AID-expressing epithelium is protected from oncogenic transformation by an NKG2D surveillance pathway

Activation-induced deaminase (AID) initiates secondary antibody diversification in germinal center B cells, giving rise to higher affinity antibodies through somatic hypermutation (SHM) or to isotype-switched antibodies through class switch recombination (CSR). SHM and CSR are triggered by AID-mediated deamination of cytosines in immunoglobulin genes. Importantly, AID activity in B cells is not restricted to Ig loci and can promote mutations and pro-lymphomagenic translocations, establishing a direct oncogenic mechanism for germinal center-derived neoplasias. AID is also expressed in response to inflammatory cues in epithelial cells, raising the possibility that AID mutagenic activity might drive carcinoma development. We directly tested this hypothesis by generating conditional knock-in mouse models for AID overexpression in colon and pancreas epithelium. AID overexpression alone was not sufficient to promote epithelial cell neoplasia in these tissues, in spite of displaying mutagenic and genotoxic activity. Instead, we found that heterologous AID expression in pancreas promotes the expression of NKG2D ligands, the recruitment of CD8+ T cells, and the induction of epithelial cell death. Our results indicate that AID oncogenic potential in epithelial cells can be neutralized by immunosurveillance protective mechanisms.


Introduction
Activation-induced deaminase (AID) is the enzyme that initiates the reactions of secondary antibody diversification: somatic hypermutation (SHM) and class switch recombination (CSR) (Muramatsu et al, 2000). These reactions enable the generation of antibodies with increased affinity for antigen (SHM) and with diversified, specialized functions for antigen removal (CSR), and are therefore critical for a competent immune response (Di Noia & Neuberger, 2007;Stavnezer et al, 2008;Alt et al, 2013;Robbiani & Nussenzweig, 2013). Accordingly, defective AID activity promotes a Hyper-IgM immunodeficiency syndrome in humans (Revy et al, 2000). AID triggers SHM and CSR by direct deamination of cytosine nucleosides in the DNA of immunoglobulin genes, resulting in the generation of U:G mismatches (Petersen-Mahrt et al, 2002;Alt et al, 2013;Robbiani & Nussenzweig, 2013). These U:G mismatches are in turn processed by alternative repair pathways that ultimately lead in SHM to the fixation of a mutation, and in CSR to a DNA doublestrand break (DSB) and a recombination reaction (Di Noia & Neuberger, 2007;Stavnezer et al, 2008;Alt et al, 2013;Robbiani & Nussenzweig, 2013).
Activation-induced deaminase activity is not confined to immunoglobulin genes and can promote mutations and DSB followed by illegitimate chromosomal translocations in other regions of the genome (Ramiro et al, 2004(Ramiro et al, , 2006Liu et al, 2008;Robbiani et al, 2008Robbiani et al, , 2009. Importantly, chromosomal translocations are the hallmark of mature B-cell lymphomas, the most frequent of all human lymphomas. Indeed, AID deficiency delays the onset of lymphomagenesis in the mouse (Ramiro et al, 2004;Kovalchuk et al, 2007;Pasqualucci et al, 2008), establishing a direct link between AID collateral genotoxic activity and neoplastic transformation in B lymphocytes.
In the last few years, it has become clear that AID expression is not, as originally thought, exclusively restricted to activated B cells. AID expression has been reported in several tissues, including gastric, hepatic, and gut epithelia (Endo et al, , 2008Matsumoto et al, 2007;reviewed in Marusawa et al (2011)). AID expression in these tissues is most frequently associated with inflammatory conditions and the activation of the NF-jB pathway Matsumoto et al, 2007) and has been claimed to promote the accumulation of mutations in epithelial cells (Matsumoto et al, 2007(Matsumoto et al, , 2010Takai et al, 2009;reviewed in Marusawa et al (2011)). Given that chronic inflammation in epithelial tissues predisposes to cancer development (Mantovani et al, 2008), the finding that the mutagenic activity of AID can be induced in an inflammatory context has fostered the idea that AID might contribute to or even constitute the link between inflammation and cancer (Takai et al, 2012;reviewed in Marusawa et al (2011)).
Several gain-of-function mouse models have been generated to address the contribution of AID to neoplastic transformation. Ubiquitous AID overexpression led mostly to early T cell neoplasia (Okazaki et al, 2003), hampering a thorough analysis of other malignancies. In contrast, B-cell-specific AID overexpression did not result in lymphomagenesis (Muto et al, 2006;Robbiani et al, 2009) unless the tumor suppressor p53 was removed (Robbiani et al, 2009). However, to date, the impact of specific AID expression in epithelial tissues, classically subject to inflammation-induced neoplastic transformation, has not been addressed. Here, we aimed to test this possibility directly by generating conditional knock-in models of AID overexpression. AID expressed in colon and pancreas epithelia was not sufficient to promote carcinogenesis, in spite of being expressed at high levels and displaying genotoxic activity. Instead, AID triggered the expression of NKGD2 ligands and the recruitment of immune cells and promoted a cytotoxic response and cell death. Our data indicate that the oncogenic potential of AID in epithelial cells is neutralized by an immunosurveillance pathway that prevents the expansion of pretumoral cells.

Inflammation-induced AID does not contribute to carcinogenesis
Inflammation is known to play a critical role in the etiology of colorectal and pancreatic ductal adenocarcinoma (reviewed in Feagins et al (2009) ;Vonderheide & Bayne (2013)). To investigate whether inflammatory conditions promote AID expression in these tissues, we treated human epithelial cell lines derived from colorectal adenocarcinoma (LoVo and SW480) and pancreatic adenocarcinoma (AsPC and PaTU) with the pro-inflammatory cytokine TNF-a and measured AID expression by qRT-PCR. TNF-a stimulation increased AID mRNA expression in all cell lines analyzed (Fig 1A  and B). To assess whether primary, non-transformed cells were also able to express AID in response to inflammatory stimuli, we generated explants from mouse pancreatic acinar cells and treated them with TNF-a. As with the human tumor cells, mouse primary epithelial cells expressed AID upon exposure to TNF-a ( Fig 1C). TNF-a treatment typically induced 4-30-fold increases in AID mRNA levels in the different cell types tested, consistent with previous findings in liver, gastric and colorectal cell lines (Endo et al, , 2008Matsumoto et al, 2007). Together, these data confirm previous results showing that inflammatory stimuli can trigger AID expression in cell lines originated from human colorectal adenocarcinoma (Endo et al, 2008), and show that pancreatic adenocarcinoma cells and primary pancreatic cells are also responsive to TNF-a treatment.
Inflammation-induced AID expression has been proposed to contribute to or even be the leading cause of some epitheliumderived tumors, such as colorectal adenocarcinoma (Marusawa et al, 2011). To address whether endogenous AID expressed in epithelium under inflammatory conditions could contribute to carcinogenesis, we made use of the well-established model of dextran sulfate sodium (DSS)-induced colitis-associated cancer (CAC) (Cooper et al, 2000). AID À/À mice or AID +/À littermates were treated for 10 cycles with 3% DSS and evaluated by pathological criteria. We found that the frequency of oncogenic lesions was not significantly different in AID À/À versus AID +/À mice ( Fig 1D). We conclude that endogenous AID does not significantly contribute to colorectal adenocarcinoma in the DSS-induced CAC model.

Conditional AID expression in epithelial cells does not promote adenocarcinoma development
The absence of a significant contribution of endogenous AID to carcinogenesis in DSS-treated mice could be explained by an insufficient amount of AID in this model. Indeed, AID expression is known to be limiting for its activity in B cells (Sernandez et al, 2008), and AID levels in B cells are typically 100-1,000 fold higher than those detected in epithelial cells under inflammatory conditions (unpublished observations). To directly evaluate whether AID expression qRT-PCR analysis of AID expression in pancreatic explants from wild-type mice (n = 3). *P = 0.0242. D AID À/À or AID +/À mice were treated with 3% DSS for 10 cycles, and colonic sections were analyzed by histologic inspection after H/E staining. Graphs represent mean frequency values of adenoma and adenocarcinoma lesions of five independent experiments. n = 28 (AID À/À males); 35 (females); 23 (AID +/À males); 25 (females). P-value: male: 0.8; female: 0.246. can contribute to carcinogenesis, we generated two mouse models for conditional AID expression in epithelial cells of colonic and pancreatic origin (Fig 2A). We introduced an AID-GFP-encoding cassette in the endogenous Rosa26 locus preceded by a transcriptional stop flanked by two loxP sites (R26AID +/KI mice). To achieve specific expression of AID in epithelial cells, we bred R26AID +/KI mice with mice expressing the Cre-recombinase under a villin promoter, which specifically drives expression in colon (el Marjou et al, 2004) (R26AID +/KI Villin-CRE +/TG mice), or the pancreasspecific Ptf1 (p48) gene (Kawaguchi et al, 2002) (R26AID +/KI p48-CRE +/KI mice). R26AID +/+ Villin-CRE +/TG and R26AID +/+ p48-CRE +/KI mice were used as controls. To confirm that the Rosa26 AID-GFP cassette was functional, we first evaluated the expression of the reporter protein GFP by immunofluorescence in colon of R26AID +/KI Villin-CRE +/TG mice and pancreas of R26AID +/KI p48-CRE +/KI mice ( Fig 2B). GFP was expressed in R26AID +/KI Villin-CRE +/TG colon and R26AID +/KI p48-CRE +/KI pancreas but not in control mice ( Fig 2B) or in other tissues (not shown). We next measured AID transcript levels by qRT-PCR. In R26AID +/KI Villin-CRE +/TG and R26AID +/KI p48-CRE +/KI mice, the amount of AID in the targeted epithelial tissues was similar to that found in B cells activated in vitro with LPS + IL4, whereas AID expression in control mice remained at background level ( Fig 2C). AID is thus expressed in the epithelium of R26AID +/KI Villin-CRE +/TG colon and R26AID +/KI p48-CRE +/KI pancreas at levels known to be functional in B cells.
To assess the contribution of AID to adenocarcinoma development, we monitored tumor incidence in R26AID +/KI Villin-CRE +/TG and R26AID +/KI p48-CRE +/KI mice. The onset of pancreatic and colorectal adenocarcinoma in a variety of mouse models ranges from 5-6 months to 1-1.5 years (Fodde & Smits, 2001;Aguilar et al, 2004;Martinelli et al, 2015). Therefore, to avoid confounding results arising from spontaneous tumorigenesis in very old mice, we set analysis end points at 75-100 weeks. Survival of R26AID +/KI Villin-CRE +/TG mice was indistinguishable from that of R26AID +/+ Villin-CRE +/TG littermate controls (Fig 2D, left). Likewise, survival of R26AID +/KI p48-CRE +/KI did not differ from that of R26AID +/+ p48-CRE +/KI controls (Fig 2D,right). To rule out the presence of early malignancies in aged animals, we performed thorough pathological analysis of colon and pancreas sections of all animals, but could not detect any tumor development in R26AID +/KI Villin-CRE +/TG and R26AID +/KI p48-CRE +/KI animals at 75-100 weeks ( Fig EV1). Expression of AID in colon or pancreatic epithelial cells is thus not sufficient to promote tumor development.

AID generates mutations and DNA double-strand breaks in pancreatic epithelium
The failure of AID expression to trigger tumorigenesis prompted us to evaluate its activity in epithelial cells. We first analyzed the in vivo mutagenic activity of ectopically expressed AID. The primary target sequences for AID mutagenic activity are immunoglobulin genes; although other genes are known to be susceptible to AIDinduced mutagenesis, this occurs at much lower rates (~10 À4 mutations/bp) and the mechanisms responsible for this susceptibility are poorly understood. One of the best-characterized requirements for AID activity is that the target sequence be transcriptionally active (Chaudhuri et al, 2003;Ramiro et al, 2003;Pavri & Nussenzweig, 2011). To simplify the mutagenesis analysis, we made use of the p48 pancreatic AID expression model to take advantage of the known low complexity transcriptome of acinar cells (MacDonald A Schematic of the constructs used for conditional expression of AID in epithelial cells. An AID-IRES-GFP cassette preceded by a transcriptional STOP flanked by LoxP sites was introduced by homologous recombination within the endogenous Rosa26 locus (R26AID +/KI mice, top). R26AID +/KI mice were bred with Villin-CRE and p48-CRE mice to achieve specific AID expression in colon and pancreas, respectively (bottom). B GFP immunofluorescence in colonic and pancreatic tissue from R26AID +/KI VillinCRE +/TG and R26AID +/KI p48CRE +/KI mice. Scale bar: 50 lm. C qRT-PCR analysis of AID expression in colonic and pancreatic tissue from R26AID +/KI VillinCRE +/TG and R26AID +/KI p48CRE +/KI mice. n = 5 (R26AID +/+ VillinCRE +/TG ); 4 (R26AID +/KI VillinCRE +/TG ); 2 (R26AID +/+ p48CRE +/KI ); 2 (R26AID +/KI p48CRE +/KI ). LPS+IL4-stimulated B cells are shown as a positive control (n = 2). Bars show mean values AE SEM normalized to LPS+IL4treated B cells. D Kaplan-Meier survival curves for R26AID +/KI VillinCRE +/TG (left) (n = 47 (R26AID +/+ VillinCRE +/TG ); 38 (R26AID +/KI VillinCRE +/TG )) and R26AID +/KI p48CRE +/KI mice (right) (n = 39 (R26AID +/+ p48CRE +/KI ); 23 (R26AID +/KI p48CRE +/KI )). . AID preferentially targets the consensus hotspots WRCY/RGYW and particularly AGCT motifs (Rogozin & Kolchanov, 1992;Pham et al, 2003;Perez-Duran et al, 2012). Based on this, we analyzed the presence of mutations in 800-900 bp downstream of the transcriptional start site of two highly transcribed genes in pancreas, Elastase1 (Ela1) and Elastase2 (Ela2a), by next-generation sequencing, which allows large number of mutations to be analyzed at a very high depth (Perez-Duran et al, 2012). This analysis revealed that mutations are specifically accumulated at the Ela1 and Ela2 genes in R26AID +/KI p48-CRE +/KI animals (Fig 3A and B) at frequencies similar to those of other non-Ig genes in B cells (Liu et al, 2008). AID activity was verified for Ela1 by conventional Sanger sequencing (Table 1). In contrast to previous reports (Matsumoto et al, 2007), we did not detect AID-induced mutations at the tumor suppressor gene Trp53, where mutation frequency was identical in R26AID +/KI p48-CRE +/KI mice and R26AID +/+ p48-CRE +/KI controls (Table 1).
To assess whether AID activity in R26AID +/KI p48-CRE +/KI mice leads not only to mutations but also to more aggressive lesions, such as DSBs, we quantified c-H2AX, a histone phosphorylation produced in response to this type of DNA damage. For this analysis, we generated acinar-cell explants from R26AID +/KI p48-CRE +/KI and control mice, stained them with anti-cH2AX, and quantified the intensity of staining per nucleus by high-throughput microscopy (HTM). We found that AID expression in R26AID +/KI p48-CRE +/KI mice promoted a significant increase in the levels of c-H2AX (Fig 3C  and D), indicating that AID generates DSBs in this cellular context.

AID induces NKG2D ligands, T cell recruitment and apoptotic cell death in pancreas
Activation of the DNA damage response (DDR) pathway induces the expression of NKG2D ligands in epithelial cells, which are in turn recognized by NKG2D receptors expressed by NK cells and subsets of T cells (Diefenbach et al, 2001;Gasser et al, 2005;Champsaur & Lanier, 2010;Raulet et al, 2013). This cross talk promotes the elimination of precancerous cells and is therefore a mechanism to prevent tumor development (Guerra et al, 2008). Given that AID expression in pancreas promotes mutations and DNA damage without leading to tumor development, we sought for the evidence of precancerous cells and found that pancreas from aged R26AID +/KI p48-CRE +/KI mice contained more proliferating cells, as assessed by Ki67 staining, than control pancreas (Fig 4A), indicating that pancreatic AID expression leads to an abnormal rate of cell division. The epithelial identity of Ki67 + cells was confirmed both by morphology ( Fig 4A, magnified micrograph on the right) and by staining with the epithelium-specific anti-cytokeratin 8 antibody (Figs 4B and EV2). We next asked whether the NKG2D immune surveillance pathway could be in play in R26AID +/KI p48-CRE +/KI mice. To test this hypothesis, we first analyzed the expression of the Raee NKG2D ligand in epithelial cells from pancreatic explants of R26AID +/KI p48-CRE +/KI and control mice by flow cytometry. Acinar cells from R26AID +/KI p48-CRE +/KI mice expressed higher levels of Raee than their control littermates, although the difference was not statistically significant ( Fig 4C). To assess whether RAE ligands were expressed by pancreatic cells in vivo, we prepared pancreas extracts from aged (75-week-old) R26AID +/KI p48-CRE +/KI mice and controls and measured the amount of five RAE isoforms by droplet digital PCR (ddPCR). With this technique, each sample is fractioned into thousands of droplets, in which PCR amplification reactions occur independently, thereby increasing the sensitivity and quantitative potential of the amplification. Amplification of RAE isoforms was detected in more drops from R26AID +/KI p48-CRE +/KI

A B
Mutation frequency   samples than from controls (Fig 4D), indicating that AID promotes the expression of NKG2D ligands in pancreas, most likely as a result of DSB and DDR. We found that primary explants from R26AID +/KI p48-CRE +/KI tended to be more sensitive to NK-mediated killing than R26AID +/+ p48-CRE +/KI littermate controls (Fig 4E), indicating that NKG2D ligand expression in AID-expressing pancreas is functional. We next asked whether the expression of RAE ligands promoted the recruitment of immune cells to R26AID +/KI p48-CRE +/KI pancreas in vivo. Hematoxylin-eosin staining of pancreas sections from aged mice clearly revealed the presence of immune infiltrates in AID-expressing pancreas of R26AID +/KI p48-CRE +/KI mice (Fig 5A). The composition of these immune infiltrates was analyzed by antibody staining to detect macrophages (F4/80), B cells (Pax5) and T cells (CD3). The vast majority of cells in the immune infiltrates of R26AID +/KI p48-CRE +/KI mice were CD3 + T cells (Fig 5B), with only a negligible contribution from B cells and macrophages (not shown). To discount age-related effects, we analyzed 20-weekold mice, finding that the accumulation of T cell infiltrates is detectable in these young animals (Fig 5C). The main NKG2D-expressing T cell subset is the CD8 + population (Raulet et al, 2013), and immunofluorescence analysis of immune infiltrates revealed that a high proportion of the CD3 + infiltrate is composed of CD8 + T cells (Fig 5C), a finding consistent with the reported recruitment of CTL cells to pancreatic islets transgenically expressing Raee (Markiewicz et al, 2012). Finally, we found that aged R26AID +/KI p48-CRE +/KI mice had significantly higher levels of pancreatic TNF-a mRNA than control littermates (Fig 5D), indicating that AID promotes the expression of effector cytotoxicity. Consistently, R26AID +/KI p48-CRE +/KI pancreas contained cells undergoing apoptotic cell death, detected by caspase-3 immunohistochemistry (Fig 5E, P = 0.054). Together, these results indicate that heterologous AID expression in pancreas promotes a cytotoxic response, most likely arising from the generation of genotoxic activity and NKG2D ligand expression and the recruitment of NKG2D-expressing CTL cells.

Discussion
In recent years, the finding that inflammatory cues induce AID expression in epithelial cells has boosted interest in the notion that AID might promote carcinogenesis and even be the causative link between inflammation and neoplastic transformation (Marusawa et al, 2011;Takai et al, 2012). In this regard, our analyses of AID expression in response to TNF-a have confirmed previous data in colonic cell lines and have expanded these observations to pancreatic cell lines. In addition, we found that primary pancreatic epithelium is also sensitive to TNF-a, suggesting that AID A Representative images of Ki67 staining in pancreas from aged (75-week-old) R26AID +/+ p48CRE +/KI (top) or R26AID +/KI p48CRE +/KI mice (bottom). Detail is shown on the right. Scale bar: 100 lm. Graph shows quantification of Ki67-positive epithelial cells per mm 2 of tissue (n = 8). *P = 0.0266. B Representative immunofluorescence staining of 20-week-old R26AID +/KI p48CRE +/KI mice: blue, DAPI; red, Ki67; green, CK8. Scale bar: 50 lm. Detail is shown on the bottom. C Quantitative FACS analysis of RAE expression in pancreatic explants from R26AID +/+ p48CRE +/KI and R26AID +/KI p48CRE +/KI mice. Left: Graph shows mean fluorescence intensity. Each dot represents an individual mouse. n = 9 (R26AID +/+ p48CRE +/KI ); 8 (R26AID +/KI p48CRE +/KI ). P = 0.077. Right: Representative FACS staining for RAE in explants from R26AID +/+ p48CRE +/KI (red) and R26AID +/KI p48CRE +/KI mice (black). D Analysis of RAE expression by ddPCR in aged (75-week-old) mice. Data are presented as the percentage of positive drops normalized to the mean control value. Each point represents an individual mouse and shows the mean amplification from two independent experiments n = 7 (R26AID +/+ p48CRE +/KI ); 9 (R26AID +/KI p48CRE +/KI ). *P = 0.012. E Analysis of killing activity. Primary explants of pancreatic cells from R26AID +/+ p48CRE +/KI or R26AID +/KI p48CRE +/KI mice were cultured with primary NK cells, and killing activity was assessed as described in Materials and Methods (n = 4). P = 0.19.
Data information: Statistical differences were analyzed by two-tailed unpaired Student's t-test. We wanted to further explore the physiological relevance of AID expression in promoting epithelium malignant transformation. We found that AID deficiency does not reduce the incidence of oncogenic lesions in an inflammation-induced carcinoma model. Our results contrast with the finding that AID deficiency reduces colon carcinogenesis in IL10 À/À mice (Takai et al, 2012). It may be that the effect reported by Takai et al is not specifically driven by epithelial cells, but rather by B cells in IL10 À/À mice, a possibility that could be tested using conditionally rather than constitutively AIDdeficient animals.
Previous reports have claimed that AID heterologous expression leads to epithelial cell neoplasia in various tissues (Endo et al, 2008;Takai et al, 2009Takai et al, , 2012Marusawa et al, 2011). In those studies, overexpression was always achieved with transgenes, none of which was epithelium specific, and the incidence of neoplasias was extremely low, and varied with the insertion sites and with ongoing mouse generations (Okazaki et al, 2003). In contrast, here we have developed epithelium-specific conditional knock-in models, thus avoiding both widespread expression and transgene-derived artifacts. Remarkably, our models allowed B-cell-like AID expression levels, a basic prerequisite for AID to be competent, given that its levels are rate limiting in its native context (Sernandez et al, 2008;Takizawa et al, 2008). We found that AID expressed ectopically in pancreatic cells is able to mutate non-immunoglobulin genes and to generate genotoxic DSBs, suggesting that in this context, there are no obvious mechanisms for negative regulation of AID activity, as they were previously proposed for transgenic AID expression in B cells (Muto et al, 2006). AID mutagenic activity was detected in two highly expressed pancreatic genes (Ela1 and Ela2), which is expected from the well-established link between AID activity and transcription of its target genes (Chaudhuri et al, 2003;Ramiro et al, 2003;MacDonald et al, 2010;Pavri & Nussenzweig, 2011).
Data information: Statistical differences were analyzed by two-tailed unpaired Student's t-test.
EMBO Molecular Medicine Vol 7 | No 10 | 2015 ª 2015 The Authors that AID promotes an NKG2D immune response in B cells infected with the Abelson murine leukemia virus (Gourzi et al, 2006). The lymphomagenic potential of AID was previously shown to be dampened in B cells (Muto et al, 2006;Robbiani et al, 2009), where p53 exerts a cell-intrinsic tumor suppressor function (Robbiani et al, 2009). Here, we provide evidence of a further protective mechanism triggered by AID activity and carried out through an extrinsic immunosurveillance pathway in epithelial tissues. These data highlight the diversity of safeguarding events in AID-expressing cells and encourage a refined view of the previously acknowledged contribution of endogenous AID to epithelial-derived tumors.

Materials and Methods
Mice R26AID mice were generated by insertion of a construct encoding mouse AID cDNA into the Rosa26 IRES-GFP targeting vector (Nyabi et al, 2009). AID cDNA was PCR-amplified from C57BL/6 mice (primers: forward 5 0 -TTCTGTGAAGACCGCAAGGCT-3 0 ; reverse 5 0 -CCCTTCCCAGGCTTTGAAA-3 0 ), cloned into the pENTR/D-TOPO Gateway vector (Invitrogen), and subsequently recombined into the Rosa 26 targeting vector, in which the cloned construct is preceded by a loxP-flanked transcriptional stop cassette and followed by an internal ribosomal entry site and GFP. The construct was linearized with PvuI before electroporation into hybrid 129/C57BL/6 ES cells.
Clones positive for homologous recombination in the Rosa26 locus were identified by Southern blot of EcoRV-digested genomic DNA hybridized with a 5 0 -arm Rosa26 probe. R26AID mice were backcrossed to C57BL/6 background for 5 generations. R26AID mice were crossed with Villin-CRE +/KI mice (el Marjou et al, 2004) and p48-CRE +/KI mice (Kawaguchi et al, 2002) to promote expression of AID in colonic and pancreatic epithelial cells, respectively. Both Villin-CRE and p48-CRE mice were backcrossed to C57BL/6 for 5 generations. Balb/c AID À/À mice were generated by backcrossing AID À/À mice (Muramatsu et al, 2000) for 6 generations by speed congenics (Ramiro et al, 2004). Mice of both genders were used unless specified otherwise. All animals were housed in the Centro Nacional de Investigaciones Cardiovasculares animal facility under a 12-h light/dark cycle with food ad libitum.
Number of animals per group to detect biologically significant effect sizes was calculated using appropriate statistical sample size formula and indicated in the biometrical planning section of the animal license submitted to the governing authority. Blinding and randomization was not applicable to the animal studies. All animal procedures conformed to EU Directive 2010/63EU and Recommendation 2007/526/EC regarding the protection of animals used for experimental and other scientific purposes, enforced in Spanish law under RD 53/2013.

Cell lines and primary acinar cell culture
Primary pancreatic acinar cells were isolated and cultured as described in Gout et al (2013). Briefly, complete pancreas from 8-week-old mice was mechanically and enzymatically digested with collagenase to obtain isolated acinar structures. Acini were grown in Waymouth's medium supplemented with 2.5% FBS, 10 mM HEPES, 0.25 mg/ml trypsin inhibitor (Sigma) and 25 ng/ml of recombinant human epidermal growth factor (Sigma). PaTU-8988S, AsPC-1, LoVo and SW480 cells were grown in DMEM supplemented with 10% FCS and 10 mM HEPES. PaTU-8988S cell line was kindly provided by Dr Thomas Gress (University of Marburg). AsPC-1, LoVo and SW480 cell lines were obtained from the ATCC. All of them were mycoplasma negative. TNF-a (50 ng/ml) was added when indicated.

DSS-induced colitis-associated cancer (CAC) experiments
8-10-week-old Balb/c AID +/À and AID À/À (Ramiro et al, 2004) mice were given 3% dextran sulfate sodium salt (DSS, Sigma) in their drinking water for 5 days followed by regular drinking water for 10 days. Colon samples were obtained from these mice after 10 cycles of DSS treatment, and Swiss roll preparations were fixed, embedded in paraffin for section, and stained with hematoxylin/ eosin (H/E).

Immunohistochemistry
Pancreases were fixed in neutral-buffered 10% formalin solution (Sigma), embedded in paraffin blocks, and cut in 5-lM sections. H/E staining was performed using standard protocols. For immunohistochemistry, sodium citrate buffer was used for antigen retrieval. The following antibodies were used: polyclonal rabbit anti-human CD3 (Dako, 1/200), rabbit anti-Ki67 (Abcam, 1/100), and biotinylated goat anti-rabbit (Abcam, 1/200). Biotinylated antibody was detected with the ABC system using diaminobenzidine as substrate (Vector Laboratories). Images were acquired with a Leica DM2500 microscope fitted with a 20× magnification lens. Reactive cells from 10 microscope fields per pancreas were counted using ImageJ. Results are shown as the number of reactive cells per mm 2 of tissue.
Next-generation sequencing for detection of mutations DNA from three R26AID +/+ p48-CRE +/KI and three R26AID +/KI p48-CRE +/KI mice was extracted and amplified using the following oligonucleotides: Elastase1 (forward) 5 0 -GCA CAG CAT CTT TTG TTT GGG TAA-3′, (reverse) 5 0 -GGG GAC AGT GGT CTA CTC TCT-3 0 ; Elastase2a (forward) 5 0 -ACG AGT CCA GGA CAA TCA GAG A-3 0 , (reverse) 5 0 -TGA TAA GGC CAC TCA TAA AAA GGA-3 0 . Amplification reactions were carried out with 2.5 U of Pfu Ultra (Stratagene) in a 50-ll reaction with the following profile: 94°C for 5 min followed by 25 cycles at 94°C for 10 s, 60°C for 30 s, and 72°C for 1 min. For NGS sequencing, the PCR product from two reactions per mouse were pooled, and an equimolar amount of DNA from each of the two mouse genotypes was pooled and mixed.

c-H2AX staining and Opera acquisition
Primary pancreatic acinar cells were grown on 96-well plates (Perkin Elmer), and c-H2AX (Millipore, 1/500) immunofluorescence was performed using standard procedures. Images were automatically acquired from each well with the Opera High-Content Screening System (Perkin Elmer). A 40× magnification lens was used, and pictures were taken at nonsaturating conditions. Images were segmented using DAPI staining to generate masks matching cell nuclei, and the mean per-cell c-H2AX signal was calculated.
Samples were transferred into a 96-well plate (Eppendorf) and cycled in a thermal cycler (Bio-Rad) under the following conditions: 95°C for 10 min followed by 38 cycles of 94°C for 30 s, 58°C for 1 min, and a final step at 98°C for 10 min. After amplification, samples were transferred to a droplet reader (QX100 Droplet Digital PCR, Bio-Rad) from which positive-drop data were extracted with Quantasoft software. Results are represented as the proportion of positive drops in duplicates of each sample.

NK killing assay in vitro
Primary pancreatic acinar cells from R26AID +/+ p48-CRE +/KI and R26AID +/KI p48-CRE +/KI mice were isolated as described above. After 6 days of culture, acinar cells were trypsinized and stained with CFSE. NK cells were isolated by cell sorting from wild-type C57BL/6 mice. Both male and female aged 6-8 weeks were used as NK cells donors. NK cells and acinar pancreatic cells were cocultured for 4 h in the presence of IL2 (2,000 U/ml, Peprotech) at a 1:10 (target cell:NK effector cell) ratio. Killing was analyzed by staining with DAPI by flow cytometry. Data are presented as the proportion of CFSE + DAPI + cells normalized to the same population in cultures lacking NK cells.

Statistics
Statistical analyses were performed with GraphPad Prism (version 6.01 for Windows, GraphPad Software, San Diego, CA, USA) using two-tailed Student's t-test. P ≤ 0.05 was considered statistically significant. Error bars in figures represent standard error of the mean (SEM). Normal distribution of data was assessed by applying a D'Agostino & Pearson omnibus normality test. F-test was used to compare variances between groups. For the survival analyses, GraphPad Prism was used and the Mantel-Cox test was applied. Differences were considered statistically significant at P ≤ 0.05. The paper explained Problem AID is a DNA-modifying enzyme expressed in germinal center B cells and required for the adaptive immune response, and its activity is linked to B cell neoplasias. A number of studies have reported AID expression in epithelial cells, usually associated with inflammatory cues. However, the contribution of AID expression to carcinogenesis remains controversial.

Results
We have developed conditional mouse models to promote epithelialspecific AID expression in pancreas and colon and found that despite high levels of AID expression and activity, these mice do not develop carcinomas. Instead, AID genotoxicity in pancreas promotes a response that involves NKG2D ligand expression, CD8 + T cell recruitment, and epithelial cell death. Therefore, our data provide evidence that AIDexpressing epithelia are protected from neoplastic transformation by an NKG2D surveillance pathway.

Impact
Our results uncover a new protective mechanism counter-acting AID oncogenic potential through an extrinsic immunosurveillance pathway in epithelial cells. These findings challenge the view that AID could contribute significantly to inflammation-related carcinogenesis.