Stomach-Specific Activation of Oncogenic KRAS and STAT3-Dependent Inflammation Cooperatively Promote Gastric Tumorigenesis in a Preclinical Model

Despite a recent decline in mortality and incidence, gastric cancer still accounts for one of the largest cancer-related mortalities (1). The most prevalent form of gastric cancer is intestinal-type gastric adenocarcinoma, which progresses from superficial gastritis through stages of chronic gastritis, atrophic gastritis, metaplastic transformation, and dysplasia to invasive adenocarcinoma (2). Although chronic infection with Helicobacter pylori (H. pylori) remains the main risk factor for intestinal-type gastric cancer, others include high-salt diet and viral infections (2).

Tumor initiation and progression is a multistep process that requires the acquisition of genetic alterations and growth-promoting conditions in the microenvironment. The most common activating mutations in gastric cancer driver genes affect KRAS, PI3K, and the receptor tyrosine kinases FGFR2, ERBB2, EGFR, MET, and related signaling molecules (3). Meanwhile, chronic inflammation of the gastric mucosa, triggered by bacterial infection or other environmental factors, can initiate premalignant metaplastic transformation, which often results in the loss of acid-producing parietal cells (4). Accordingly, the excessive inflammatory response of the gastric epithelium of gp130^F/F mice, a validated preclinical model for early stage intestinal-type gastric cancer in humans, is associated with intestinal metaplasia and inevitably triggers adenoma formation in the most distal part of the stomach, referred to as antrum (5). Meanwhile, aberrant activation of Kras induces hyperplasia in many epithelial tissues and, in the gastric epithelium, is associated with some hallmarks of spasmolytic peptide–expressing metaplasia (SPEM) and pseudointestinal metaplasia (6). However, oncogenic Kras activation in isolation is usually insufficient to trigger the formation of gastric adenomas (7, 8). Thus, akin to the paradigms for the intestinal epithelium, gastric tumorigenesis may also depend on the sequential acquisition of mutations within the precancerous metaplastic epithelium or require a permissive tumor microenvironment characterized by subclinical inflammation.

The formation of solid tumors is widely believed to depend on cancer stem cells, which, at least in the case of the intestinal epithelium, express similar markers as their nontransformed tissue stem cells. Accordingly, the gastrointestinal stem cell marker locus Lgr5 can confer Cre recombinase–mediated cellular transformation at the base of the glands of the pylorus and corpus region of the stomach (9). However, the higher abundance of Lgr5-positive stem cells in the intestinal epithelium results in lethal intestinal tumorigenesis, preventing the analysis of gastric tumor formation in corresponding Lgr5^CreERT2;Apc^fl/fl mice, which harbor conditional null alleles of the tumor suppressor gene Apc (9). Meanwhile, Cre expression that is largely confined to committed gastric epithelial progenitors in Atp4b-Cre or Tg(Tff2-CreERT2) mice is insufficient to reproducibly and reliably transduce tumorigenesis, and neither of these Cre-drivers have been tested for their capacity to confer recombinase activity within established tumors (10, 11). We therefore exploited the regulatory sequences of the stomach-specific trefoil factor 1 (Tff1) gene, which is primarily expressed in the epithelium of the glandular stomach within the corpus and antrum, and to a lesser extent also in the neoplastic counterpart in gastric cancer (12, 13). Importantly, Tff1 deficiency can give rise to gastric tumorigenesis in mice through a cell-intrinsic mechanism (14), arguing that Tff1-expressing gastric epithelium can provide the cell of origin for gastric cancer. We therefore exploited bacterial artificial chromosome (BAC) transgenesis to confer tamoxifen-inducible CreERT2 activity in the glandular stomach of corresponding transgenic Tg(Tff1-CreERT2) mice. This confers long-term retention of conditional β-galactosidase (β-Gal) or yellow fluorescent protein (YFP) reporter activity, or of somatically mutated genes, in the gastric epithelium and in corresponding tumor cells.

Homozygous mice for a mutated version of the IL6 signal transducer gp130, encoded by the Il6st gene, and referred to as gp130^Y757F (gp130^F/F), or for a floxed Stat3 allele (Stat3^fl/fl), as well as heterozygous Kras^LSL-G12D/+, Braf^LSL-V600E/+; LSL-LacZ13 (LacZ), and Rosa26-LSL-YFP (YFP) mice were propagated on a mixed C57BL/6 × 129/Sv background (15–20). Mice were cohoused under specific pathogen-free conditions, and age- and gender-matched littermates were used for experiments. Animal studies were approved by the Animal Ethics Committees of the Ludwig Institute for Cancer Research Institute, the Walter and Eliza Hall Institute (Melbourne, Australia), and the Olivia Newton-John Cancer Research Institute (Heidelberg, Australia).

To generate the Tff1-CreERT2 BAC, we used “DNA recombineering” to insert the CreERT2 open reading frame with a polyadenylation signal immediately downstream of the endogenous ATG initiation codon in the Tff1 locus. We injected the resulting BAC transgene in the pronucleus of CBB6F1-derived one-cell embryos and genotyped transgenic founder mice and their offspring by PCR (refer to Supplementary Methods).

Tamoxifen (Sigma T5648) was dissolved in 100% ethanol (Sigma) and diluted 1:10 in sunflower oil (Sigma). Mice were injected intraperitoneally (100 μL) for 5 consecutive days with tamoxifen (1 mg/20 g body weight; twice daily) or vehicle, unless indicated otherwise. The tamoxifen-containing emulsions were prepared freshly every 3 days.

Stomachs and intestines were opened longitudinally, washed three times in PBS by vigorous shaking, and pinned out on silicone-coated plates. Tissues were prefixed in 4% paraformaldehyde (PFA; Sigma) for one hour at 4°C, washed three times with β-Gal Wash Buffer (20 minutes/room temperature), and incubated in β-Gal Staining Solution (refer to Supplementary Methods) at 37°C overnight. Photographs were taken and the organs were fixed in 2% PFA (Sigma) at 4°C overnight, transferred to 70% ethanol, paraffin sectioned, and counterstained with Nuclear Fast Red.

Following stomach dissection, specimens were washed in PBS, fixed in 4% PFA solution (Electron Microscopy Sciences) overnight, incubated in 30% and 50% sucrose/PBS (w/v) solution, embedded in OCT (Tissue-Tek Sakura), and snap frozen on dry ice. Frozen stomach sections (8 μm) were counterstained with DAPI before mounting. YFP (shown as green) and DAPI (shown as blue) were imaged using a confocal microscope (Zeiss LSM 780).

Following dissection, tissues were washed in PBS, fixed in 10% neutral-buffered formalin solution (pH 7.4) overnight, and processed. Sections (4 μm) were stained with hematoxylin and eosin, Alcian blue or periodic acid-Schiff. Immunohistochemical analysis was performed as described previously (5).Histopathology scoring of gastric mucosa and tumors was performed as described previously (21, 22).

Protein lysates from snap-frozen tissues were prepared using the TissueLyser II (Qiagen) and RIPA lysis buffer (Sigma), separated by SDS-PAGE, and transferred to nitrocellulose membranes by iBlot (Invitrogen). Protein bands were visualized and quantified using the Odyssey Infrared Imaging System (LI-COR Biosciences).

Unless otherwise stated, comparisons among mean values were performed by ANOVA or a two-tailed Student t test as appropriate using Prism 5 software (GraphPad). P < 0.05 was considered statistically significant.

Further information is provided in the Supplementary Methods.

To confer Cre activity to the gastric epithelium without affecting endogenous gene expression, we took advantage of BAC transgenesis and DNA recombineering. We generated a BAC vector encoding the CreERT2 recombinase inserted at the translational initiation sites of the Tff1 locus (Supplementary Fig. S1A–S1C). We derived four independent Tg(Tff1-CreERT2) founder lines and confirmed retention of the transgenic BAC after germline transmission (Supplementary Fig. S1D).

We first determined CreERT2 transgene activity through the assessment of β-Gal or YFP activity conferred by activation of the conditional LacZ or YFP reporter allele in the corresponding Tg(Tff1-CreERT2) compound mutant strain. Six hours after a single tamoxifen administration (1 mg/20 g body weight), we detected β-Gal or YFP-positive cells in the pit regions of gastric glands in the corpus and antrum and also often along entire antral glands (Fig. 1A–C and Supplementary Fig. S1E). These staining patterns suggest CreERT2 expression within all cells of the antral glands but only some cells in the corpus, thereby reflecting the distribution of endogenous Tff1 protein (Supplementary Fig. S2A). Likewise, the absence of reporter activity in the forestomach, and the gradually increased expression of the Tff1-CreERT2 transgene from the squamous forestomach to the antrum, with a sharp decrease towards the proximal small intestine, coincided with a similar expression pattern of endogenous Tff1 (Supplementary Fig. S2B; refs.12, 13). We also observed β-Gal–positive cells towards the base of antral glands, where Lgr5-positive stem cells reside (9). Indeed, FACS-sorted GFP-positive cells from stomachs of Lgr5^{CreERT2-IRES-GFP} mice also expressed Tff1, although we found stronger Tff1 expression in GFP-negative stomach epithelial cells (Supplementary Fig. S2C; ref. 2). Consistent with expression of the Tff1-CreERT2 transgene in long-lived stem cells, approximately 20% of antral glands stained entirely β-Gal–positive 120 days after tamoxifen administration, with another 40% staining partially positive for the reporter (Fig. 1B). However, we did not detect completely β-Gal–traced glands in the corpus, suggesting that the Troy-positive stem cells at the base of the corpus, which can reconstitute entire corpus glands (23), do not express the Tff1-CreERT2 transgene.

To induce more widespread and uniform CreERT2 activity, we administered tamoxifen (1 mg/20 g mouse, twice daily) over 5 consecutive days. Three days later, we again observed >20% of entirely β-Gal–traced glands in the antrum, but not in the corpus (Fig. 2A and B). Over a 360-day follow-up period, the frequency of entirely stained glands increased slightly in the antrum, but was never observed in the corpus, where only 20% of all oxyntic glands showed some β-Gal staining. A transient increase of entirely stained glands 10 days after tamoxifen administration is consistent with the mucosal repair elicited by the transient loss of parietal cell associated with high doses of tamoxifen administration (i.e., >3 mg/20 g body weight), and which spontaneously resolves within three weeks (24). In contrast, stomachs of vehicle-treated Tff1-CreERT2–positive or tamoxifen-treated Tff1-CreERT2–negative littermates showed no β-Gal staining (Supplementary Fig. S3A). We also observed β-Gal and YFP (data not shown) staining in the Brunner glands and in some airway epithelial cells (Supplementary Fig. S3B), but not in the pancreas (Supplementary Fig. S3C and S3D). As all four transgenic founder lines yielded similar reporter staining patterns, we concluded that the Tff1-CreERT2 transgene replicates the expression of endogenous Tff1 (25, 26), and confers recombinase activity to long-lived stem/progenitor cells, particularly within the antrum.

To explore whether the oncogenic Kras^G12D mutation in epithelial stem/progenitor cells is sufficient to initiate tumorigenesis in the gastric antrum, we treated adult Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice with tamoxifen. Nine months later, we detected large adenomas in one third of the mice analyzed (Fig. 3A; Supplementary Table S1), whereas stomachs of tamoxifen-treated Tg(Tff1-CreERT2)-negative littermates remained histologically normal and lacked evidence for the recombined Kras^G12D allele (Fig. 3A and B and Supplementary Fig. 4A). Kras^G12D-induced tumors were characterized by glandular structures at the luminal edges and regions with intestinal goblet-like cells (Fig. 3C) that stained positive for acidic mucus and neutral glyco and mucoproteins (Supplementary Fig. S4B). The tumor-adjacent antral mucosa showed similar signs of mucus metaplasia, coinciding with epithelial hyperplasia and mononuclear cell infiltrates in the lamina propria (Fig. 3C and Supplementary Fig. 4C). In contrast, the corpus revealed only mild inflammation and occasionally dilated and disorganized glands, which therefore clearly demarcated the corpus–antrum border in Kras^G12D-expressing mice (Fig. 3C).

As expected from the widespread recombination of the latent Kras^LSL-G12D allele (Supplementary Fig. S4A), we observed staining of the phosphorylated (activated) Erk1/2 isoforms at the luminal edges of Kras^G12D-induced tumors and more prominently in the adjacent hyperplastic antrum, irrespective of the presence of tumors (Fig. 4A and B). Surprisingly, all tumors in Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice displayed extensive staining of the activated, tyrosine-phosphorylated form of Stat3 (Fig. 4A and B), which mediates the tumor-promoting activity of the inflamed microenvironment (27, 28). Collectively, these results suggest that oncogenic Kras^G12D triggers excessive Mek/Erk pathway activation, which, in a tumor-specific manner, coincides with aberrant Stat3 activation, as indicated by excessive staining for phosphotyrosine (pY)-Stat3.

We detected widespread expression of the SPEM marker Tff2 (27, 29) in all Kras^G12D-induced adenomas (Fig. 4C) and at the base of the adjacent antral mucosa (Supplementary Fig. S4D). This coincided with Alcian blue staining in cells of the deep antral glands (Supplementary Fig. S4B). We also detected signs of pseudointestinal metaplasia within the tumor-adjacent glandular epithelium and in the adenomas, comprising prominent appearance of goblet-like cells alongside the induction of the intestinal marker gpA33 (30, 31) and to a lesser extent of Cdx2 (Fig. 4C and D and Supplementary Fig. S4E and S4F). Consistent with the discordant expression between gpA33 and the intestinal master regulator Cdx2, with the latter being expressed at the base of the glands (31, 32), ectopic expression of the intestinal marker Cdx2 was limited to some epithelial cells at the base of antral glands in tumor-bearing mice (Supplementary Fig. S4E). Meanwhile, we detected the intestinal mucin Muc2 only in a few tumor cells (Fig. 4C), albeit reminiscent of MUC2 expression in humans being most prominent in gastric cancer tissues (33). Thus, SPEM and pseudointestinal metaplasia appear in transformed hyperplastic epithelium and tumors irrespective of the mucosal cell type(s) that express Kras^G12D in response to Tff1-CreERT2 activation or transgenic expression from the K19 promoter (6).

To establish whether the observations associated with Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice also occurred with other mutations within the Ras–Erk pathway, we generated Tg(Tff1-CreERT2);Braf^LSL-V600E/+ compound mice to mimic a rare mutation in human gastric cancer (34). We observed large adenomas in the antrum of 10/15 Tg(Tff1-CreERT2);Braf^LSL-V600E/+ mice 8 months after tamoxifen administration (Supplementary Fig. S5A; Supplementary Table S1). Strikingly, these lesions replicated the histopathologic features observed in Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice, including disorganized glandular structures, SPEM, and regions of pseudointestinal metaplasia with goblet-like cells that stained positive for acidic mucus (Supplementary Fig. S5A–S5C). Predictably, the antrum of all Braf^V600E-expressing mice showed elevated Mek1/2 and Erk1/2 phosphorylation, irrespective of the presence of tumors, and all tumors also displayed extensive staining of pY-Stat3 (Supplementary Fig. S5D and S5E).

Homozygous gp130^F/F mice, which carry a constitutive tyrosine to phenylalanine substitution mutation at amino acid 757 to prevent binding of the negative regulator Socs3, as well as activation of the Ras signalling cascade, spontaneously develop gastric adenomas in response to excessive IL11-dependent Stat3 activation with molecular hallmarks of intestinal-type gastric cancer (5, 35). Although Tff1 expression is reduced in tumors of gp130^F/F mice (Supplementary Fig. S6A; ref.15), we observed extensive β-Gal staining throughout the tumor epithelium of gp130^F/F;Tg(Tff1-CreERT2);LacZ mice and detected robust, long-term retention of β-Gal (Supplementary Fig. S6B). Although the latter was concentrated towards the tumor edges, immunohistochemical analysis for β-Gal revealed staining of most tumor glands (Supplementary Fig. S6C). We therefore hypothesized that the Tff1-CreERT2 transgene–mediated genetic ablation of tumor promoters could counteract their effect, and treated tumor-bearing gp130^F/F;Tg(Tff1-CreERT2);Stat3^fl/fl and gp130^F/F;Stat3^fl/fl control mice, harboring a “floxed” Stat3 allele, with tamoxifen. We observed a significant reduction in tumor burden and numbers in mice of the CreERT2-positive cohort four weeks later (Fig. 5A–C). This coincided with reduced levels of total Stat3 expression in the tumors and adjacent mucosa and a reduction of nuclear accumulation of the pY-Stat3 (Fig. 5D–F). Thus, the Tff1-CreERT2 transgene is sufficient to effectively confer “oncogene” inactivation in the gastric mucosa and established tumors.

Although tumor latency and incidence differs between gp130^F/F and Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice (Supplementary Table S1), both models are associated with hallmarks of inflammation and metaplastic transformation. Furthermore, as Stat3 activity is moderately elevated in Kras^G12D-induced gastric tumors (Fig. 4A and B), we hypothesized that further augmentation of Stat3 and other signaling molecules downstream of Kras in compound gp130^F/F;Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice may increase tumor frequency and progression of the tubular adenomas that spontaneously arise in gp130^F/F mice. Indeed, 3 months after tamoxifen administration to 8-week-old tumor-bearing gp130^F/F;Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice, we detected increased tumor numbers when compared with those from the gp130^F/F;Kras wild-type (Kras-wt) cohort, which comprised gp130^F/F;Tg(Tff1-CreERT2) and gp130^F/F;Kras^LSL-G12D/+ mice (Fig. 6A).

Surprisingly, the extent of epithelial hyperplasia and the extent of inflammatory cell infiltration were similar in tumors of all genotypes (Fig. 6B and C). Strikingly however, goblet-like cells and associated mucus metaplasia as well as dysplastic regions were consistently more prominent in tumors of gp130^F/F;Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice (Fig. 6B–E). Importantly, approximately half of all tumors analyzed from 3-month-old gp130^F/F;Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice contained clusters of submucosal glands, likely to reflect an increased propensity of the infrequent epithelial invasion through the muscularis mucosae seen in 8- to 10-month-old gp130^F/F mice (Fig. 6C and E; ref. 36).

We morphologically investigated the nontumor mucosa for quantifiable evidence of metaplasia and compared stomachs from 3-month-old CreERT2-positive gp130^F/F;Kras^LSL-G12D/+ and gp130^F/F;Kras-wt mice with those obtained from 9-month-old CreERT2-positive gp130^+/+;Kras^LSL-G12D/+ mice. Despite the different ages, we found a similar prevalence of glandular disorganization, hyperplasia, goblet-like cell metaplasia, and inflammatory cell infiltrates in the antral mucosa of gp130^+/+;Kras^LSL-G12D/+ and gp130^F/F;Kras-wt mice, which was less prominent in the corpus (Fig. 7A and B and Supplementary Fig. S7A). Most likely attributable to the gp130^F/F mutation (5, 15), the gastric mucosa of the antrum and corpus of gp130^F/F;Kras^LSL-G12D/+ mice showed a striking transformation into an intestinal phenotype associated with extensive hyperplasia, signs of inflammation, and large numbers of goblet-like cells (Fig. 7A and B and Supplementary Fig. S7A). Indeed, these mucus-producing cells had replaced most of the gastric cell types in the corpus, which showed a severe loss of parietal and chief cells (Fig. 7B and Supplementary Fig. S7B).

Collectively, our observations suggest that simultaneous excessive Stat3 activation and oncogenic Kras^G12D expression cooperate to increase tumor progression and to trigger the formation of submucosal glands and invasion, with the latter coinciding with extensive metaplastic transformation and inflammatory cell accumulation in the tumor-adjacent precancerous gastric mucosa.

Here, we describe the generation and validation of Tg(Tff1-CreERT2) mice as a novel gastric tissue–specific Cre-driver strain and conclude that the employed BAC-based transgene confers and restricts CreERT2 activity to Tff1 expressing epithelial cells along the entire length of the antral glands and some cells in the pit region of the corpus (12). As the gastric mucosa continuously and rapidly undergoes complete renewal, the abundant long-term retention of LacZ or YFP reporter activity in the antrum, tumors, and to a lesser extent, in the corpus, indicates Tff1-CreERT2 activity within the long-lived stem and progenitor cell compartment. This provides an important advantage over many existing Cre-driver strains when modeling diseases with a long latency (i.e., cancer) or in situations where stem cell loci (i.e., Lgr5) also confer recombinase activity to other compartments outside of the stomach.

High concentrations of tamoxifen (>3 mg/20 g mouse) can induce transient mucosal damage and parietal cell atrophy 3 days after the first tamoxifen administration; however, these cytotoxic side effects fully resolve within 1 to 3 weeks through increased gastric stem cell proliferation (24).Our data suggest an involvement of Tff1-expressing, possibly Lgr5-positive, stem cells in this process, because we observe a maximal number of traced antral glands 10 days after tamoxifen administration. Although high tamoxifen concentrations could also induce transient metaplastic changes (23, 24), this is less likely to interfere in situations of “long latency phenotypes” (i.e., Kras^G12D and Braf^V600E–induced tumorigenesis) than with short-term lineage tracing experiments reported by Huh and colleagues (24).

The frequency and extended latency period of tumor development in Tg(Tff1-CreERT2);Kras^LSL-G12D/+ and Tg(Tff1-CreERT2);Braf^LSL-V600E/+ mice is reminiscent of similar findings in mice with systemic Kras activation (7, 8) and suggest the need to overcome tumor-suppressing mechanisms, including Ink4a/Arf–mediated cell-cycle arrest (37). As inflammation has been shown to overcome oncogene-induced senescence (38) and to enable epithelial-to-mesenchymal transition (39), we speculated that inflammation-associated Stat3 activation may also facilitate these processes. This hypothesis is supported by our observation of extensive Stat3 activation within the tumors of Kras^G12D mice, but not in the adjacent preneoplastic epithelium. We functionally validated this correlation in gp130^F/F;Tg(Tff1-CreERT2); Kras^LSL-G12D/+ mutant mice by demonstrating tumor-promoting synergies between excessive, inflammation-driven Stat3 and oncogenic Kras^G12D signaling. Our findings complement similar observations in Kras-driven human xenografts, including pancreas cancer models, where neoplastic growth required persistent Stat3 activity (38). Thus, inflammation may not only promote metaplastic transformation of the (Kras-) mutant glandular epithelium, but indeed may overcome oncogene-induced intrinsic barriers that safeguard against aberrant proliferation.

It still remains unclear whether chronic inflammation and metaplastic transformation, which are commonly associated with human gastric cancer, are required for tumorigenesis or whether they account for epiphenomena. Likewise, whether persistent inflammation and metaplasia develop in parallel or whether one condition causes the other has not been resolved yet. Here, we find Tff2 expression in Alcian blue–positive antral glands of the Kras^G12D-mutant epithelium, at least in tumor-bearing mice, with metaplasia characterized by some intestinal hallmarks, including the presence of goblet-like cells and the expression of intestinal marker gpA33 (30, 31). Our observations also clarify the contribution of gp130-dependent excessive Stat3 activity to Kras-induced tumorigenesis. First, in the Kras^G12D-mutant epithelium, elevated Stat3 activity correlates with enhanced metaplastic transformation and associated loss of parietal and chief cells and promotes tumor development. As the gastric epithelium of tumor-bearing and tumor-free Kras^G12D or Braf^V600E mice showed goblet-like cell metaplasia, we conclude that the tumor-promoting secondary events, which are implicated by the long latency period required for tumors to emerge, are likely to occur in metaplastically transformed cells. Second, excessive Stat3 activation in Kras^G12D or Braf^V600E cells confers the potential for hyperplastic/dysplastic transformation and, when further stimulated through excessive gp130 signaling, the appearance of submucosal glands. It is therefore tempting to speculate that interference with gp130/Stat3–activating cytokines might provide a novel therapeutic target for Kras^G12D-dependent gastric tumorigenesis. Indeed, we have observed that systemic limitation of Stat3 expression in tamoxifen-induced compound Tg(Tff1-CreERT2);Braf^LSL-V600E/+;Stat3^+/− markedly reduced tumor formation (Eissmann and Ernst, unpublished data). Furthermore, the notion of inflammatory cytokines promoting Kras-initiated gastric tumors (39) is consistent with observations by Okumura and colleagues (6) and fits with emerging findings that inflammation downregulates mismatch repair proteins and the clinical observations that KRAS mutations occur more frequently in gastric cancers with DNA mismatch repair deficiency (40).

Although various animal models have been employed to identify molecular mechanisms involved in gastric tumorigenesis, there is a need for mouse models that allow for the consecutive induction of mutations to mimic advanced human disease. Here, we replicate in gp130^F/F;Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice, as a proof of concept, the sequential genetic changes that collectively facilitate progression from tubular adenomas to invasive adenocarcinomas. To our knowledge, gp130^F/F;Tg(Tff1-CreERT2);Kras^LSL-G12D/+ mice are the first conditional gastric tumor model in which two oncogenic pathways can be activated independently and alongside the CEA/SV40-T, the Tff1 knockout and the double-transgenic K19-C2mE (GAN) strains, one of a few models that reproducibly show submucosal invasion (41–43). Moreover, as we show here functionally for Stat3 gene expression, the extensive Tff1-CreERT2 transgene expression throughout the proliferative tumor epithelium provides a novel Cre-driver strain to assess the contribution of putative oncogenes to gastric tumor progression. Furthermore, the widespread Cre expression throughout the glandular stem cell compartment of Tg(Tff1-CreERT2) mice enables functional testing of the contribution of candidate genes from human genome-wide association studies to gastric cancer initiation and progression.

No potential conflicts of interest were disclosed.

Conception and design: S. Thiem, M.F. Eissmann, T.L. Putoczki, M. Ernst

Development of methodology: S. Thiem, M.F. Eissmann, J. Elzer, T.L. Putoczki

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Thiem, M.F. Eissmann, J. Elzer, A. Jonas, T.L. Putoczki, A. Poh, A. Preaudet, D. Flanagan, P. Waring, M. Buchert

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Thiem, M.F. Eissmann, T.L. Putoczki, M. Buchert, A. Jarnicki, M. Ernst

Writing, review, and/or revision of the manuscript: S. Thiem, M.F. Eissmann, M. Buchert, M. Ernst

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Thiem, A. Jonas, P. Nguyen, D. Flanagan, E. Vincan

Study supervision: T.L. Putoczki, M. Ernst

Other (obtained funding): M. Ernst

The authors thank the LICR and WEHI Histology and Animal Facility Staff for expert technical assistance. The authors also thank Sue Bath for the pronuclear microinjections and Joan Heath for helpful discussions.

This work was supported in part by a Project Grant (APP1007523) and a Program Grant (APP487922) from the NHMRC and by funds from the Operational Infrastructure Support Program provided by the Victorian Government, Australia (M. Ernst). This work was also initiated through the generous financial support of Ludwig Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.