AID ‐expressing epithelium is protected from oncogenic transformation by an NKG 2D surveillance pathway

Research Article17 August 2015Open Access AID-expressing epithelium is protected from oncogenic transformation by an NKG2D surveillance pathway Arantxa Pérez-García Arantxa Pérez-García B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Pablo Pérez-Durán Pablo Pérez-Durán B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Thomas Wossning Thomas Wossning B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Isora V Sernandez Isora V Sernandez B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Sonia M Mur Sonia M Mur B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Marta Cañamero Marta Cañamero Roche Diagnostics GmbH, Penzberg, Germany Search for more papers by this author Francisco X Real Francisco X Real Epithelial Carcinogenesis Group, Spanish National Cancer Research Centre (CNIO), Madrid, Spain Search for more papers by this author Almudena R Ramiro Corresponding Author Almudena R Ramiro B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Arantxa Pérez-García Arantxa Pérez-García B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Pablo Pérez-Durán Pablo Pérez-Durán B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Thomas Wossning Thomas Wossning B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Isora V Sernandez Isora V Sernandez B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Sonia M Mur Sonia M Mur B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Marta Cañamero Marta Cañamero Roche Diagnostics GmbH, Penzberg, Germany Search for more papers by this author Francisco X Real Francisco X Real Epithelial Carcinogenesis Group, Spanish National Cancer Research Centre (CNIO), Madrid, Spain Search for more papers by this author Almudena R Ramiro Corresponding Author Almudena R Ramiro B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain Search for more papers by this author Author Information Arantxa Pérez-García1, Pablo Pérez-Durán1,4, Thomas Wossning1,5, Isora V Sernandez1,6, Sonia M Mur1, Marta Cañamero2, Francisco X Real3 and Almudena R Ramiro 1 1B Cell Biology Lab, Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC), Madrid, Spain 2Roche Diagnostics GmbH, Penzberg, Germany 3Epithelial Carcinogenesis Group, Spanish National Cancer Research Centre (CNIO), Madrid, Spain 4Present address: Department of Pathology, NYU Cancer Institute, New York University School of Medicine, New York, NY, USA 5Present address: Klinik für Tumorbiologie, Freiburg, Germany 6Present address: Otsuka Pharmaceutical, Barcelona, Spain *Corresponding author. Tel: +34 91 4531200; Fax: +34 91 4531245; E-mail: [email protected] EMBO Mol Med (2015)7:1327-1336https://doi.org/10.15252/emmm.201505348 PDFDownload PDF of article text and main figures. Peer ReviewDownload a summary of the editorial decision process including editorial decision letters, reviewer comments and author responses to feedback. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions Figures & Info Abstract 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. Synopsis AID expression in epithelial cells does not lead to inflammation-related carcinogenesis; rather, AID genotoxicity in pancreas promotes a protective NKG2D immunosurveillance pathway that seemingly prevents neoplastic transformation. Mouse models were developed for specific expression of AID in pancreatic and colonic epithelia. AID expression in epithelial cells promoted mutations and genotoxic lesions, but mice did not develop carcinomas. AID-expressing epithelium triggered an NKGD2 immunosurveillance response that very likely neutralized AID oncogenic potential. Our 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. 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 double-strand 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, 2006; Liu et al, 2008; Robbiani et al, 2008, 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, 2007, 2008; Matsumoto 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-κB pathway (Endo et al, 2007; Matsumoto et al, 2007) and has been claimed to promote the accumulation of mutations in epithelial cells (Matsumoto et al, 2007, 2010; Takai 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. Results 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-α and measured AID expression by qRT-PCR. TNF-α 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-α. As with the human tumor cells, mouse primary epithelial cells expressed AID upon exposure to TNF-α (Fig 1C). TNF-α 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, 2007, 2008; Matsumoto 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-α treatment. Figure 1. Inflammation-induced AID expression does not contribute to carcinogenesis A–C. AID expression was analyzed in colon and pancreatic human cell lines and in pancreas explants from C57BL/6 mice. Samples were treated as indicated with 50 ng/ml TNF-α. (A) qRT–PCR analysis of AID expression in LoVo and SW480 colon cell lines. n = 5 (LoVo); 3 (SW480). **P-value: LoVo: 0.0017; SW480: 0.0079. (B) qRT–PCR analysis of AID expression in AsPC and PaTU-8988S pancreatic cell lines (n = 2). *P-value: AsPC: 0.0369; PaTU-8988S: 0.0119. (C) qRT–PCR analysis of AID expression in pancreatic explants from wild-type mice (n = 3). *P = 0.0242. D. AID−/− or AID+/− 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−/− males); 35 (females); 23 (AID+/− males); 25 (females). P-value: male: 0.8; female: 0.246. Data information: All data are mean values ± SEM. Statistical differences were analyzed by two-tailed unpaired Student's t-test. Download figure Download PowerPoint Inflammation-induced AID expression has been proposed to contribute to or even be the leading cause of some epithelium-derived 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−/− mice or AID+/− 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−/− versus AID+/− 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 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+/KIVillin-CRE+/TG mice), or the pancreas-specific 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+/KIVillin-CRE+/TG mice and pancreas of R26AID+/KIp48-CRE+/KI mice (Fig 2B). GFP was expressed in R26AID+/KIVillin-CRE+/TG colon and R26AID+/KIp48-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+/KIp48-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+/KIVillin-CRE+/TG colon and R26AID+/KIp48-CRE+/KI pancreas at levels known to be functional in B cells. Figure 2. Heterologous AID expression does not promote carcinoma development 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). GFP immunofluorescence in colonic and pancreatic tissue from R26AID+/KIVillinCRE+/TG and R26AID+/KIp48CRE+/KI mice. Scale bar: 50 μm. qRT–PCR analysis of AID expression in colonic and pancreatic tissue from R26AID+/KIVillinCRE+/TG and R26AID+/KIp48CRE+/KI mice. n = 5 (R26AID+/+ VillinCRE+/TG); 4 (R26AID+/KIVillinCRE+/TG); 2 (R26AID+/+p48CRE+/KI); 2 (R26AID+/KIp48CRE+/KI). LPS+IL4-stimulated B cells are shown as a positive control (n = 2). Bars show mean values ± SEM normalized to LPS+IL4-treated B cells. Kaplan–Meier survival curves for R26AID+/KIVillinCRE+/TG (left) (n = 47 (R26AID+/+ VillinCRE+/TG); 38 (R26AID+/KIVillinCRE+/TG)) and R26AID+/KI p48CRE +/KI mice (right) (n = 39 (R26AID+/+p48CRE+/KI); 23 (R26AID+/KIp48CRE+/KI)). Download figure Download PowerPoint To assess the contribution of AID to adenocarcinoma development, we monitored tumor incidence in R26AID+/KIVillin-CRE+/TG and R26AID+/KIp48-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+/KIVillin-CRE+/TG mice was indistinguishable from that of R26AID+/+ Villin-CRE+/TG littermate controls (Fig 2D, left). Likewise, survival of R26AID+/KIp48-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+/KIVillin-CRE+/TG and R26AID+/KIp48-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. Click here to expand this figure. Figure EV1. Heterologous AID expression does not promote carcinoma development Representative H/E stainings in colonic tissue from 75-week-old R26AID+/+VillinCRE+/TG (top) and R26AID+/KIVillinCRE+/TG (bottom) mice. Scale bar: 500 μm. Representative H/E stainings in pancreatic tissue from 75-week-old R26AID+/+p48CRE+/KI (top) and R26AID+/KIp48CRE+/KI (bottom) mice. Scale bar: 500 μm. Download figure Download PowerPoint 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 AID-induced mutagenesis, this occurs at much lower rates (~10−4 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 et al, 2010). 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+/KIp48-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+/KIp48-CRE+/KI mice and R26AID+/+p48-CRE+/KI controls (Table 1). Table 1. Analysis of AID mutagenic activity by Sanger sequencing Genotype Total clones analyzed Mutations Total bp sequenced Frequency (×104) Elastase1 R26AID+/+ p48-CRE+/KI 84 4 70,018 0.571 R26AID+/KI p48-CRE+/KI 82 13 69,355 1.87 Trp53 R26AID+/+ p48-CRE+/KI 66 0 59,472 0 R26AID+/KI p48-CRE+/KI 59 1 53,936 0.185 Figure 3. AID expression in pancreas promotes DNA lesions A, B. Analysis of AID mutagenic activity in Elastase1 (A) and Elastase2a (B) by next-generation sequencing. Pancreatic DNA was isolated from pools of R26AID+/KIp48CRE+/KI and control R26AID+/+p48CRE+/KI 20-week-old mice, and then PCR-amplified with specific primers and sequenced as previously described (Perez-Duran et al, 2012). Graphs show cytosine mutation frequency overall (total) or at AGCT hotspots. n = 2. *P-value: Elastase1: 0.0382; Elastase2a: 0.009. C. HTM-mediated quantification of γH2AX intensities per nuclei in pancreas explants cultured in vitro for 6 days. Red lines show mean values. Results of two independent experiments are shown. ****P < 0.0001. D. Representative images of γH2AX staining in pancreas explants from R26AID+/+p48CRE+/KI (top) or R26AID+/KIp48CRE+/KI mice (bottom) (40× magnification). White arrow points a representative positive cell. Data information: Statistical differences were analyzed by two-tailed unpaired Student's t-test. Download figure Download PowerPoint To assess whether AID activity in R26AID+/KIp48-CRE+/KI mice leads not only to mutations but also to more aggressive lesions, such as DSBs, we quantified γ-H2AX, a histone phosphorylation produced in response to this type of DNA damage. For this analysis, we generated acinar-cell explants from R26AID+/KIp48-CRE+/KI and control mice, stained them with anti-γH2AX, and quantified the intensity of staining per nucleus by high-throughput microscopy (HTM). We found that AID expression in R26AID+/KIp48-CRE+/KI mice promoted a significant increase in the levels of γ-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+/KIp48-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+/KIp48-CRE+/KI mice. To test this hypothesis, we first analyzed the expression of the Raeε NKG2D ligand in epithelial cells from pancreatic explants of R26AID+/KIp48-CRE+/KI and control mice by flow cytometry. Acinar cells from R26AID+/KIp48-CRE+/KI mice expressed higher levels of Raeε 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+/KIp48-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+/KIp48-CRE+/KI 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+/KIp48-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. Figure 4. AID expression in pancreas promotes proliferation and NKG2D ligand expression Representative images of Ki67 staining in pancreas from aged (75-week-old) R26AID+/+p48CRE+/KI (top) or R26AID+/KIp48CRE+/KI mice (bottom). Detail is shown on the right. Scale bar: 100 μm. Graph shows quantification of Ki67-positive epithelial cells per mm2 of tissue (n = 8). *P = 0.0266. Representative immunofluorescence staining of 20-week-old R26AID+/KIp48CRE+/KI mice: blue, DAPI; red, Ki67; green, CK8. Scale bar: 50 μm. Detail is shown on the bottom. Quantitative FACS analysis of RAE expression in pancreatic explants from R26AID+/+p48CRE+/KI and R26AID+/KIp48CRE+/KI mice. Left: Graph shows mean fluorescence intensity. Each dot represents an individual mouse. n = 9 (R26AID+/+p48CRE+/KI); 8 (R26AID+/KIp48CRE+/KI). P = 0.077. Right: Representative FACS staining for RAE in explants from R26AID+/+p48CRE+/KI (red) and R26AID+/KIp48CRE+/KI mice (black). 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+/KIp48CRE+/KI). *P = 0.012. Analysis of killing activity. Primary explants of pancreatic cells from R26AID+/+p48CRE+/KI or R26AID+/KIp48CRE+/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. Download figure Download PowerPoint Click here to expand this figure. Figure EV2. Ki67 is expressed in pancreatic epithelial cellsRepresentative immunofluorescence stainings of pancreatic tissue of 20-week-old R26AID+/KIp48CRE+/KI mice: blue, DAPI; red, Ki67; green, CK8. Scale bar: 50 μm Download figure Download PowerPoint We next asked whether the expression of RAE ligands promoted the recruitment of immune cells to R26AID+/KIp48-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+/KIp48-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+/KIp48-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-week-old 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 Raeε (Markiewicz et al, 2012). Finally, we found that aged R26AID+/KIp48-CRE+/KI mice had significantly higher levels of pancreatic TNF-α mRNA than control littermates (Fig 5D), indicating that AID promotes the expression of effector cytotoxicity. Consistently, R26AID+/KIp48-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. Figure 5. AID expression in pancreas promotes immune infiltration and cell death Hematoxylin–eosin (HE) staining of pancreas from aged (75-week-old) R26AID+/+p48CRE+/KI and R26AID+/KIp48CRE+/KI mice. Left: Representative HE staining showing an immune infiltrate in a R26AID+/KIp48CRE+/KI mouse. Scale bar: 200 μm. Right: Quantification of number of foci per mm2 of tissue. n = 13 (R26AID+/+p48CRE+/KI); 21 (R26AID+/KIp48CRE+/KI). **P = 0.0098. CD3 immunohistochemistry of pancreas from 75-week-old R26AID+/+p48CRE+/KI and R26AID+/KIp48CRE+/KI mice. Left: Representative image of a CD3 infiltrate in a R26AID+/KIp48CRE+/KI mouse. Scale bar: 100 μm. Right: Qua