Transgenic Expression of Interferon-γ in Mouse Stomach Leads to Inflammation, Metaplasia, and Dysplasia

Transgenic Mice and Helicobacter felis Infection

Mouse IFN-γ cDNA was amplified from IMAGE clone ID 8733812 using forward primer 5′-CTACCTGACTGGATCCTCTGAGACAATGAACGCTAC-3′ and reverse primer 5′-GGAGTCGCTGCTGATTCGGATCCTTGACACATC-3′ and subcloned into the BamHI site to replace the Ctox cDNA in the H/K-Ctox transgene,²¹ which contains 1059 bp of mouse H/K ATPase β promoter.²² The resulting H/K-mIFN-γ transgene was verified by DNA sequencing, excised from vector with HindIII and SacII, and submitted to the University of Michigan Transgenic Animal Model Core for microinjection into F2 embryos from SJL × C57BL/6 parents. Eleven H/K-mIFN-γ founders were produced, and three exhibited severe SPEM associated with other histopathological changes. We studied mouse lines from founders 944 (H/K-IFN-γ⁹⁴⁴) and 53 (H/K-IFN-γ⁵³), with most data generated from line H/K-IFN-γ⁹⁴⁴. Transgenic mice were backcrossed onto a C57BL/6J background (The Jackson Laboratory, Bar Harbor, ME). For Helicobacter felis infections, 2-month-old C57BL/6J mice were gavaged three times over 3 days with 10⁸ bacteria (CS1 strain), in 100 μL of Brucella broth.

Histopathological Scoring

Stomachs were cut along the greater curvature, placed on filter paper, fixed in 4% buffered paraformaldehyde overnight at 4°C, cut into strips extending from the forestomach to the proximal duodenum, and transferred to 70% ethanol until processing and paraffin embedding. Sections (5 μm) were stained using hematoxylin and eosin (H&E) and examined by a board-certified veterinary pathologist (K.A.E.) blinded to the experimental groups. Scoring was performed to quantify the presence of inflammation, SPEM, parietal and chief cell atrophy, gland atrophy, dysplasia, and tumor development. Histological scoring was performed on a scale of 0 to 3; 0, absence of detectable inflammation; 1, multiple focal areas of inflammation confined to the lamina propria; 2, inflammation that was widespread and/or that extended to the submucosa; and 3, transmural infiltration. For quantification of SPEM, the absence of detectable SPEM was scored as 0; mild or multifocal SPEM was scored as 1; the presence of SPEM in most fields was scored as 2; and widespread SPEM in all fields was scored as 3. Loss (atrophy) of parietal and chief cells was graded as 0, no loss; loss of cells detected was graded as 1; cell loss easily identified was graded as 2; and severe loss with few parietal or chief cells remaining was graded as 3. Gland atrophy was scored according to the appearance of the glands: 0, no change; 1, small or thin glands; 2, decreased gland size with space between glands; and 3, glands widely spaced with many missing. Dysplasia was scored as positive if there was cellular atypia with disorganized glands. Elderly H/K-IFN-γ mice occasionally developed more advanced lesions in their antra. Lesions that were focal but largely sessile, well differentiated, and had minimal dysplasia were interpreted as polyps. One focal proliferative lesion with marked dysplasia was interpreted as an adenoma. A single lesion with marked dysplasia, cellular atypia, and a high mitotic index was interpreted as an adenocarcinoma. For data comparing phenotypes in the corpus and antrum, inflammation and hyperplasia were scored as described above, and dysplasia was scored as follows: 0, no gland or cellular atypia; 1, the presence of occasional disorganized glands; 2, the presence of many disorganized glands; and 3, cellular atypia or expansile lesion.

Immunostaining, Antibodies, and Flow Cytometry

Immunohistochemical analysis was performed as previously described,²³ with minor modifications. Primary antibodies detecting the following antigens were used for immunostaining: mucin 5AC (#MS-145; NeoMarkers, Fremont, CA); H/K-ATPase α subunit (#D031-3; Medical and Biological Laboratories, Nagoya, Japan); intrinsic factor (a gift from Dr. David Alpers; Washington University, St. Louis, MO); Ki-67 (#RM-9106; Thermo Scientific, Fremont, CA); cleaved caspase 3 (D175, #9661; Cell Signaling, Danvers, MA); CD45 (#14–0451; eBioscience, San Diego, CA); CD3 (#A0452; DakoCytomation, Glostrup, Denmark); F4/80 (#T-2006; BMA Biomedicals, Augst, Switzerland); myeloperoxidase (#N1578; DakoCytomation); STAT1 (#ab31369; Abcam, Cambridge, MA); pSTAT1 (Y701) (#9167; Cell Signaling); pSTAT3 (Y705) (#9145; Cell Signaling); β-catenin (#C7207; Sigma-Aldrich, St. Louis, MO); TFF2 (anti-spasmolytic polypeptide antibody, #ab49536; Abcam); smooth muscle actin (SMA) (#MS-113-P0; NeoMarkers); keratin 5 (AF138, #PRB-160P; Covance, Princeton, NJ); keratin 8 (TROMA-I; DSHB, Iowa City, IA); and IFN-γRβ (Ifngr2) (N-20, #sc-971; Santa Cruz, Santa Cruz, CA). Fluorescein-conjugated GSII lectin (Vector Laboratories, Burlingame, CA) was used to identify mucous-containing neck cells in normal stomach and SPEM in H/K-IFN-γ transgenic mice. The following antibodies were used for immunoblotting: PCNA (#RB-9055; Thermo Scientific); Shh (#sc-1194; Santa Cruz); H/K-ATPase β subunit (#D032-3; Medical and Biological Laboratories); cleaved Caspase 3 (#9661; Cell Signaling); and GAPDH (#sc-25778; Santa Cruz). Flow cytometry was performed as previously described,²⁴ with minor modifications. Antibodies for sorting included PE rat anti-mouse Ly-6G and Ly-6C (Gr1) (#561084; BD Biosciences Pharmingen, San Diego, CA) and PE-Cy7 rat anti-mouse CD11b (#561098; BD Biosciences Pharmingen).

Immunoblot Analysis

Mucosal scrapings from freshly harvested stomach were homogenized for 20 minutes on ice by vortexing every 5 minutes in ice-cold radioimmunoprecipitation assay (RIPA) lysis buffer containing phosphatase inhibitor cocktail and protease inhibitor cocktail, both from Roche Applied Science (Indianapolis, IN). Lysates were cleared by centrifugation at 12,000 × g for 15 minutes at 4°C, aliquoted, and stored at −80°C. A 50-μg quantity of protein per lane was separated by gradient SDS–polyacrylamide gel electrophoresis and transferred onto a polyvinylidene fluoride membrane. Immunoblotting and detection were performed as described,²³ with minor modifications.

RT-qPCR Analysis

Stomach tissue samples were obtained from three to six independent mice per group, as described in the text, and frozen in RNAlater (Qiagen, Hilden, Germany). Total RNA was extracted using an RNeasy Mini Kit (Qiagen); single-stranded complementary DNA was synthesized using 2 μg of total RNA with SuperScript III Reverse Transcriptase (Invitrogen, Carlsbad, CA), and 2 μL of a 1:5 dilution of cDNA was used in a 20-μL quantitative RT-PCR reaction mix containing 2 μL of SYBR Green (1:10,000 dilution; #S-7563; Invitrogen), 100 nmol/L of each primer, 20 nmol/L fluorescein (#170–8780; BioRad, Hercules, CA), 0.1 μL Platinum Tag polymerase (#10966–034, Invitrogen), and ultrapure water (#10977; Invitrogen). Primer sequences are available upon request. Real-time quantitation was performed in triplicate for each sample using the iCycler iQ5 Real-Time PCR System (Bio-Rad) and normalized to glyceraldehyde-3-phosphate dehydrogenase. Results are expressed as relative increase in the level of gene expression, compared to controls, using the threshold cycle (Ct) slope method.

Statistical Analysis

Statistical analysis was performed using GraphPad Prism (version 5) software. Statistical significance was determined using the two-tailed, unpaired, Student's t-test. Data are expressed as mean ± SEM. A P value of less than 0.05 was considered significant. Comparison of phenotypic scores for inflammation, hyperplasia, and dysplasia, in the corpus and antrum, was performed using Spearman's rank correlation coefficient. For quantification of histological changes over time, a Wilcoxon-Mann-Whitney test (nonparametric) was run with a multiple comparison P value adjustment.

Production of Stomach-Targeted H/K-IFN-γ Transgenic Mice

To determine the consequences of IFN-γ overexpression in stomach, we generated transgenic mice using a 1059-bp fragment of the H/K ATPase β subunit promoter²² to target mouse IFN-γ to gastric parietal and preparietal cells (Figure 1A). We focused our analysis on transgenic mouse lines derived from two founders, designated H/K-IFN-γ⁹⁴⁴ and H/K-IFN-γ⁵³. Gastric corpus isolated from both lines of mice exhibited elevated expression of IFN-γ mRNA, with levels in the H/K-IFN-γ⁹⁴⁴ line substantially higher than the H/K-IFN-γ⁵³ line (Figure 1B). Expression of IFN-γ target genes Mig (Cxcl9) and IP10 (Cxcl10) was also elevated (Figure 1C), indicating that IFN-γ overexpression elicited the expected transcriptional response. IFN-γ mRNA in H/K-IFN-γ⁹⁴⁴ mice was maintained at comparably high levels at 7 weeks and 15 months of age, whereas 15-month-old H/K-IFN-γ⁵³ mice expressed reduced levels of IFN-γ relative to 7 weeks, and this was reflected in reduced expression of IFN-γ target genes (Figure 1, B and C).

To estimate whether the level of IFN-γ signaling in our mice was in the same range as in the setting of gastritis, we compared the expression of IFN-γ target genes in stomachs of H. felis–infected mice and H/K-IFN-γ mice, relative to controls. Mig mRNA expression levels, relative to controls, were elevated 63-fold in H. felis–infected mice, 46-fold in H/K- IFN-γ⁹⁴⁴ mice, and 37-fold in H/K- IFN-γ⁵³ mice (Figure 1D). For the IFN-γ target gene IP10, expression levels relative to controls were elevated 2.6-fold in H. felis–infected mice, 17-fold in H/K-IFN-γ⁹⁴⁴ mice, and 14-fold in H/K-IFN-γ⁵³ mice (Figure 1D). Taken together, these data support the notion that the level of IFN-γ signaling in H/K-IFN-γ transgenic mice, assessed by the expression level of IFN-γ target genes, is generally comparable to that observed in H. felis–infected mice.

Grossly, the stomachs from both lines of mice revealed a thickened and enlarged corpus region, and more than 50% of transgenic mice also exhibited thickening of the squamous limiting ridge at the junction of the forestomach and corpus (Figure 2A, arrow; see also Supplemental Figure S1 at http://ajp.amjpathol.org). Interestingly, 38% of mice from line 53 also contained areas of ectopic squamous epithelium within the glandular corpus (Figure 2A, arrowheads; see also Supplemental Figure S1 at http://ajp.amjpathol.org). This was not detected in control or H/K-IFN-γ⁹⁴⁴ mice. Stomach weights from H/K-IFN-γ mice were higher than those from control mice at all time points examined (Figure 2B), and there was an approximately twofold increase in mucosal thickness of the corpus at both 5 and 15 months of age (Figure 2C). Overexpression of IFN-γ in the corpus of transgenic mice thus leads to sustained elevation of transcripts encoding IFN-γ and IFN-γ target gene transcripts; this is associated with gastric hypertrophy, focal squamous hyperplasia, and, in one line of mice, ectopic squamous tissue within the corpus.

Parietal Cell Atrophy and Spasmolytic Peptide-Expressing Metaplasia in H/K-IFN-γ Transgenic Mice

Stomachs from H/K-IFN-γ transgenic mice revealed regions with marked histological changes involving multiple gastric cell types as well as a prominent inflammatory infiltrate. H/K-IFN-γ mice contained strikingly fewer parietal cells and chief cells, which were replaced by cells with abundant, foamy cytoplasm resembling the morphology of metaplastic cells comprising SPEM (Figure 3, A and B), a putative preneoplastic lesion both in humans and in mice.¹⁵ Histopathological scoring of stomachs collected from control and H/K-IFN-γ mice at different ages quantified the presence of SPEM, parietal and chief cell atrophy, gland atrophy, and inflammation (Figure 3C), as described in Materials and Methods. Each of these histological features was either not detected or minimal in control mice but markedly altered in H/K-IFN-γ mice (Figure 3C).

The foamy cells in stomachs of H/K-IFN-γ transgenic mice contained abundant mucin identified by staining with PAS-Alcian blue (Figure 4, A and B), bound GSII lectin (Figure 4, C and D), and expressed TFF2 (Figure 4, E and F), supporting a diagnosis of SPEM. Expression of the mucous pit marker, Muc5ac, was generally reduced (Figure 4, G and H). Immunostaining for the α subunit of H/K ATPase (H/K ATPase α) revealed both a striking reduction in the number of parietal cells (Figure 4, I and J), in keeping with the histopathological scoring (Figure 3C). Similarly, the number of cells expressing the chief cell marker, intrinsic factor, was also greatly reduced (Figure 4, K and L). In contrast, there was a marked expansion of mesenchymal cell expressing smooth muscle actin (SMA) (Figure 4, M and N), a marker of myofibroblasts.

IFN-γ Expression in Stomach Leads to Increased Proliferation and Cell Death

Additional immunostaining revealed increased proliferation in the corpus of H/K-IFN-γ transgenic mice. In contrast to control mice, where expression of the proliferation marker Ki-67 is largely restricted to a narrow band of cells within the isthmus region of each gastric gland, in H/K-IFN-γ mice there was both an increase in the number of Ki-67–positive cells and a broader distribution, with expansion deeper into gastric glands and more superficially (Figure 5, A–D). Co-immunostaining for Ki-67 and lineage markers revealed that proliferating cells in H/K-IFN-γ mice did not express Muc5ac (Figure 5, E and F) or H/K ATPase β (Figure 5, G and H), but some cells that bound GSII lectin were also Ki-67 positive (Figure 5, I and J). Thus, although neither the differentiated mucous pit nor the parietal cells are induced to proliferate in response to stomach-targeted expression of IFN-γ, a subset of metaplastic cells do express Ki-67.

In addition to enhanced proliferative activity, immunostaining for cleaved caspase 3 revealed increased apoptosis in the corpus of H/K-IFN-γ transgenic mice, and occasionally these cells were also positive for H/K ATPase β (Figure 5, K–N), suggesting that the loss of parietal cells is due at least in part to increased cell death. Immunoblotting studies examining markers of proliferation, differentiation, and cell death confirmed immunostaining results. Mucosal lysates from corpus of H/K-IFN-γ transgenic mice contained higher levels of PCNA and cleaved caspase 3, with reduced levels of the H/K ATPase β subunit and Sonic hedgehog (Shh) (Figure 5O), both of which are expressed in normal parietal cells, relative to controls.

We performed immunostaining to localize the expression pattern of the β subunit of the IFN-γ receptor, Ifngr2, in control and H/K-IFN-γ transgenic mice (see Supplemental Figure S2 at http://ajp.amjpathol.org). We co-immunostained with fluorescent GSII lectin, as a recent study reported that Ifngr2 is localized to mucous neck cells.²⁵ In control mice, Ifngr2 co-localizes with GSII lectin but also stains most cells in the lower portion of the oxyntic glands in the gastric corpus, where chief cells are localized. In H/K-IFN-γ transgenic mice, epithelial expression of Ifngr2 is reduced or absent in regions with pronounced SPEM, and only a small subset of metaplastic GSII+ cells express appreciable levels of Ifngr2, in contrast to GSII+ mucous neck cells in the control stomach. Although Ifngr2 expression is reduced in regions with SPEM, it is not possible to ascertain in this setting whether this is due to chronic stimulation by IFN-γ with down-regulation of IFN-γ receptors, or to loss of some of the normal cell type(s) that express these receptors. Interestingly, scattered Ifngr-expressing cells are detected in regions of corpus of H/K-IFN-γ mice, and these appear to be nonepithelial (see Supplemental Figure S2 at http://ajp.amjpathol.org).

IFN-γ Overexpression in Stomach Has Biphasic Effects on Hedgehog Signaling Activity

Although early stages of gastric tumorigenesis are characterized by reduced Shh levels and lowered Hh pathway activity, likely due largely to the loss of the Shh-expressing parietal cells, progression to gastric cancer is associated with elevated levels of Hh ligands and signaling activity.^26,27 The reduced level of Shh protein in H/K-IFN-γ mice (Figure 5O) was associated with lower Hh pathway activity in regions with prominent SPEM, detected by x-gal staining of stomachs from 10-week-old H/K-IFN-γ;Gli1-lacZ Hh target gene reporter mice (see Supplemental Figure S3 at http://ajp.amjpathol.org). Quantitative PCR at 7 weeks of age revealed similar levels of mRNA encoding Shh, Indian hedgehog (Ihh), and the Hh target genes Gli1 and Hhip in the corpus of H/K-IFN-γ and control mice, with the target gene Ptch1 expressed at lower levels in H/K-IFN-γ⁹⁴⁴ mice than in controls (see Supplemental Figure S3 at http://ajp.amjpathol.org). At 15 months of age, however, statistically significant increases were noted in the expression of Shh and Ihh, as well as Ptch1 and Hhip, in H/K-IFN-γ transgenic mice relative to controls (see Supplemental Figure S3 at http://ajp.amjpathol.org).

Robust Inflammation, Cytokine Induction, and Activation of STAT1 and STAT3 in H/K-IFN-γ Transgenic Mice

Inflammatory cells in H&E-stained sections of H/K-IFN-γ mice were distributed in regions throughout the corpus in the mucosa and occasionally submucosa (Figure 3, A and B), and contained variable amounts of mononuclear cells, including plasma cells, and neutrophils (Figure 6, A and B). Immunostaining to detect CD45 confirmed increased numbers of leukocytes in the corpus of H/K-IFN-γ transgenic mice (Figure 6, C and D). The infiltrate included T lymphocytes (CD3-positive) (Figure 6, E and F), macrophages (F4/80 antigen-positive) (Figure 6, G and H), and neutrophils (myeloperoxidase-positive) (Figure 6, I and J).

The robust inflammation in H/K-IFN-γ mice was accompanied by elevated expression of mRNAs encoding markers for distinct subsets of T cells and multiple cytokines, measured at both 7 weeks and 15 months of age. Quantitative PCR revealed upregulation of transcripts encoding T helper cell markers T-bet and Foxp3, markers of Th1²⁸ and Treg²⁹ cells, respectively (Figure 7). In addition, H/K-IFN-γ mice expressed elevated levels of transcripts encoding the Th1-related cytokines tumor necrosis factor–α (TNF-α) and IL-12 (p40); as well as the Treg-related anti-inflammatory cytokine IL-10 (Figure 7). Transcripts encoding the pro-inflammatory cytokines IL-1β and IL-6, both implicated in tumorigenesis in stomach, were also upregulated (Figure 7). IL-11 was significantly elevated in H/K-IFN-γ⁹⁴⁴ mice, but only at 15 months of age.

H/K-IFN-γ mice did not express significantly higher levels of mRNAs encoding the Th17 marker RORγT and Th17 cytokines IL-17A, IL-17B; Th17-related IL-23 (p19); Th1-related IL-12 (p35); and the Th2 marker Gata3 and Th2 cytokine IL-4 (see Supplemental Figure S4 at http://ajp.amjpathol.org). These findings are in keeping with the inhibitory effects of IFN-γ on Th2 and Th17 development. Overexpression of IFN-γ in mouse corpus thus activates a robust, sustained inflammatory response with a predominant Th1 profile. Interestingly, despite markedly higher levels of IFN-γ in H/K-IFN-γ⁹⁴⁴ than H/K-IFN-γ⁵³ mice (Figure 1B), upregulation of mRNA for most of the cytokines examined, including those previously implicated in gastric metaplasia and transformation (IL-6, TNF-α, and IL-1β), was generally comparable in both lines of H/K-IFN-γ mice (Figure 7). This suggests that a threshold sufficient for near-maximal signaling may have been reached even in the lower-expressing H/K-IFN-γ⁵³ mice.

To test the requirement for acquired immunity in the phenotypic changes detected in H/K-IFN-γ mice, we generated mice deficient in B and T cells by breeding H/K-IFN-γ mice with mice carrying Rag1 null alleles.³⁰ The histopathological score for H/K-IFN-γ and H/K-IFN-γ; Rag1^−/− mice was not appreciably different (see Supplemental Figure S5 at http://ajp.amjpathol.org): epithelial changes and inflammation were present in both groups of mice. As expected, immunostaining for CD3 confirmed the absence of T-cells in Rag1^−/− and H/K-IFN-γ;Rag1^−/− mice (see Supplemental Figure S5 at http://ajp.amjpathol.org). These data show that the marked phenotypic changes seen in corpus of H/K-IFN-γ transgenic mice are not dependent on acquired immunity and reflect similar findings in H/K ATPase-targeted IL-1β-expressing mice, where T-cell function is dispensable but myeloid-derived suppressor cells (MDSCs) may play an important role in the development of gastritis and dysplasia.¹¹ In light of these findings, we performed flow cytometry to assess whether IFN-γ overexpression also leads to recruitment of MDSCs. As shown in Supplemental Figure S6 (available at http://ajp.amjpathol.org), the proportion of MDSCs in H/K-IFN-γ⁹⁴⁴ mouse stomach is 29-fold higher than in control mice (1.933% versus 0.0667%; P < 0.0001). There is also a 5.5-fold increase in MDSCs in H/K-IFN-γ⁵³ mice (0.3667%), relative to controls (0.0667%), but this is not statistically significant (see Supplemental Figure S6 at http://ajp.amjpathol.org). These results are in keeping with the concept that infiltrating MDSCs may contribute to the phenotype that develops in H/K-IFN-γ transgenic mice, but additional studies would be needed to rigorously address this possibility.

Given the central role of signal transducers of transcription (STATs) in cellular responsiveness to cytokines, we examined the expression and phosphorylation/activation status of STAT1 and STAT3 in stomach of H/K-IFN-γ mice. Immunostaining for STAT1, pSTAT1, and pSTAT3 revealed very low expression of these signaling molecules in control stomach. In contrast, there was marked upregulation of these proteins in affected regions of corpus of H/K-IFN-γ transgenic mice (Figure 8, A–I). Co-immunostaining with the epithelial cell-surface marker β-catenin revealed pSTAT1 primarily in epithelial cells, whereas pSTAT3 was detected both in epithelial and in nonepithelial cells (Figure 8, J and K). Cells expressing pSTAT3 included parietal cells, GSII lectin-binding SPEM cells, and SMA-positive mesenchymal cells (Figure 8, L–T), indicating that STAT3 signaling is activated in stromal and multiple epithelia cell populations in H/K-IFN-γ mice. Immunostaining results for STAT1 and STAT3 proteins were corroborated by immunoblotting (Figure 8U).

Preneoplastic Progression and Antral Tumor Development in H/K-IFN-γ Transgenic Mice

Because SPEM is considered a preneoplastic process and H/K-IFN-γ mice developed prominent and persistent SPEM, we were interested in assessing whether H/K-IFN-γ mice were at increased risk for developing gastric cancer. In addition to metaplasia and other histological changes quantified in Figure 3C, H/K-IFN-γ mice developed dysplasia, cystic gland dilatation with occasional herniation into the submucosa, and gastric gland atrophy (Figure 9, A–F). Dysplasia was common in H/K-IFN-γ mice: at 3 to 5 months of age, 46% of H/K-IFN-γ⁹⁴⁴ and 50% of H/K-IFN-γ⁵³ mice exhibited areas of dysplasia; in mice over 12 months of age, 65% of H/K-IFN-γ⁹⁴⁴ and 44% of H/K-IFN-γ⁵³ mice had regions of dysplasia (Figure 9F). In contrast, dysplasia was not detected in any of the control mice at similar ages (Figure 9F). The similar levels of dysplasia in H/K-IFN-γ⁹⁴⁴ and H/K-IFN-γ⁵³ mice suggests that even the lower-expressing strain 53 produces sufficient IFN-γ to drive this preneoplastic change. This concept is supported by the comparable expression levels of IFN-γ target genes, as well as transcripts encoding multiple cytokines, in H/K-IFN-γ⁹⁴⁴ and H/K-IFN-γ⁵³ mice (Figures 1 and 7).

Four of 39 (10%) H/K-IFN-γ mice over 1 year of age developed either polyps or more advanced lesions, and, interestingly, all of these were located in the antrum (Figure 9, G, H, J, and K). One H/K-IFN-γ⁹⁴⁴ mouse developed an antral adenoma at 18 months of age (Figure 9, G and H); two H/K-IFN-γ⁵³ mice developed antral polyps, one at 14 months and another at 21 months of age; and one H/K-IFN-γ⁵³ mouse developed an antral adenocarcinoma (Figure 9, J and K) at 21 months of age. This tumor was characterized by marked dysplasia, cellular atypia, and a high mitotic index. In addition, this tumor contained cells with β-catenin localized prominently to nuclei (see Supplemental Figure S7 at http://ajp.amjpathol.org), consistent with activated canonical Wnt/β-catenin signaling, which has previously been linked to gastric cancer development.^31,32 Despite the appearance of advanced lesions in transgenic but not control mice, this difference was not statistically significant (Fisher's exact test, P = 0.32).

The development of antral changes was nonetheless surprising, as IFN-γ expression in our mice was targeted to the gastric corpus. We therefore examined several parameters to look for any correlation between alterations in the corpus and the antrum. Interestingly, we found a moderate correlation for hyperplasia (r = 0.45; P = 0.0003) but no correlation for inflammation (r = 0.12; P = 0.351) or dysplasia (r = 0.09; P = 0.474) (see Supplemental Figure S8 at http://ajp.amjpathol.org). Additional studies will be required to gain insight into the mechanism underlying antral hyperplasia in this setting and its long-term consequences.