with GraphPad Prism4. For the overall survival analysis, Kaplan–Meier curves were analyzed by log-rank test. In all cases, we chose a group size that produced statistically unambiguous results.

No potential conflicts of interest were disclosed.

Conception and design: S. Baumgart, J.T. Siveke, M.E. Fernandez-Zapico, M. Eilers, T.M. Gress, R. Urrutia, V. Ellenrieder

Development of methodology: S. Baumgart, J.-S. Zhang, S.K. Singh, I. Esposito, G. Singh, V. Ellenrieder

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Baumgart, N.-M. Chen, J.T. Siveke, A. König, J.-S. Zhang, S.K. Singh, E. Wolf, I. Esposito, J. Reinecke, J. Nikorowitsch, M. Brunner, G. Singh, T. Smyrk

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Baumgart, N.-M. Chen, J.T. Siveke, A. König, S.K. Singh, E. Wolf, M. Bartkuhn, I. Esposito, E. Heßmann, J. Reinecke, J. Nikorowitsch, G. Singh, W.R. Bamlet, A. Neesse, R. Urrutia, V. Ellenrieder

Writing, review, and/or revision of the manuscript: S. Baumgart, N.-M. Chen, J.T. Siveke, E. Heßmann, M.E. Fernandez-Zapico, T. Smyrk, W.R. Bamlet, A. Neesse, T.M. Gress, D.D. Billadeau, D. Tuveson, R. Urrutia, V. Ellenrieder

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Baumgart, E. Heßmann, G. Singh, T.M. Gress, R. Urrutia

Study supervision: S. Baumgart, S.K. Singh, R. Urrutia, V. Ellenrieder

The authors thank Dr. Laurie Glimcher (Weill Cornell Medical College, Provost of Medical Affairs, Cornell University, Ithaca, NY) for NFATc1^Δ/Δ mice. The authors are grateful to Kristina Reutlinger, Bettina Geisel (Philipps University), and Susanne Haneder (Technische Universität, Munich, Germany) for technical support, and Dr. Lukas Rycak (GenXPro GmbH) for the statistical analyses.

This work was generously supported by the Deutsche Forschungsgemeinschaft (KFO210 and SFB-TR17, to V. Ellenrieder); the German Cancer Research Foundation (no. 109423 “Inflammation and Cancer” and AK “Mildred Scheel” Fellowship, to V. Ellenrieder; “Max Eder” Fellowship, to A. Neesse); the University Medical Centre Giessen and Marburg (to A. Neesse); Mayo Foundation for Medical Research, NIH grants DK52913 and P30DK084567 (to R. Urrutia); and NCI Pancreas SPORE Grant P50 CA102701 (to M.E. Fernandez-Zapico and D.D. Billadeau).