Epithelial memory of inflammation limits tissue damage while promoting pancreatic tumorigenesis

Abstract

Inflammation is a major risk factor for pancreatic ductal adenocarcinoma (PDAC). When occurring in the context of pancreatitis, mutations of KRAS accelerate tumor development. We discovered that long after its complete resolution, a transient inflammatory event primes pancreatic epithelial cells to subsequent transformation by oncogenic KRAS. Upon recovery from acute inflammation, epithelial cells of the pancreas display an enduring adaptive response associated with sustained transcriptional and epigenetic reprogramming. Such adaptation enables the prompt reactivation of acinar-to-ductal metaplasia (ADM) upon subsequent inflammatory events, thus efficiently limiting tissue damage via rapid decrease of zymogen production. We propose that since activating mutations of KRAS maintain an irreversible ADM, they may be beneficial and under strong positive selection in the context of recurrent pancreatitis.

Graphical Abstract

The association between tumors and inflammation is a long-established clinical observation (1). Many studies have demonstrated that the inflammatory microenvironment can promote tumor growth through the activation of survival and proliferation programs in cancer cells (2, 3). However, the reasons why inflammation, an evolutionarily conserved response to damage aimed at reestablishing tissue integrity upon injury, might be integral to tumorigenesis remain unknown, although recent evidence suggests that inflammation-induced chromatin changes may play an important role (4).

PDAC, a tumor characterized by poor prognosis (5), represents a distinctive example of cooperation between inflammation and activated oncogenes. Frequently developed in the context of chronic pancreatitis, PDAC is associated with an inflammatory microenvironment (6). As supported by a substantial body of evidence across a multitude of experimental models, induction of inflammation in pancreatic tissue expressing oncogenic KRAS hastens tumor progression (7, 8), inducing the appearance of neoplastic precursor lesions, such as ADM and pancreatic intraepithelial neoplasia (PanIN). These precursors can evolve into invasive tumors (9–11), although alternative models of PanIN-independent progression have been proposed (12, 13), for example when oncogenic signaling is activated in the ductal compartment (14).

Additionally, preneoplastic pancreatic alterations, specifically ADM, have been identified previously in acute and chronic pancreatitis in the absence of oncogene activation (15–19). Because ADM involves the rapid shutting down of the expression of pancreatic enzymes, it may represent an adaptive response to inflammation aimed at limiting tissue damage. In this conceptual framework, any genetic event able to promote or stabilize ADM, such as activating mutations of KRAS, may result in the positive selection of mutant cells within an inflamed tissue.

To better understand the relationship between inflammation and pancreatic tumorigenesis, we investigated the long-term consequences of transient inflammatory events in response to acute pancreatic damage and how resolved inflammation cooperates with activated oncogenes to drive transformation of normal epithelial cells.

Results

Transient inflammation promotes tumorigenesis long after its resolution

To investigate the long-term effects of inflammation on the transformation of normal pancreatic epithelial cells, we used caerulein (hereafter CAE), a decapeptide analog of cholecystokinin (20), to trigger damage and subsequent inflammation in a well-characterized PDAC mouse model (inducible KRAS, iKRAS) in which oncogenic KRAS^G12D expression is induced in the pancreas via doxycycline administration (TetO-LSL-Kras^G12D_ROSA26-LSLrtTa-IRES-GFP_p48-Cre) (21–23). To avoid confounding effects linked to chronic CAE administration, such as stromal and microenvironmental remodeling, we used a 2-day CAE administration protocol to induce acute inflammation (24), while control animals received intra-peritoneal injections with PBS on the same schedule (Fig. 1A). Immediately after CAE administration, we observed transient pancreatic inflammation, with edema and inter/intra-lobular infiltration of inflammatory cells, followed by rapid restoration of tissue integrity by day 7 (Fig. S1A). Immunostaining was consistent with the histological analysis, revealing that inflammatory cell infiltration (CD45-positive cells) and proliferation (Ki67 staining) present at day 1 (D1) post-CAE treatment returned to pre-CAE levels by day 28 (Fig. S1B), indicative of a complete histological resolution. In addition, we observed strong nuclear staining of two pivotal transcriptional activators of inflammation, NF-κB (25) and STAT3 (26), in epithelial and stromal cells immediately following the induction of pancreatitis (Fig. 1B and Fig. S1C–E). The normalization of these two transcription factors, together with recovery of the normal pancreatic histology, suggested that the inflammatory response was extinguished one week post-CAE treatment. Therefore, we explored the effects of KRAS^G12D after resolution of this single inflammatory event by inducing its expression at 28 days after the initial CAE protocol and monitoring tumor development (Fig. 1A). CAE-treated mice developed tumors with high penetrance and succumbed to disease earlier than control animals (median survival of 190 days in CAE-treated versus undefined in control animals; p=0.01) (Fig.1C and Fig. S1F). This observation was also confirmed by magnetic resonance imaging (Fig. S1G) and histological analysis at different time points over a period of 4 months after KRAS activation, which demonstrated an earlier onset and larger neoplastic lesions in CAE-treated animals (Fig.1D–F). Strikingly, the cooperation between oncogenic KRAS and resolved inflammation extended far beyond 28 days, as we documented accelerated tumorigenesis even when KRAS was activated 3 months after a single inflammatory event (median survival of 217 days in CAE-treated versus undefined in untreated animals; p=0.0026) (Fig. 1G–I). This phenomenon was independent of the modality used to induce inflammation (Fig. S1H) and appeared to have similar penetrance as conventional protocols of accelerated pancreatic tumorigenesis in which inflammation is induced in the context of ongoing oncogenic signaling (Fig. S1H). Overall, these data demonstrate that transient inflammatory events have persistent effects on normal epithelial cells and can cooperate with oncogene activation long after their resolution.

Figure 1. Transient inflammation promotes tumor progression long after resolution.

A) Schematics representing the experimental design. Briefly iKRAS mice are treated for two days (-D1, D0) with caerulein (CAE) to induce acute pancreatitis and followed for 4 weeks. When pancreata fully recovered from pancreatitis (D28), CAE-treated and control mice were put on doxycycline to induce the expression of mutated KRAS and observed for tumor development. B) Immunofluorescence for NF-κB (p-Ser536, red), E-cadherin (green), and DAPI (blue) in pancreatic samples at different time points before and after CAE treatment. Scale bar=200 μm for main images, scale bar= 50 μm for insets. n=3 mice per group. C) Kaplan-Meier survival plot of mice previously exposed to inflammation (caerulein, n=21) or control mice (untreated, n=10) after KRAS induction. D) Schematic representing the experimental design for (E) and (F). In brief, KRAS was induced in iKRAS mice at 1 month after acute pancreatitis, and pancreata were sampled for histological analysis at 2, 3, and 4 months after KRAS induction. Mice that did not receive acute inflammation were used as control. E) Histopathological evaluation and quantification of KRAS-driven changes over time. For each time point, 4 images are shown per experimental group: upper images, H&E staining at low magnification (left) and corresponding quantification mask (right); lower images, high magnification details of the H&E (left) and corresponding quantification mask (right). Tumor lesions in high magnification H&E panels are highlighted by orange lines. Color code for quantifying mask: Red=tumor lesion, Yellow=normal pancreatic tissue, Black=spleen or lymphoid tissues. Scale bar=2000 μm in upper images, scale bar=100 μm in lower images. F) Quantification of tumor burden (ratio of lesions to normal pancreas) over time as in (E). Data are shown as mean ± SD, n=3 mice per time point per group. Statistical significance was assessed with unpaired Student’s t-test. ns= non-significant. G) Schematic representing the experimental design for (H) and (I). In brief, KRAS is induced in iKRAS mice at 3 months after acute pancreatitis. Mice without acute inflammation are included as controls. Animals were followed for tumor development. H) Kaplan-Meier survival analysis of mice exposed to inflammation (caerulein, n=22) or control mice (untreated, n=12) 3 months before KRAS induction as described in (G). I) H&E staining shows pancreas histology of matched samples as in (H). Scale bar= 200 μm.

Long-term effects of resolved inflammation are cell-autonomous

To determine whether differences in outcome between CAE-treated and untreated animals resulted from epithelial cell-autonomous effects or from the influence of enduringly activated stroma, we established epithelial cultures from pancreata of mice 4 weeks after acute CAE treatment or PBS treatment in control animals. Pancreatic cells were cultured as spheroids, an effective culturing system that allows the expansion of 3D epithelial structures. Spheroids derived from CAE-treated wild-type animals, although numerically and morphologically similar (Fig. S2A–C), showed an increased size with respect to spheroids derived from control animals, suggesting that epithelial cells previously exposed to inflammation can expand more efficiently in vitro (Fig. 2A,B). This difference was further exacerbated in 2D conditions where only cells derived from CAE-treated mice were able to reach confluency after the first passage. Replicative exhaustion in primary cultures has been linked to senescence (27), and suppression of senescence programs during pancreatitis has been reported (8). We thus evaluated markers of senescence, including β-Gal activity, histone H3 lysine 9 trimethylation (H3K9me3), and p16 expression, which revealed suppression of the senescence program in CAE-treated compared to control pancreata (Fig. 2C–E), indicating that inflammation-induced reprograming can allow epithelial cells to evade senescence.

Figure 2. Cell-autonomous effects of resolved inflammation.

A) Quantification of spheroid size derived from pancreata of wild-type mice recovered from inflammation (CAE) or controls (CTRL). Size was evaluated as pixel log10 scale (9 fields for each condition from three independent experiments; CAE n=69, CTRL n=58; p<0.01). Representative images of spheroids are shown. Scale bar= 50 μm. B) Flow cytometry analysis and quantification of BrdU incorporation by epithelial spheroids derived from wild-type mice pre-exposed or not to caerulein. DAPI is used for DNA content. Data are shown as mean ± SD of three independent experiments. Statistical significance is assessed with unpaired Student’s t-test (p=0.022). C) β-Gal staining of 2D epithelial cultures derived from pancreata of wild-type mice recovered from inflammation (CAE) or control animals (CTRL). Scale bar =100 μm. D) Immunoblots for p16 and vinculin on epithelial cells derived from wild-type pancreata recovered from inflammation (CAE) or controls (CTRL). E) Immunofluorescence for H3k9me3 (Red) on epithelial cultures derived from pancreata of P48-Cre_mT/mG mice recovered from inflammation (CAE) or controls (CTRL). GFP (Green) indicates pancreatic origin of the cells. Cell nuclei are stained with DAPI (Blue). Scale bar =20 μm. F) Schematic representing the experimental design. Briefly, spheroids derived from iKRAS pancreas recovered from animals with pancreatitis (4-weeks recovery) or controls were orthotopically injected into recipient animals never exposed to inflammation. Then KRAS expression was induced and mice followed for tumor development. G) Kaplan-Meier survival plot of mice transplanted with iKRAS spheroids derived from pancreata recovered from pancreatitis (caerulein, n=7) and control mice (untreated, n=5) after KRAS induction. H) H&E staining and immunofluorescence for GFP (Green) and CD45 (Red) of orthotopic tumors and corresponding liver metastases developed from animals injected with spheroids derived from the pancreata of animals previously exposed to inflammation. Cell nuclei are stained with DAPI (Blue). Scale bar=200 μm or scale bar= 50 μm for main or inset images, respectively.

In parallel, pancreatic epithelial spheroids derived from iKRAS animals were orthotopically transplanted into inflammation-naïve recipients (250,000 cells/animal) after 5 weeks in culture, and KRAS was induced (Fig. 2F). Mice that received cells derived from CAE-treated pancreata developed with high penetrance aggressive tumors (Fig. 2G) characterized by poorly differentiated histology and metastasis to the liver (Fig. 2H and S2D, left panels). Although these tumors did not display classical features of pancreatic differentiation, the expression of GFP and the exclusion of CD45 immunoreactivity confirmed their pancreatic origin (Fig. 2H and S2D, middle and right panels). In line with this observation, upon digestion and subcutaneous re-transplantation of isolated primary tumor cells, the tumors acquired a more differentiated histology, as indicated by the organization of epithelial structures and acquisition of CK19 expression (Fig. S2E). Together, these data indicate that the inflammation-induced long-lasting epithelial modifications that cooperate with oncogenic signaling are cell-autonomous and maintained by pancreatic epithelial cells irrespectively of the microenvironment.

Transient inflammatory events induce sustained gene expression and chromatin reprogramming

Next, to identify the molecular changes in epithelial cells responsible for the long-term effect on pancreatic tumorigenesis, we performed single-cell transcriptomic analysis on WT pancreata at different time points (1, 7, and 28 days) after treatment with CAE as well as on normal animals as a control. To minimize inter-animal variability, for each time point we pooled pancreata from three independent animals, and samples were depleted of CD45-positive cells to maximize epithelial contribution especially at early time points after inflammation. Exocrine epithelial cell populations were identified by differentiation markers (Fig. S3A,B), and subsequent clustering analysis revealed the kinetics of transcriptomic deregulation specifically in the acinar compartment upon inflammation (Fig. 3A). Indeed, acinar cells underwent profound transcriptomic changes immediately after inflammation (day 1). Among them, an impairment of pancreatic digestive functions, including suppression of acinar zymogens (such as amylase, elastase, and pancreatic ribonuclease) and the concomitant upregulation of ductal markers (such as cytokeratins 18 and 19 and Mucin1), were apparent at day 1 but resolved by day 28 (Fig. 3B). Notably, these changes are compatible with the molecular program of acinar to ductal metaplasia (ADM) (15). Beyond these transient changes, CAE-treated acinar cells retained a modified gene transcriptional program at day 28 after inflammation that was not observed in the ductal compartment (Fig. 3A). Ingenuity Pathway Analysis (IPA) of signature genes that define different acinar states uncovered the activation of programs related to cell death and survival, proliferation, and cell transformation specifically in CAE-treated animals (Fig. 3C, Table S1). Moreover, a set of genes coregulated during pancreatic embryonic development and tumorigenesis (28) was also significantly enriched (p<4.37E⁻¹²)(Fig. S3C). These findings support the notion that epithelial acinar cells maintain an adaptive response to tissue damage that involves the sustained activation of multiple gene expression programs, including embryonic programs reactivated during cancer progression.

Figure 3. Pervasive transcriptional deregulation and chromatin changes in epithelial cells recovered from inflammation.

A-D) Single-cell RNA sequencing of ex-vivo pancreatic cells before (CTRL) and after acute inflammation at different time points (day 1_D1, day 7_D7, and day 28_D28). A) Epithelial clusters (acinar and ductal compartments) as identified based on cell type-specific markers reveal transcriptomic deregulation after inflammation mainly in acinar cells. Specifically, acinar cells at D28 remained in a different transcriptional state with respect to control cells (left panel). Unsupervised clustering (right panel) recognized acinar cells at day 28 (D28) after inflammation (orange circle) belonging to separate clusters with respect to untreated control (CTRL) (blue circle), while ductal CTRL and D28 cells clustered together (cyan circle). B) Gene expression dynamics of acinar and ductal markers in acinar compartment across different time points. C) Diseases and biological processes scoring in pathway analysis (IPA) associated with genes expressed in acinar cells at D1, D7, and D28 vs CTRL. D) TF binding sequences enriched in motif analysis of promoter regions of deregulated genes at D28 vs CTRL in acinar cells. E-I) ATAC-seq of ex-vivo epithelial pancreatic cells before (CTRL) and after acute inflammation (D1, D7, and D28). E,F) Density plots and heat maps depicting dynamics of enriched open regions, proximal (E) or distal (F) to gene TSS, at D28 vs CTRL across different time points. G) Pathway analysis (GREAT) of enriched chromatin open region-associated genes at D28 vs CTRL. H,I) Motif enrichment analysis of open chromatin regions at D28 vs CTRL identifying overrepresented sequences recognized by transcription factor families in proximal (H) or distal (I) regions. J) Expression of transcription factors identified via motif enrichment analysis in single-cell RNA sequencing across different time points before and after acute inflammation (D, H and I).

To identify the regulatory networks responsible for this sustained transcriptional signature, we identified DNA sequence motifs statistically over-represented in the promoters of deregulated genes at day 28, which are recognized by distinct transcription factor families including Sp, Kruppel-like factors (KLF), E2F, and EGR (Fig. 3D, Table S2). To investigate whether the long-term transcriptomic response to inflammation is associated with persistent chromatin modifications, we mapped chromatin accessibility of pancreatic cells after inflammation. ATAC-seq analysis of pancreata at different time points during and after the inflammatory episode revealed increased chromatin accessibility at thousands of transcription start site (TSS)-proximal and TSS-distal genomic regulatory elements (corresponding to promoters and putative enhancers, respectively) in CAE-treated animals compared to controls (Table S3). Both gene promoters and putative enhancers that showed increased accessibility early during inflammation (day 1) either maintained or increased their accessibility after full recovery of the pancreatic tissue (day 28), demonstrating that the long-term transcriptomic response to a transient pancreatitis is accompanied by persistent chromatin modifications in acinar cells (Fig. 3E,F). Not surprisingly, functional analysis of genes associated with the open genomic regions identified biological processes involved in abnormal inflammation, development, and cancer (Fig. 3G). Moreover, motif enrichment analysis of differentially accessible regions mapped to the same families of transcription factors that we identified as enriched in the promoters of transcriptionally deregulated genes (Fig. 3H,I). Expression of some members of these transcription factor families by acinar cells after inflammation, as indicated by single cell RNA-seq data, further corroborates their putative role in sustaining the enduring response to inflammation (Fig. 3J and Fig. S3D). Taken together, these data demonstrate that, after normal pancreatic epithelial cells histologically recover from a transient episode of inflammation, they acquire a lasting adaptive response maintained by a persistent transcriptional and epigenetic reprogramming.

Persistent inflammation-induced reprogramming in acinar progenitor cells

Because on the one hand epithelial spheroids derived from post-CAE animals support KRAS-mediated tumorigenesis upon orthotopic transplantation (Fig. 2F–H) and on the other hand acinar cells are the major target of inflammation-induced reprogramming in vivo, we investigated the relationship between epithelial spheroids and acinar cells. Epithelial spheroids derive from Doublecortin-Like Kinase 1 (DCLK1)-positive progenitor cells (29–32). To confirm that in our experimental setting epithelial spheroids originated from DCLK1-positive cells, we used a mouse model in which a green fluorescent protein (ZsGreen) is expressed under the control of the Dclk1 gene promoter (Dclk1-DTR-ZsGreen) (29) (Fig. S4A), and demonstrated that the only cells in the pancreas able to generate spheroids were in the ZsGreen-positive fraction (Fig. 4A and S4B). To further demonstrate the role of DCLK1-positive cells in maintaining spheroid cultures, we used the diphtheria toxin (DT) receptor coexpressed alongside ZsGreen to specifically target DCLK1-positive cells in epithelial spheroids. Exposure to DT dramatically impacted the expansion of epithelial spheroids derived from Dclk1-DTR-ZsGreen mice but did not have any effect on spheroids derived from wild-type animals, confirming the specificity of the results and the critical importance of DCLK1-positive cells in spheroid maintenance (Fig. S4C). Because the vast majority of DCLK1-positive cells display acinar differentiation (29, 30, 33) and sustain acinar regeneration upon tissue damage in vivo (29), we functionally assessed the ability of DCLK1-positive progenitor cells from epithelial spheroids to regenerate normal pancreatic tissue upon damage. To this aim, we derived GFP-positive spheroids from P48_Cre_R26-mT/mG animals and then transplanted them orthotopically into syngeneic recipients in which pancreata were previously damaged by CAE treatment. Four weeks after implantation, GFP-positive acinar lobules were clearly detected in the pancreata, and their differentiation was confirmed by amylase immunostaining (Fig. 4B). Lineage-tracing experiments also demonstrated that after inflammation, spheroids derived from Mist1-Cre^ERT2-mT/mG animals had an intense positivity for GFP, suggesting that the DCLK1-positive cells at the origin of the spheroids have acinar properties (Fig. 4C). Together, these findings further corroborate previous observations that DCLK1-positive cells represent a reservoir of pancreatic progenitor cells able to sustain both epithelial spheroids in vitro and acinar regeneration in vivo upon tissue damage (29).

Figure 4. Progenitor cells undergo persistent reprogramming after inflammation, and EGR1 is a critical player in promoting tumorigenesis.

A) Relative spherogenic potential of cells sorted from single cell suspension of pancreata isolated from Dclk1-DTR-ZsGreen mice based on their fluorescence: ZsGreen positive (DCLK1+), ZsGreen negative (DCLK1-) (n=3). Data are shown as mean ± SD. Statistical significance was assessed by unpaired Student’s t-test. Scale bar= 50 μm. B) Green spheroids derived from untreated P48-Cre_mT/mG mice were orthotopically transplanted into pancreata of WT animals 48 hours after CAE treatment. Cryosections of pancreata from mice euthanized at 4 weeks after implantation revealed GFP-positive lobules (Green) colocalized with acinar functional marker amylase (Purple). Cell nuclei were stained with DAPI (Blue). Scale bar =20 μm. C) Upper panels: Mist1-CreERT2_mT/mG mice were treated with tamoxifen to induce nuclear transfer of Cre recombinase followed by acute inflammation induction. Pancreata were then collected and part of the tissues was stained against GFP (Green) and amylase (Red) to show efficiency of Cre recombination in acinar compartment. DAPI was used to counterstain nuclei (Blue). White arrow indicates a non-recombined ductal and stromal structure. Scale bar=200 μm or scale bar= 50 μm for main or inset images, respectively. Middle and bottom panels: the other part of collected tissues was used to establish epithelial spheroid cultures. The GFP-positivity of spheroids indicates the acinar origin. D-G) ATAC-seq of isolated DCLK1-positive cells before (CTRL) and after acute inflammation (D1 and D28). D,E) Density plot and heat map depicts dynamics of enriched open TSS proximal (D) or distal (E) regions at D28 vs CTRL across different time points in DCLK1-positive cells. F,G) Motif analysis of enriched chromatin open regions at D28 vs CTRL identifies over-represented sequences recognized by transcription factor families in proximal (F) or distal (G) regions. H,I) Bulk RNA-seq on wild-type mouse pancreatic spheroids derived from pancreas pre-exposed or not to inflammation. H) Pathway analysis (IPA) of genes associated with diseases or biological functions enriched after inflammation vs. control. Highest-ranked terms are shown. I) GSEA enrichment plots showing the hallmark signature Kras signaling and Development and Progression signature including genes coregulated during development and carcinogenesis in pancreatic cells (28). The p53 Pathway signature, which is enriched in down-regulated genes, is also shown. Genes are ranked from left to right based on signed p-value, with genes on the left showing significantly higher expression after inflammation treatment. NES, normalized enrichment score; FDR, false discovery rate. J) Motif enrichment analysis of TF binding site over-representation at promoters and distal regions. The over-represented families of TFs at the promoters of up-regulated (Up-P) and down-regulated (Down-P) genes in spheroid bulk RNA-seq relative to all Refseq genes are shown on the left. The right panel shows the over-represented TF families at the differentially acetylated TSS-distal regions in spheroid ChIP-seq (using the FANTOM5 collection of enhancers as background). The heat map shows the negative logarithm of the enrichment P-value determined by a two-tailed Welch’s t-test. K) Kaplan-Meier survival plot of Egr1-WT_iKRAS mice (n=6) and Egr1-null_iKRAS mice (n=9) after oncogenic KRAS induction at day 28 after acute inflammation. I) H&E staining of Egr1-WT_iKRAS mice and matched Egr1-null_iKRAS mice. Scale bar=500 μm or scale bar= 50 μm for main or inset images, respectively.

Next, to understand the gene regulatory bases underlying reprogramming, we performed ATAC-Seq on DCLK1-positive cells isolated from untreated or CAE-treated pancreata (Table S4). Compared to controls, DCLK1-positive cells isolated 28 days after CAE treatment were characterized by increased chromatin accessibility at thousands of genomic regions (Fig. 4D,E), and differentially accessible chromatin regions were enriched in motifs similar to those previously identified in acinar cells ex vivo (EGR, Sp, KLF, E2F) (Fig. 4F,G). We also characterized epithelial spheroids derived from wild-type animals and performed a transcriptomic analysis 9 weeks after CAE treatment or control, namely after 4 weeks of recovery in vivo prior to 5 weekly passages ex vivo (Fig. S4D, Table S5). IPA of deregulated genes in epithelial spheroids derived from CAE-treated animals showed activation of gene expression programs similar to those previously identified in acinar cells ex vivo, which included genes controlling development, organismal injury and abnormalities, cell movement, survival, and cancer (Fig. 4H). Gene Set Enrichment Analysis (GSEA) demonstrated that key PDAC tumorigenic pathways, such as KRAS and p53 as well as functional categories related to pancreatic embryonic development and tumorigenesis (28) were deregulated as well (Fig. 4I). Moreover, ChIP-sequencing experiments on paired spheroids from inflamed and non-inflamed pancreata using antibodies for H3K27Ac, which detects active genomic regulatory elements, revealed a large number of persistent differences in histone acetylation, 90% of which were located distally from gene promoters and thus represented putative enhancers (Fig. S4E, Table S6). DNA sequences statistically over-represented in the promoters of deregulated genes relative to all other RefSeq genes, as well as in the differentially acetylated distal regions (Fig. 4J), were characterized by the overrepresentation of motifs recognized by similar families of transcription factors and the expression of some of the TFs of these families was concordantly deregulated, suggesting a possible causative role (Fig. S4D). Of note, TF families like EGR, NK, FOX, and SOX overlapped with those identified in DCLK1-positive cells ex vivo (Fig. 4F,G), thus suggesting that inflammation-induced reprogramming of DCLK1-positive cells is maintained in culture conditions. Our data support a model in which, upon damage and tissue regeneration, the long-term epigenomic modifications and gene expression programs acquired by DCLK1-positive cells can be maintained in their differentiated progeny, represented by epithelial spheroids in vitro or acinar cells in vivo.

EGR1 promotes pancreatic tumorigenesis after recovery from inflammation

DCLK1-positive cells have been reported to be the cell of origin of pancreatic cancer (29, 30, 34–36). Indeed, the intense DCLK1 staining of tumors derived from the transplantation of epithelial spheroids is consistent with the notion that the reprogramming of pancreatic progenitor cells may play an important role in inflammation-promoted tumorigenesis (Fig. S4F). To better understand how inflammation-induced reprogramming impacts oncogenic KRAS-mediated transformation in pancreatic epithelial cells, we chose to study the biological relevance of EGR1, a transcriptional regulator of the early growth response (EGR) gene family involved in tissue regeneration (37) that consistently scored across all omics analyses reported above as one of the most prominent transcription factors involved in cell reprogramming. We first assessed EGR1 expression in pancreatitis tissue. EGR1 was highly expressed in the nuclei of the vast majority of epithelial cells 1 day after CAE treatment, whereas it was undetectable in control tissue. EGR1 expression persisted in pancreatitis tissue through 28 days post-CAE treatment, though with lower intensity (Fig. S4G,H). EGR1 expression was also upregulated in human samples of acute and chronic pancreatitis compared to normal pancreas tissue, as analyzed by immunostaining (Fig. S4I). Next, to determine whether EGR1 is involved in pancreatic tumorigenesis in the context of resolved inflammation, we generated Egr1-null iKRAS animals by crossing Egr1^−/− mice with TetO-LSL-Kras^G12D_ROSA26-LSL-rtTA-IRES-GFP_p48-Cre animals. Although inflammatory cell recruitment was not affected in absence of EGR1 (Fig. S5A), oncogenic KRAS activation in matched Egr1^+/+ and Egr1^−/− iKRAS animals 4 weeks after inflammation revealed that Egr1^+/+ iKRAS animals developed tumors with high penetrance and had dramatically shorter survival than Egr1^−/− iKRAS mice (310 days median survival in Egr1^+/+ animals versus undetermined in Egr1^−/− animals; p<0.01) (Fig.4K). As expected, when mice were euthanized at the endpoint of the study, histological analysis of Egr1^−/− iKRAS pancreata revealed minimal tumor burden (Fig.4L), strongly suggesting that EGR1 is a critical player in pancreatic tumorigenesis and epithelial cell reprogramming induced by inflammation.

IL-6 mediates epithelial reprogramming during inflammatory events

To test whether epithelial reprogramming is dependent on the activity of inflammatory cells and to identify mediators responsible for this phenotype, we cultured epithelial spheroids derived from iKRAS pancreata with medium conditioned by CD45-positive cells isolated from mice with acute pancreatitis. After one week, spheroids were transferred to conventional medium and maintained in culture for additional 4 weeks to minimize the acute effects of cytokine exposure (Fig. 5A). Spheroids exposed to CD45-conditioned medium or control spheroids were then orthotopically transplanted into recipient mice, and KRAS expression was induced. Only mice injected with CD45-conditioned cells developed tumors that histologically resembled those obtained from transplantation of spheroids derived from CAE-treated pancreas (Fig. 5B and S5B upper panel). The epithelial origin of these tumors was confirmed by GFP positivity and the mutual exclusivity of CD45 staining (Fig. S5B middle and lower panels). Thus, epithelial cells undergo reprogramming in vitro in response to soluble molecules released by the inflammatory cells that mediate inflammation-induced changes in the pancreatic epithelium in vivo. ELISA analysis of CD45-conditioned medium revealed high concentrations of IL-6 and G-CSF (Fig. 5C). Because the G-CSF receptor is not expressed in pancreatic cells according to our data set (Table S5), we posited that IL-6, for which a role in PDAC progression is supported by a large body of evidence (11, 38–41), was likely a critical mediator of epithelial reprogramming. Exposure of spheroids to CD45-conditioned medium, as well as to recombinant Hyper-IL6, a potent chimeric molecule able to engage gp130 trans-signaling (41), confirmed strong induction of STAT3 phosphorylation at Tyr705 (Fig. 5D). Similarly, in vivo immunostaining analysis detected expression of IL-6-positive infiltrating cells and nuclear phospho-STAT3 signal in virtually all acinar cells in pancreatic samples immediately after CAE-treatment (D1) (Fig. 5E and S5C).

Fig. 5. IL-6 is a mediator of epithelial memory.

(A) Schematic representing the experimental design. Briefly, spheroids derived from iKRAS mice were cocultured in the presence or absence of CD45-positive cells isolated from animals with acute pancreatitis. After 1 week, conditioned spheroids were moved to conventional medium for another 4 weeks and then transplanted orthotopically into recipient mice and KRAS induced. (B) Kaplan-Meier survival plot of mice transplanted with conditioned (CD45, n = 5) or nonconditioned spheroids (CTRL, n = 7) after KRAS induction. (C) Cytokine array of medium conditioned for 1 or 7 days with CD45. Absorbance for different antibodies is reported. Data are shown as mean ± SD. (D) Immunoblotting for pSTAT3 (phospho-Tyr⁷⁰⁵), STAT3, and vinculin in spheroids exposed to CD45 conditioned medium (top) or hyper-IL-6 200 ng/ml (bottom) for the indicated amounts of time. (E) Immunofluorescence staining for IL-6 (red), pSTAT3 (green), and DAPI (blue) in pancreatic samples at day 1 after CAE treatment showing multiple pSTAT3 nuclear–positive cells, including many acinar structures (yellow dashed lines), interspersed among IL-6–positive cells. Scale bar is 120 μm. (F) CyTOF immunophenotyping of CD45-positive cells infiltrating the pancreas during acute pancreatitis. tSNE plots for CD68, CD11b, F4/80, and IL-6 are presented. (G) Immunoblotting for pSTAT3 (phospho-Tyr⁷⁰⁵), STAT3, EGR1, and vinculin in spheroids exposed to hyper-IL-6 200 ng/ml for 24 hours and then sampled at the indicated time points after hyper-IL-6 wash-out. NT, nontreated. (H) Pancreas from WT or Il6-null mice at day 1 after acute inflammation were harvested and immunostained for EGR1 (red). Cell nuclei were counterstained with DAPI (blue). Green channel (BG), although unstained, has been acquired and used to highlight tissue architecture and vessels. Scale bars are 200 μm. (I) Quantification of nuclear signal as pixel log10 intensity for EGR1 as in (H). Data are shown as violin plots (n = 3 mice per group). Statistical significance was assessed by unpaired Student’s t test. AU, arbitrary units.

Mass cytometry immunophenotyping of CD45 cells recruited to the pancreas upon acute CAE exposure (D1) revealed a massive infiltration of macrophages (CD68+, F4/80+, CD11+) (Fig. 5F) with only a marginal contribution of lymphoid cells (CD4, CD8, B220, NK1.1) (Fig. S5D). Furthermore, tSNE representation of immunoreactivity for IL-6 completely overlapped with CD68, F4/80, and CD11 markers, identifying macrophages as the major source of IL-6 production in vivo (Fig. 5F). Next, we measured the expression of the key transcription factor EGR1 by immunoblot analysis that revealed its strong and sustained upregulation after exposure of spheroids to Hyper-IL6 for 24 hours (Fig. 5G). To determine the role of IL-6 in EGR1 expression in vivo, we analyzed pancreata from Il6-null mice. As anticipated, even though inflammatory cell recruitment was not affected (Fig. S5E), EGR1 expression was markedly impaired in Il6-null pancreas tissue after CAE treatment compared with Il6-proficient controls (Fig. 5H,I). These in vitro and in vivo findings demonstrate that IL-6 is a key mediator of the inflammation-induced reprogramming of epithelial cells.

Acinar to ductal metaplasia is facilitated by epithelial memory to limit tissue damage

As any adaptive process, epithelial memory of previous inflammation should improve organismal fitness. Because of the transcriptional reprogramming of the acinar compartment in vivo, one possibility is that such memory provides a defense mechanism in case of recurring inflammatory events that are common in the pancreas and that would otherwise result in the repeated release of pancreatic enzymes and cumulative tissue damage. To understand how a discrete inflammatory episode can influence subsequent inflammatory events, we rechallenged wild type animals that had recovered from CAE-induced acute pancreatitis with a second inflammatory trigger (Fig. 6A). Early evaluation of pancreatic enzymes (Fig. 6B) as well as lactate dehydrogenase (LDH, a marker of cell lysis) (Fig. 6C) in the blood of wild-type mice 24 h after pancreatitis induction revealed that both enzymes in the rechallenged group were comparable to control animals. This suggests that a sustained adaptive response triggered by the first inflammatory event attenuated pancreatic damage induced by a second episode of acute inflammation. Indeed, at the histological level, only the pancreata of animals receiving the first inflammatory trigger had extensive acinar damage (Fig. S6A), as further confirmed by immunostaining for cleaved caspase 3 (CC3) (Fig. 6D). Unexpectedly, the pancreata of rechallenged animals responded to the second inflammatory event by undergoing an extensive acinar-to-ductal metaplasia (ADM), as indicated by the increased CK19/amylase staining ratio and the associated morphological changes (Fig. 6E). Such ADM was much more pervasive than that observed in animals exposed to a single challenge and completely manifested within 48 hours post-CAE administration. In spite of its pervasive character, ADM was completely resolved by day 7 post-CAE administration, as demonstrated by the full recovery of normal pancreatic tissue (Fig. S6B).

Figure 6. ADM as a physiological and reversible adaptation to limit tissue damage.

A) Schematic representing the experimental design. To investigate the role of epithelial memory, wild type or iKRAS mice were rechallenged with a second acute pancreatitis after complete recovery from a previous one. Pharmacologic modulation of ADM or KRAS induction was achieved by treating mice with EGF, MEK inhibitor, or doxycycline (KRAS induction) 2 days before and during the administration of caerulein. B,C) Concentrations of amylase (B) and LDH (C) detected in the peripheral blood at 24 h after the induction of acute pancreatitis (D0) in WT mice; untreated mice (CTRL, n=3), mice without memory after a single inflammation (Single, n=3), mice with memory after rechallenge (Rechallenge, n=3). Data are shown as mean ± SD. Statistical significance was assessed with one-way ANOVA. D) Immunofluorescence for cleaved caspase 3 (CC3-Red) and DAPI (Blue) of pancreata at 24 h after the induction of acute pancreatitis (D0) in WT mice with (Rechallenged) or without memory (Single inflammation). Green channel (BG), although unstained, was acquired and used to highlight tissue architecture and vessel. Scale bar= 200 μm or scale bar=100 μm for main or inset images, respectively. Quantification of cleaved caspase-3 in single inflammation and rechallenged group is shown (right panel). Each dot represents CC3-positive area per field. n=4 mice per group. Statistical significance was assessed by unpaired Student’s t-test. E) Immunofluorescence for CK19 (Green), amylase (Red), and DAPI (Blue) at 24 h (Day 1) after a 2-day inflammation in wild-type mice (Untreated) or mice with (Rechallenge) or without memory (Single inflammation). Scale bar=200 μm or scale bar=50 μm for main and inset images, respectively. Quantification of CK19/Amy ratio in the same experimental groups is shown in the lower panel. Each dot represents CK19/Amy ratio per field. n=3 mice per group. Data are shown as mean ± SD. Statistical significance was assessed with one-way ANOVA. F) Upper panels: H&E staining of pancreata of iKRAS mice at day 1 after rechallenge in presence/absence of pharmacological treatment with EGF, MEK inhibitor, or induction of KRAS. Scale bar=100 μm. Middle panel: Immunofluorescence for CK19 (Green), amylase (Red), and DAPI (Blue) at 24 h (Day 1) after rechallenge in presence/absence of pharmacological treatment with EGF, MEK inhibitor, or induction of KRAS. Scale bar=200 μm or scale bar= 50 μm for main or inset images, respectively. Lower panels: Damage evaluation in iKRAS-rechallenged mice, immunofluorescence for cleaved caspase 3 (CC3-Red) and DAPI (Blue) at 24 h (Day 1) after rechallenge in presence/absence of pharmacological treatment with EGF, MEK inhibitor, or induction of KRAS. Green channel (BG), although unstained, has been acquired and used to highlight tissue architecture and vessels. Scale bar=200 μm or scale bar= 50 μm for main or inset images, respectively. G) Extent of ADM in (F) middle panels is quantified as CK19/Amy ratio. Each dot represents CK19/Amy ratio per field. n=3 mice per group. Data are shown as mean ± SD. Statistical significance was assessed with one-way ANOVA. H) Pancreatic damage quantification evaluated as cleaved caspase-3-positive area (Log 10 scale), same setting as in (F) lower panels. Each dot represents CC3-positive area per field. n=3 mice per group. Data are shown as mean ± SD. Statistical significance was assessed with one-way ANOVA. I) Concentrations of LDH detected in the peripheral blood at 24 h (Day 1) after rechallenge in presence/absence of pharmacological treatment with EGF, MEK inhibitor, or induction of KRAS (n=3 for each group). Data are shown as mean ± SD. Statistical significance was assessed with one-way ANOVA.

Thus, the sustained adaptive response triggered in the pancreatic epithelium by an acute inflammatory event resulted in a markedly attenuated response to subsequent inflammatory episodes. Such decreased tissue damage was accompanied by the rapid dedifferentiation of acinar cells that lasted for the length of the stimulus, and from which the tissue appeared to promptly and completely recover.

To explore the hypothesis that ADM is a physiologic, fast, and reversible adaptation mediated by epithelial memory that limits the detrimental effects of repeated pancreatitis, we evaluated the effects of pharmacological modulation of ADM in iKRAS animals subjected to repeated inflammation. Because ADM is mediated by the activation of MAPK signaling (18, 42, 43), we counteracted or promoted ADM formation with a highly selective clinical MEK1/2 inhibitor (trametinib) or EGF (a MAPK activator), respectively (Fig. 6A). Mice that were pretreated with EGF before and during CAE rechallenge had a further increase of ADM formation with respect to control mice rechallenged with CAE alone (~3-fold relative area increase, p<0.01), with decreased tissue damage as indicated by CC3 immunostaining (~8-fold, p<0.01) (Fig. 6F–H and Fig. S6C) and LDH concentrations in peripheral blood (Fig. 6I). Conversely, animals pre-treated with trametinib developed minimal ADM (Fig. 6F,G and Fig. S6C), accompanied by a very severe pancreatitis with massive apoptosis and extensive acinar loss (Fig. 6F–I). Taken together, these data support a model in which sustained epithelial adaptation in response to inflammation involves the facilitation of ADM. By blocking the production of acinar zymogens during sequential inflammatory events, ADM provides strong protection from tissue damage.

Because ADM has protective effects against pancreatic damage, we posited that selection of mutations that confer constitutive activation of MAPK signaling, such as mutations of KRAS, may be beneficial and thus be under strong positive evolutionary pressure. Toward an initial evaluation of this possibility, we studied the impact of inducing mutant KRAS prior to a second inflammatory event. Indeed, in animals with epithelial memory, constitutive activation of KRAS signaling prior to the second CAE exposure resulted in massive ADM (Fig. 6F,G and Fig. S6D) and virtually no tissue damage (Fig. 6F–I).

Discussion

Inflammation evolved as a nonspecific damage response coordinating activation of innate immunity with tissue repair to accelerate parenchymal regeneration and the restoration of tissue integrity after injury (39, 44). Whereas the cellular and molecular events that lead to the activation of the immune compartment during an inflammatory response have been extensively characterized, the short- and long-term effects of inflammation on parenchymal cells are much less understood. Although quiescent stem cells resident in tissues can be activated by a variety of damage signals during injury (29, 45–47), only recently have adaptations of the stem cell compartment in response to inflammation-inducing damage been reported. In two seminal papers, it was shown that the regenerative capability of stem cells is strongly increased following tissue damage, suggesting the existence of a cellular memory-like phenotype, namely a sustained adaptive response, in the stem cell compartment (48, 49). In this context, the ability to sustain the acute adaptive response would represent an evolutionary advantage, because it enables faster recovery after injury. However, the long-term effects of inflammation on tissue properties, and specifically on the resilience of tissues exposed to repeated inflammation, have never been assessed. In the specific case of the pancreas, the production of highly active and potentially tissue-destructive enzymes by acinar cells brings about a high risk of widespread organismal damage in case of inflammation-induced acinar damage. Using a model of acute pancreatitis, we demonstrated that a short inflammatory episode has a long-lasting effect on pancreatic epithelial cells that results in marked protection of tissue integrity. Specifically, upon a second discrete inflammatory event, acinar cells promptly underwent massive ADM, a dedifferentiation process that leads to the accumulation of progenitor-like cells that have ductal features and are thus unable to secrete zymogens (15, 30). This fast and reversible adaptation appears to be of critical importance in light of the pathogenesis of pancreatitis, wherein zymogens, normally excreted in the duodenum, are prematurely activated within the pancreatic tissue and also released in the circulation. During repeated inflammatory events, the proclivity of adapted acinar cells to activate the ADM program may serve two critical functions: to shut down zymogen production that would otherwise fuel further damage, and to prompt acinar cells to acquire ductal features that render them more resistant to the detrimental effects of enzymatic activation. That ADM represents a physiologic process limiting pancreatic damage during pancreatitis is an emerging concept (50) further supported by the sustained adaptation described here. Indeed, by facilitating a rapid acquisition of ADM, memory promptly counteracts damage, reducing inflammation and promoting tissue regeneration. Because ADM is mediated by the activation of MAPKs (18, 42, 43), stabilization of the metaplastic phenotype through constitutive signaling plausibly exerts an important protective role to limit tissue damage, making activating mutations of KRAS beneficial events in the context of recurrent pancreatitis. It is thus not surprising that activating mutations of KRAS are found in ~25–37% of chronic pancreatitis cases, as well as in the vast majority of early PanIN lesions (51, 52).

This study, consistent with recent observations in other organ contexts (53), unmasks a previously unappreciated role of oncogenic mutations in preserving tissue homeostasis during repeated damage. Such a positive effect, independent from the eventual contribution to tumorigenesis, can explain why some oncogenic mutations, such as KRAS in pancreas, represent a nearly universal event in specific disease contexts. In light of these observations, we speculate that pharmacological induction of ADM with MAPK agonists, such as TGFα and EGF, or epigenetic drugs might ameliorate pancreatitis by protecting acinar cells from tissue damage while also reducing the positive pressure to mutate KRAS and hindering the progression to PDAC.

Materials and Methods

Animal models

Mice were housed in a pathogen-free facility at the University of Texas MD Anderson Cancer Center (MDACC). All manipulations were performed under the Institutional Animal Care and Use Committee (IACUC)-approved protocol (00001843-RN00). Male and/or female animals were used for experiments.

iKRAS (TetO-LSL-Kras^G12D; ROSA26-LSLrtTa-IRES-GFP; p48_Cre) mice were generated as previously described (21) and provided by Dr. Haoqiang Ying.

Dclk1-DTR-ZsGreen was generated and provided by Timothy C. Wang’s laboratory as described here. The DTR-2A-Zsgreen-pA-FrtNeoFrt cassette was ligated into a pL451 plasmid. A BAC clone RP23–283D6 containing an approximately 50-kb 5′ sequence of the Dclk1 gene–coding region (CHORI) was isolated and transferred into SW105-competent cells. The correct sequence was confirmed by using restriction enzyme digestion and PCR in the region of interest. The purified DTR-2A-Zsgreen-pA-FrtNeoFrt with a probe containing a 75-bp sequence homologous to the BAC sequence directly upstream and downstream of the ATG in exon 2 of mouse Dclk1 gene was electroporated into SW105 Dclk1-BAC–containing cells. BAC DNA (Fig. S4A) was isolated, linearized, and then microinjected into the pronucleus of fertilized CBA × C57BL/6J oocytes at the Columbia University Transgenic Animal Core facility. One positive founder was identified and backcrossed to C57BL/6J mice.

B6.129(Cg)-Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo}/J (referred to as R26-mT/mG) mice were generated in Liqun Luo’s laboratory (54) and purchased from The Jackson Laboratory (Stock No: 007676). These animals have been bred with B6.129-Bhlha15tm3(cre/ERT2)Skz/J and B6 p48-Cre for tracing experiments.

B6.129S2-Il6tm1Kopf/J (referred to as Il6-null) mice were generated by Manfred Kopf and Georges Kohler (Max Planck Institut Fur Immunbiologie, Freiburg, Germany) (55) and purchased from The Jackson Laboratory (Stock No: 002650).

B6N;129-Egr1tm1Jmi/J (referred to as Egr1-null) mice were generated by Jeffrey Milbrandt, (Washington University School of Medicine, St. Louis) (56) and purchased from The Jackson Laboratory (Stock No: 012924).

B6.129-Bhlha15tm3(cre/ERT2)Skz/J (referred to as Mist1_Cre^ERT2) mice were generated by Stephen Konieczny (Purdue University) (57) and purchased from The Jackson Laboratory (Stock No: 029228).

p48-Cre (Ptf1a-Cre) mice have been generated as described in Kawaguchi et al., 2002 (58) and purchased from The Jackson Laboratory (Stock No: 023329). The mice were then backcrossed to C57BL/6J.

NCR-NU immunodeficient mice used as recipients for transplantation experiments of iKRAS spheroids have been purchased from Taconic.

All wild-type animals (C57BL/6J) used as controls or recipients for transplantation experiments of p48-Cre_ R26-mT/mG spheroids have been purchased from The Jackson Laboratory (Stock No: 000664).

Human samples

Human tissue slides containing cases of acute and chronic pancreatic inflammation were purchased from US Biomax, Inc. and used for immunofluorescence staining following the protocol described below.

Data and materials availability

All sequencing data related to the manuscript are available at GEO https://www.ncbi.nlm.nih.gov/geo/, using access number GSE180212. Further information and request for resources and reagents should be directed to the lead contact.