YAP/TAZ Initiates Gastric Tumorigenesis via Upregulation of MYC

Gastric cancer is the fifth most common type of cancer worldwide and ranks as the third most frequent cause of cancer-related deaths (1). Despite numerous studies focusing on the genetic alterations associated with gastric cancer, our knowledge is limited about the signaling pathways that drive malignant transformation (2–6). The Hippo signaling pathway (also known as Hippo-YAP/TAZ signaling) is a growth control and tumor suppressor pathway associated with carcinogenesis, regeneration, and metabolism (7–11). In this pathway, the core kinases MST1/2 and LATS1/2 negatively regulate the effector molecules YAP and TAZ. The scaffolding proteins SAV1 and MOB1A/B bind to MST1/2 and LATS1/2, respectively, to modulate their kinase activities. When the Hippo pathway is activated, LATS1/2 phosphorylates YAP/TAZ, inhibiting their activity. Conversely, when the pathway is deactivated, YAP/TAZ accumulates in the nucleus, driving gene expression by forming complexes with transcription factors (most notably with TEAD). This effect leads to cell proliferation and inhibition of apoptosis.

Several studies have suggested a potential link between YAP/TAZ signaling and gastric cancer (12–15). YAP expression is upregulated in gastric tumor samples compared with normal tissue, and the total amount of YAP expression as well as its nuclear localization are significantly correlated with poorer clinical outcomes (12–14). Moreover, a peptide mimicking the role of VGLL4 (a candidate for inhibiting the YAP–TEAD interaction) inhibits tumor growth in both xenograft and carcinogen-induced murine gastric tumor models (15). However, the clear significance of Hippo-YAP/TAZ signaling in gastric cancer has never been established in human genomic/transcriptomic studies with a large number of samples (2–6). Most important, the lack of a genetically engineered mouse model has made it difficult to clarify whether the relationship between the upregulation of YAP/TAZ and carcinogenesis is direct and causal. The fundamental question of whether YAP/TAZ activation is the cause or consequence of gastric cancer in vivo thus remains to be answered.

Anatomically, the murine stomach can be roughly subclassified into three parts. The proximal part is the forestomach, which is covered with squamous epithelium. The distal part is the glandular stomach, which can be subdivided into corpus and pylorus (antrum). Surface mucous epithelial cells in the gastric mucosa are columnar, and epithelial cells in the glands are cuboidal, but their morphologic details and cellular makeup differ (16, 17). Various markers are available for stem cell populations in the corpus and pylorus, and some of them overlap. Stem cells in the corpus express Sox2, Troy, Mist1, and eR1, whereas stem cells in the pyloric antrum express Sox2, Lgr5, CCK2R, eR1, Axin2, and Mist1 (18–25). Mouse model studies have shown that oncogenic events in gastric stem cells can frequently lead to carcinogenesis, although mutations targeted to chief cells also can cause these cells to dedifferentiate and acquire the potential to grow and regenerate gastric glands, sometimes forming metaplasia (17, 20, 21, 22, 26–29).

In this study, we asked whether pyloric stem cells expressing Lgr5 may serve as the cells of origin for cancers of the distal stomach that exhibit YAP/TAZ activation (22). For this purpose, we used conditional knockouts of Lats1 and Lats2 to activate YAP/TAZ and showed that this perturbation was sufficient to trigger dysplastic changes and eventually tumorigenesis in the pyloric epithelium. We performed transcriptome analysis to scrutinize the underlying molecular mechanisms and uncovered MYC as a critical target of YAP in the tumorigenic pathway. We corroborated the significance of MYC by showing that its pharmacologic inhibition could rescue YAP-induced oncogenic phenotypes. Using human gastric cancer tissue microarray samples, we confirmed a positive correlation of YAP and MYC. Finally, by analyzing public microarray data, we show that YAP signature genes are highly enriched in a subset of gastric tumors, supporting its importance in carcinogenesis of human gastric cancer.

All mouse experiments were performed under approval by the Institutional Animal Care and Use Committee of Korea Advanced Institute of Science and Technology. Lats1^fl/fl and Lats2^fl/fl mice were described previously (30–32). Lgr5-EGFP-IRES-creERT2 mice were kindly provided by Dr. Young-Yun Kong (Seoul National University) and described previously (33). R26-tdTomato reporter mice were purchased from Jackson laboratory, and also described previously (34). Inducible Cre-mediated gene deletion was initiated by single dose intraperitoneal injections into 8- to 10-week-old mice of 75 mg tamoxifen/kg body weight. Tamoxifen (Cayman Chemical) was dissolved in corn oil (Sigma) at a concentration of 20 mg/mL.

Mouse stomach was harvested and cut open through the greater curvature, fixed with 10% neutral-buffered formaldehyde at 4°C overnight, and embedded in paraffin. Sections (3 μm) were cut and stained with hematoxylin and eosin (H&E). For IHC staining, slides were deparaffinized, quenched for endogenous peroxidases with 3% hydrogen peroxide, and subjected to antigen retrieval in citrate buffer (10 mmol/L sodium citrate, 0.05% Tween 20, pH 6.0) using a pressure cooker. Blocking was performed with 3% BSA and 0.3% Triton X-100 in PBS for 1 hour at room temperature. The primary antibodies used were anti-MYC (Millipore, 06-340, 1:100), and anti-Hes1 [gift from Dr. Ryoichiro Kageyama (Kyoto University, Kyoto, Japan), rabbit polyclonal, 1:500]. Primary antibody detection was performed using goat anti-rabbit horseradish peroxidase (Jackson ImmunoResearch, catalog no. 111-035-003, 1:500) and the Dako EnVision System (catalog no. K5007). Immunofluorescence staining was performed with anti-YAP (Cell Signaling Technology, 4912s, 1:200), anti-TAZ (Sigma, HPA007415, 1:200), anti-Ki67 (Abcam, ab16667, 1:100), anti-RFP (Rockland Immunochemicals, RK600-401-379, 1:200), anti-MUC2 (Abcam, ab90007, 1:150), and anti-β-catenin (BD Biosciences, 610154, 1:100). Goat anti-rabbit Alexa Fluor 594 (Thermo Fisher Scientific, A-11037, 1:1,000), donkey anti-rabbit Alexa Fluor 488 (Thermo Fisher Scientific, A-21206, 1:1,000), or donkey anti-mouse Alexa Fluor 488 (Thermo Fisher Scientific, A-21202, 1:1,000) was used to detect fluorescent signals. Images were obtained with a Leica DMLB microscope (Leica) or confocal microscope (Zeiss, LSM 780, and LSM 800). Alcian blue staining was performed using Alcian Blue (pH 2.5) Stain Kit (Vector Laboratories, H-3501). Staining intensities were calculated using ImageJ software. Mouse histology slides for H&E staining were confirmed by a pathologist (C. Choi).

Cells were trypsinized and 1,000 cells per well in 100 μL of DMEM with 10% FBS were seeded in 96-well ultralow attachment plates (Corning, 3474). For drug treatment, JQ1 (Cayman Chemical, final concentration 500 nmol/L), 10058-F4 (Sigma, final concentration 50 μmol/L) or vehicle (DMSO, MP Biomedicals) was added to the media. Colonies were counted 5–7 days later. Quantification of cell viability was measured with a WST-1 assay (LPS solution, CYT3000) according to the manufacturer's protocol.

Vehicle for injection was prepared with 10% cyclodextrin (Sigma, H-107) in distilled water. JQ1 was first dissolved in DMSO at a concentration of 50 mg/mL, and then diluted 1:10 with vehicle to a final concentration of 5 mg/mL. Mice were intraperitoneally injected daily with JQ1 at 50 mg/kg bodyweight for 5 weeks.

Microarray data from the Asian Cancer Research Group cohort were downloaded from NCBI Gene Expression Omnibus (GSE62254). Gene set enrichment analysis (GSEA) was performed using C6 Oncogenic Signatures from the Molecular Signatures Database (MSigDB) version 5.1 (35). The microarray samples were reclassified based on YAP signature gene expression using the Signature Evaluation Tool (36). Survival analysis was performed using the log rank test and Kaplan–Meier method via GraphPad Prism.

Gastric adenocarcinoma tissue microarrays were obtained from Chonnam National University Hwasun Hospital National Biobank of Korea. The patients from whom the samples were derived underwent surgical tumor resections between 2009 and 2012 at Chonnam National University Hwasun Hospital. All samples were obtained with informed consent and approved by the Institutional Review Board of Chonnam National University Hwasun Hospital (CNUHH-2017-086).

All statistical analyses, including two-tailed t tests and log-rank tests for survival analyses, were performed using GraphPad Prism software (version 7). Analyses were performed with two-tailed t tests, unless otherwise indicated.

The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE104823 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE104823).

To activate YAP/TAZ, we used conditional knockout mice for Lats1 and Lats2 because LATS kinases are direct negative upstream regulators of YAP/TAZ. We mated Lats1 and Lats2-floxed mice with Lgr5-EGFP-IRES-creERT2 to generate Lats1^fl/fl;Lats2^fl/fl;Lgr5-CreER mice, designated as Lats1/2^iΔLgr5 mice, for activating YAP/TAZ specifically in Lgr5⁺ epithelial stem cells in a tamoxifen-inducible manner (Supplementary Fig. S1A). These mice were born in Mendelian ratios and showed no noticeable defects unless injected with tamoxifen for Cre induction. Tamoxifen was intraperitoneally injected into Lats1/2^iΔLgr5 mice at 8–10 weeks of age to induce gene deletion, consequently activating YAP/TAZ. Lats1^fl/fl;Lats2^fl/fl littermates of Lats1/2^iΔLgr5 mice (referred to here as wild type; WT) were used as control and injected with identical tamoxifen doses at the same frequencies. By tracing Cre expression with reporter mice (Lgr5-CreER;tdTomato) and performing recombination PCR on the pyloric epithelial tissue, we confirmed efficient Cre-induced gene deletion upon tamoxifen injection in Lats1/2^iΔLgr5 mice (Supplementary Fig. S1B and S1C).

Lgr5 marks adult stem cells in various tissues, so we harvested all organs known to contain Lgr5⁺ stem cells (i.e., stomach, small intestine, colon, skin, liver, pancreas, and ovary; refs. 22, 33, 37–40). To our surprise, we observed Lats1 and Lats2 knockout phenotypes only in the pyloric stomach and skin (hair follicular hyperplasia). Unexpectedly, the small intestine and colon, the best known locations for Lgr5⁺ stem cell populations, appeared to be histologically normal (Supplementary Fig. S1D). The reason for this result is unclear, but to investigate the role of YAP/TAZ in gastric cancer, we focused only on the stomach in this study.

To analyze tumorigenesis in the stomach, we sacrificed mice at sequential time points after tamoxifen induction (Supplementary Fig. S1A). At gross examination, there was no significant mass protrusion on the gastric epithelial surface, but there were diffuse flat elevations in the pyloric region (Supplementary Fig. S1E). In histology sections, we found evidence of microscopic changes in the pyloric epithelium in Lats1/2^iΔLgr5 mice beginning at 8 weeks postinduction (Fig. 1A). These changes include slight tubule distortion and mild nuclear size variation, indicating hyperplasia. Dysplastic changes appeared 12 weeks postinduction with some nuclei showing a slight loss of polarity and mild cytologic atypia (low-grade intraepithelial neoplasia). At 16 weeks, these lesions became more severe, characterized by crowding of tubules lined by atypical cells (high-grade intraepithelial neoplasia). At 20 and 24 weeks, cells in the lesions displayed hyperchromatic nuclei and loss of polarity with glands budding and invading into the lamina propria (intramucosal invasive carcinoma/intramucosal carcinoma; Fig. 1A and B). The intense Ki67 staining of these lesions suggested that they were highly proliferative (Supplementary Fig. S2A). Pyloric stomach of WT mice exhibited weak YAP expression in the antral glands, while TAZ expression was restricted in the isthmus region and lost at the base of glands where Lgr5⁺ cells reside (Supplementary Fig. S2B). However, Lats1/2^iΔLgr5 mice exhibited intense nuclear immunostaining of YAP/TAZ throughout the period of analysis, reflecting elevated activity of these signaling effectors (Fig. 2A; Supplementary Fig. S2C). Lats1/2^iΔLgr5 mice also showed strong increases in staining for markers of intestinal metaplasia (MUC2 and Alcian blue), indicating that metaplasia preceded dysplastic changes (Supplementary Fig. S3A). These lesions appeared at very high penetrance (Supplementary Table S1A), confirming the robustness of our YAP/TAZ–induced gastric tumor model.

To confirm that these tumors originated from gastric stem cells lacking Lats1 and Lats2, we bred Lats1/2^iΔLgr5 mice with R26-tdTomato reporter mice to generate Lats1^fl/fl;Lats2^fl/fl;Lgr5-CreER;R26-tdTomato (Lats1/2^iΔLgr5;tdTomato) mice. Gastric lesions formed after Cre induction in these mice were positive for tdTomato, indicating that they were derived from Lats1/2-deleted stem cells (Fig. 2B). As pyloric Lgr5⁺ cells are reportedly almost quiescent, we asked whether Lgr5⁺ cell proliferation changes early after Lats knockout (22, 23). When we compared Lgr5-CreER;tdTomato mice with Lats1/2^iΔLgr5;tdTomato mice 1 week post-tamoxifen induction, we noticed that while there was no difference in the number of Lgr5⁺cells, loss of Lats markedly increases Lgr5⁺ cell proliferation as tracked by tdTomato expression (Supplementary Fig. S3B). Also, whereas most Lgr5⁺ cells in Lgr5-CreER;tdTomato mice were quiescent, Lats-deficient Lgr5⁺ cells showed significantly more Ki67 staining, implying loss of Lats triggers more rapid cell cycling in these cells by enhancing YAP and TAZ (Supplementary Fig. S3B and S3C). Together, these results indicated that YAP/TAZ activation by Lats1/2 deletion in pyloric stem cells triggered stepwise progression from low-grade intraepithelial neoplasia to intramucosal carcinoma in the normal gastric epithelium, validating this signaling cascade as a critical driver of gastric tumorigenesis in vivo.

To verify the oncogenic role of YAP/TAZ in vitro, we asked whether activation of YAP/TAZ could promote cellular transformation and confer the hallmark features of cancer on a gastric epithelial cell line (HFE-145). YAP-5SA and TAZ-4SA harbor serine-to-alanine mutations in LATS phosphorylation sites that result in phenotypes similar to those caused by hyperactive mutant forms of YAP and TAZ (41, 42). To recapitulate our in vivo results, we overexpressed YAP-5SA and TAZ-4SA in HFE-145 cells. As an index for cellular transformation, we asked whether these active mutants promote colony formation in an anchorage-independent condition. In a soft agar assay, both YAP-5SA and TAZ-4SA overexpression led to 3.25-fold and 1.74-fold increases in colony size, while only YAP-5SA overexpression led to a 5.42-fold increase in the number of colonies (Fig. 2C; Supplementary Fig. S3D). The differential effect of YAP and TAZ on colony formation suggested a more active role of YAP in tumorigenesis.

We next asked whether active YAP/TAZ promotes invasion through the extracellular matrix, another hallmark of malignant cells. In a transwell invasion assay, YAP-5SA and TAZ-4SA overexpression enhanced cellular invasion through the extracellular matrix by 3.58- and 1.89-fold, respectively (Fig. 2D; Supplementary Fig. S3E). Collectively, these in vitro data, combined with the in vivo findings, indicated that YAP/TAZ activation promotes malignant transformation of gastric epithelial cells.

Because YAP and TAZ are transcriptional coactivators, we next attempted to identify the downstream targets responsible for relaying tumorigenic signals. For this purpose, we performed RNA sequencing on gastric lesions formed 8 weeks after tamoxifen induction in our model mice. We first harvested the epithelial lining of the stomach and verified the activation of YAP/TAZ by quantifying expression levels of the known downstream target genes Ctgf, Cyr61, and Ankrd1 (Supplementary Fig. S4A). By analyzing and comparing transcriptome data from the gastric tissues of Lats1/2^iΔLgr5 and WT mice, we identified almost 2700 differentially expressed genes (i.e., 1158 upregulated genes and 1543 downregulated genes in the Lats1/2^iΔLgr5 mice compared with WT mice). A subsequent gene ontology analysis of biological process terms revealed significant enrichment of genes related to cell-cycle progression (Fig. 3A).

We also analyzed the differentially expressed genes using GSEA to identify the signaling pathways downstream of the oncogenic cascade. First, we found significant enrichment of YAP signature gene sets in the Lats1/2^iΔLgr5 group, validating our sequencing results (Fig. 3B). Consistent with our gene ontology analysis, we found that E2F targets, G2M checkpoint-related genes, and MYC target genes, all molecular signatures related to cell-cycle progression, topped the GSEA list (Fig. 3C and D; Supplementary Fig. S4B). Epithelial-to-mesenchymal transition (EMT)-related gene sets were also highly upregulated in the tumors, suggesting a link between this process and YAP/TAZ signaling (Supplementary Fig. S4B and S4C).

Among the enriched gene sets, we focused on MYC target genes. MYC is widely known for its role as an oncogene, promoting cell-cycle progression and regulating the expression of many important genes critical for cell growth and proliferation (43). It also has been suggested as a downstream target of YAP (44, 45). In an effort to extend our own observations and the results in these previous reports, we asked whether MYC may act as a critical downstream target relaying YAP/TAZ signals in gastric tumorigenesis.

First, we confirmed whether or not MYC was active in our model mice. Using IHC with mice at 8 weeks post-tamoxifen induction, we found that pylorus of Lats1/2^iΔLgr5 mice were mostly covered with intense nuclear MYC expression, while MYC expression in WT mice was limited to the basal layer (Fig. 4A). Because YAP reportedly regulates Wnt/β-catenin and Notch signaling, and because MYC acts downstream of these pathways, it is possible that the upregulation of MYC we observed is indirect (46, 47). In our RNA sequencing results comparing WT and Lats1/2^iΔLgr5 mice, however, we did not observe any significant enrichment of Wnt/β-catenin or Notch target gene sets. (Supplementary Fig. S5A). Moreover, we could not detect any significant difference in the staining of β-catenin and Hes1 in the pyloric epithelium of WT and Lats1/2^iΔLgr5 mice (Supplementary Fig. S5B).

To further confirm that YAP/TAZ activation upregulates MYC, we examined MYC expression in the YAP-5SA– and TAZ-4SA–overexpressed HFE-145 cells. As expected, YAP-5SA and TAZ-4SA overexpression increased MYC protein levels (Fig. 4B, left), but to our surprise, MYC mRNA levels remained unaltered (Fig. 4B, right).

To determine whether this regulation occurs at the posttranscriptional level, we overexpressed YAP-5SA/S94A, which harbors another serine-to-alanine mutant preventing YAP binding to the TEAD transcription factor (48). YAP5SA/S94A overexpression also increased MYC protein expression but not MYC mRNA levels (Fig. 4C). We additionally checked on whether protein stability is altered with YAP5SA overexpression, performing the cycloheximide chase assay; however, the half-life of MYC did not change upon YAP activation (Fig. 4D), indicating that changes in protein stability did not mediate the MYC upregulation.

YAP reportedly regulates MYC through repression of miRNA processing (44), so we asked whether the same mechanism operates in the gastric epithelial cells. We sequenced miRNAs from HFE-145 cells overexpressing YAP-5SA or a vector control. Among the miRNAs significantly downregulated upon YAP activation, we found two (miR-34a-5p and miR-21-5p) that are known to repress MYC and three (miR-374b-5p, miR-34a-5p, and miR-449b-5p) that are predicted to target the MYC 3′-UTR (Fig. 4E; Supplementary Fig. S5C; refs. 49, 50). Using a miRNA reporter assay, we confirmed all four miRNAs from our sequencing results can bind the MYC 3′-UTR (Fig. 4F). Moreover, we verified that two of these miRNAs (miR-374b-5p and miR-449b-5p) functionally repress MYC protein expression in the HFE-145 and AGS cell lines (Supplementary Fig. S5D).

We went on to validate detailed mechanisms that have been suggested to regulate MYC through miRNA. According to the literature (44), nuclear YAP sequesters p72 (a regulatory component of the miRNA processing machinery) to prevent its binding to DROSHA/DGCR8, leading to a decrease in mature miRNA processing. This causes downregulation of the miRNAs that repress MYC, thereby resulting in increased MYC expression upon YAP activation. Thus, we asked whether overexpression of p72 against a background of high YAP activity could restore miRNA processing, consequently repressing MYC. Indeed, we found that overexpression of DDX17, the gene that encodes p72, in the YAP-5SA-expressing cells significantly downregulated MYC protein expression, even under YAP active conditions (Fig. 4G). On the basis of these data and on previous findings, we confirmed that MYC is upregulated by YAP at the posttranscriptional level via its regulation through miRNA processing.

We then asked whether YAP/TAZ also transcriptionally regulates MYC. YAP5SA overexpression did not increase MYC mRNA levels (Fig. 4B and C), so we evaluated MYC levels upon YAP/TAZ depletion. We tested this effect in HFE-145 cells and in two gastric cancer cell lines (AGS, and MKN74) that display high YAP/TAZ and MYC protein expression. Knockdown of YAP and TAZ in these cell lines via the expression of short hairpin RNAs induced a significant reduction in both MYC mRNA and protein levels (Fig. 5A and B; Supplementary Fig. S6A).

To confirm MYC as a direct transcriptional target of YAP, we performed a chromatin immunoprecipitation assay in the YAP-5SA–overexpressing HFE-145 cells. In the resulting data, we looked for YAP binding at the previously reported MYC enhancer regions and in promoter regions with at least three TEAD-binding consensus motifs (CATTCC; ref. 45). As shown in Fig. 5C, MYC enhancers 4 and 6 showed enrichment of YAP binding (Fig. 5C; Supplementary Fig. S6B). Altogether, these results are consistent with a role for YAP and TAZ in the transcriptional regulation of MYC.

After confirming that YAP/TAZ regulates MYC at the transcriptional and posttranscriptional levels, we incorporated these two mechanisms into a new model of MYC regulation (Fig. 5D). In this model, under baseline conditions, YAP binds to MYC enhancers to transcriptionally control MYC expression. At the same time, miRNAs processed by the DROSHA–DGCR8–p72 complex repress MYC. These two regulatory mechanisms oppose one another to maintain tight, homeostatic control of MYC levels. When YAP is activated, however, there is no additional transcriptional regulation; instead, increased nuclear YAP sequesters p72 and downregulates miRNA processing. The outcome is decreased MYC repression, causing a shift toward elevated MYC expression (Fig. 5D).

To confirm MYC's functional importance as a key signaling molecule downstream of YAP/TAZ, we asked whether inhibition of MYC limits tumorigenic potential. We took pharmacologic approaches and first used JQ1, a BET bromodomain inhibitor, which has been widely employed to inhibit MYC both in vitro and in vivo (51–53). As expected, JQ1 treatment in HFE-145 cells effectively inhibited MYC even in the presence of YAP-5SA overexpression (Fig. 6A and B). Because YAP activation promotes cellular transformation (Fig. 2C), we next asked whether this transformation can be rescued by addition of JQ1. We performed a low-attachment growth assay, which facilitates test drug efficiency during cellular transformation compared with the soft agar assay (54). Consistent with our other results, YAP-5SA overexpression increased colony formation in low attachment conditions. In line with our hypothesis, the addition of JQ1 effectively hindered the cellular transformation phenotype enhanced by YAP-5SA (Fig. 6C and D).

To corroborate our findings, we tested the effect of a more specific MYC inhibitor under the same conditions. 10058-F4 inhibits the dimerization of MYC-MAX, thereby blocking transcription of MYC downstream targets (55), as shown in Fig. 6E. Similar to JQ1, 10058-F4 inhibition of MYC also suppressed the anchorage-independent growth intensified by YAP-5SA overexpression (Fig. 6C and D).

Extending these results, we asked whether MYC inhibition blocks the tumorigenic phenotype in vivo. 10058-F4 was unsuitable for in vivo application because of its rapid metabolism, so we tested this effect with JQ1 (56). Starting 3 weeks after tamoxifen induction, we treated Lats1/2^iΔLgr5 mice with either JQ1 or vehicle daily for 5 weeks and compared their gastric epithelium with that of vehicle-treated WT mice (Fig. 6F). As expected, the vehicle-treated Lats1/2^iΔLgr5 animals showed gastric epithelial hyperplasia at 8 weeks post-tamoxifen induction. Mice treated with JQ1, however, were comparable with vehicle-treated WT mice, with no significant evidence of hyperplasia. The JQ1-treated mice also showed reduced MYC expression, supporting the efficacy of JQ1 in MYC blockade (Fig. 6G; Supplementary Table S1B). Together, these data strongly suggested that pharmacologic inhibition of MYC blocked the oncogenic activity of YAP/TAZ. They also corroborate the role of MYC as a key signaling mediator downstream of YAP/TAZ in gastric tumorigenesis.

Having shown that MYC mediates the oncogenic signals associated with YAP/TAZ activation, we next looked for a correlation between YAP and MYC expression in human gastric cancers. After staining a tissue microarray composed of 1,090 samples from patients with gastric cancer for the YAP and MYC proteins, we measured the signal intensity for each in a semiquantitative manner (Fig. 7A). We then converted these intensity values and the area of distribution for each sample into numeric scales to create a scoring system ranging from 0 to 12 (see Supplementary Methods for a more detailed description). Because these scores reflect the activity of each molecule, we aligned the scores of YAP and MYC and found a significant correlation between them (Fig. 7B). When we split the scores into two groups at their median value (6, with scores ≥6 designated as high and scores <6 as low), we observed high levels of MYC concurrent with high levels of YAP (Fig. 7C). This finding indicated a positive correlation between YAP and MYC and firmly supported our conclusion that YAP depends on MYC to initiate gastric carcinogenesis.

YAP and TAZ are reportedly increased in a subset of human gastric cancers, most of which are associated with poor clinical outcomes (12–14). Until now, this association has been shown only via IHC (12–14). Moreover, the importance of YAP/TAZ signaling has never been addressed in human genomics/transcriptomics studies with large numbers of samples, leaving unclear the significance of this signaling cascade in human gastric cancer (2–6). We thus reanalyzed transcriptome data from a public dataset to evaluate the global expression patterns of YAP/TAZ target genes in human gastric cancer (3). The microarray dataset reported by the Asian Cancer Research Group (GSE62254) includes gene expression profiles from 300 patients with gastric cancer along with associated clinical information. In the original study, the authors classified gastric cancers into four distinct groups based on gene expression patterns: (i) microsatellite instability (MSI), (ii) microsatellite stability/epithelial-to-mesenchymal transition (MSS/EMT), (iii) MSS/TP53+, and (iv) MSS/TP53-. Of these four groups, MSS/EMT showed the worst prognosis, which was validated in other patients with gastric cancer cohorts (Fig. 7D; ref. 3).

YAP activity has also been reported as an indicator of poor prognosis, and we observed an enrichment of EMT gene sets in our RNA sequencing data from Lats1/2^iΔLgr5 tumors (Supplementary Fig. S4B and S4C). Therefore, we hypothesized that the MSS/EMT subgroup may show upregulation of YAP signature genes. We performed a GSEA based on the classifications provided in the original study and found that the MSS/EMT subgroup did indeed show significant enrichment of YAP signature genes (Fig. 7D; Supplementary Table S2). Assuming that the expression level of well-known YAP target genes can be used as a readout of YAP activity (57), we asked whether YAP activity can act as a prognostic indicator in this dataset. After splitting the dataset based on YAP signature gene expression levels using bioinformatics software (Fig. 7E; ref. 36), we found that patients with high YAP activity showed significantly shorter periods of disease-free and overall survival (Fig. 7F). When this subclassification was applied to the original classification of ACRG, 98% of the MSS/EMT group was consisted of YAP upregulated cases, while the composition was around 30 % in the other three groups (Fig. 7G). On the basis of these analyses, we concluded that YAP is significantly activated in a subset of patients with gastric cancer and that its activity is closely related to their clinical prognosis.

To assess the similarity of the molecular expression patterns between our RNA sequencing data and the published human microarray data, we performed GSEA in the dichotomized GSE62254 dataset based on YAP signature gene expression. We found highly significant levels of overlap among the gene sets enriched in the YAP-activated tumor subgroup and in the RNA sequencing data from our mouse model (Figs. 3C and D and 7H). And in line with the sequencing results, Wnt signaling was not enriched in either MSS/EMT or YAP-activated subgroup when the dataset was reclassified (Fig. 7I). This also supports that upregulation of MYC is a direct consequence of YAP activation, rather than an indirect effect through Wnt signaling pathway. Altogether, these results corroborate that our YAP/TAZ-induced gastric tumor model mimics the molecular expression pattern of human gastric cancer with YAP activation.

In this study, we developed a novel gastric tumor model by activating YAP/TAZ in gastric epithelial stem cells. Although gastric cancer has been reported to arise from mature differentiated cells and from stem cells, we add a layer of evidence that pyloric stem cells can serve as the origin of tumorigenesis. We demonstrate that YAP/TAZ activation by Lats1/2 deletion in pyloric stem cells induces a step-wise tumorigenic progression from low-grade intraepithelial neoplasia to intramucosal carcinoma. This finding suggests that Lats1/2-mediated inhibition of Yap/Taz is required to restrict the proliferation of Lgr5-postive gastric epithelial stem cells. It is necessary to keep the mice for a longer period to confirm whether these mice develop advanced stages of cancer and eventually metastasize into distant organs. However, most of Lats-deleted mice became moribund at around 6 months post-tamoxifen injection, which precluded long-term follow-up. It seems their development of skin lesions (i.e., hair follicular hyperplasia and crust formation all over the body) may be one of the causes of morbidity. Because Lgr5 is the most prominent intestinal stem cell marker in crypt basal cells (33), it is unclear why we observed no related phenotypes in the small or large intestines and saw no histologic changes in any of the other organs (i.e., liver, pancreas, and ovary) reported to harbor Lgr5⁺ cells. Further studies are ongoing to disclose the mechanism underlying the differences in these phenotypes.

It is interesting that Lats knockout can cause distinct phenotypes in different organs. Previously, our group reported that liver-specific Lats depletion induces an expansion of immature biliary epithelial cells and fibroblasts but not tumor formation (58). Importantly, rather than causing hyperproliferation, we found the activation of YAP in hepatocytes induces a transdifferentiation into cells of the biliary epithelial lineage as a consequence of TGFβ upregulation. This change also triggers senescence mechanisms in YAP-activated hepatocytes, as reflected in the RNA sequencing results. In contrast to what we observed in the gastric epithelium, we observed p53 pathway upregulation and E2F target downregulation. This implies the LATS-YAP signaling cascade regulates diverse physiologic functions via various mechanisms in a context-dependent manner.

We injected tamoxifen to induce Cre expression at the desired time point. Intriguingly, one study reports that injections of 5 mg tamoxifen per 20 g of mouse body weight can trigger injury to the gastric epithelium and cause metaplasia (59). In our conditions, however, we detected no pathologic changes with tamoxifen injections alone, even at early (24- to 120-hour) time points. This distinction may be attributable to the smaller doses of tamoxifen we injected (75 mg/kg mouse body weight = 1.5 mg/20 g).

Using RNA sequencing, we identified MYC as the key downstream molecular target of YAP/TAZ in gastric tumorigenesis. MYC is a well-known oncogene involved in cell-cycle progression (43), so our transcriptome analysis points to the activation of cell-cycle–related genes, corroborating the significance of MYC as the key molecular target. In functional studies, we confirmed that MYC inhibition significantly rescues the cancer-related phenotypes of YAP. Furthermore, we found a positive correlation between YAP and MYC in human gastric cancers, also supporting YAP's regulation of MYC as a key molecular mediator of gastric tumorigenesis. Of note, Yorkie (the Drosophila homolog of YAP) and dMyc directly regulate one another in flies (60, 61). Two other studies showed that YAP/TAZ regulates MYC in mammalian cells, but the mechanisms by which YAP/TAZ did so were discrepant (44, 45). Mori and colleagues found that YAP regulates MYC at the posttranscriptional level through miRNA repression. Nuclear YAP directly interacts with p72 (encoded by DDX17), a regulatory component of the miRNA processing machinery, and sequesters it from DROSHA and DGCR8, thereby inhibiting miRNA processing (44). As a consequence, miRNAs that inhibit MYC are downregulated upon YAP activation, leading to increased MYC protein expression. However, Zanconato and colleagues (45) suggested a different mechanism, showing that MYC is a direct transcriptional target of YAP/TAZ. In their study, they found that YAP/TAZ binds MYC enhancers and that knockdown of YAP/TAZ reduces MYC mRNA levels.

In our own experiments, we confirmed the existence of both transcriptional and posttranscriptional regulatory mechanisms and integrated them into a model explaining the tight regulation of MYC expression by YAP/TAZ. Under physiologic conditions, MYC is maintained in an equilibrium between transcription and repression. When oncogenic YAP is activated, however, it induces a shift toward increased MYC expression through downregulating miRNA repression of MYC. Nevertheless, MYC is a transcriptional target of YAP that is involved in regulating cell growth, and that role has been conserved from flies to humans.

It is important to note that CTGF, AREG, and BIRC5 are all well-known targets of YAP that are reportedly related to gastric cancer (62–64). It is therefore plausible that these genes may contribute directly to YAP-induced gastric tumorigenesis. We are convinced, however, that it is MYC that has the most influence relaying YAP signals because MYC inhibition rescues the observed YAP-dependent oncogenic phenotypes.

The gastric tumor model we present here seems to reflect well the molecular pattern of a subgroup of human gastric cancers. As described above, none of the human genomics/transcriptomics studies have revealed the significance of Hippo-YAP/TAZ signaling in gastric cancer. Here, through analyzing a public microarray dataset containing a large number of patient samples, we show that a subset of human gastric cancers (the MSS/EMT subtype) shows strong YAP target gene enrichment. When we reclassified the dataset based on YAP signature gene expression, we observed an inverse correlation between clinical outcomes and YAP activity. Furthermore, the enriched gene sets in the YAP-activated human gastric cancer subgroup were similar to the upregulated genes in our murine gastric tumors, reflecting the clinical relevance of this mouse model.

Previous studies reporting poor clinical outcome with YAP activity show inconsistent results. Song and colleagues reported that nuclear YAP1 expression did not correlate with overall survival when all patients with gastric cancer were analyzed together, but was an independent prognostic factor only for the intestinal type (13). Other studies by Hu and colleagues, and Kang and colleagues showed that YAP expression was correlated with poor clinical outcome regardless of the histologic subtype (12, 14). Analyses from the original article of our microarray data showed that MSS/EMT subgroup, which is mainly consisted of diffuse-type cancer, displayed the worst clinical outcome (3). The underlying reasons to the differences shown in these reports remain elusive at this point. One of the possibilities is that classifying gastric cancer by Lauren's subtypes could have high inter-observer variabilities (65, 66). As there is no “gold standard” for defining the Lauren's classification, discordance between pathologists vary from 20% to 30% and there also exists a considerable portion of mixed-type gastric cancer. Another possibility is the difference in measuring YAP activity between the studies. Previous studies used IHC staining, whereas our data used YAP signature gene expression as the indicator for YAP activity.

YAP/TAZ is a well-known potent oncogene in many different cancers. Nevertheless, mutations in YAP/TAZ or the core upstream Hippo pathway regulators are rarely reported (11). This lack of genomic evidence leads to the hypothesis that various extracellular cues may be responsible for the upregulation of YAP/TAZ in cancers. Gastric cancer is associated with diverse carcinogenic stimuli including Helicobacter pylori infection, smoking, and diet (e.g., ingested nitrates/nitrites and high salt). These triggers induce repetitive injury and regeneration cycles in which the gastric epithelial cells and their microenvironment are affected by mechanical tension and extracellular ligands. Because YAP/TAZ seems to act as a nexus for these various signals, it may be possible to link such changes as potential cues for YAP/TAZ activation (10, 67). This idea is supported by the upregulation of YAP expression in Helicobacter-infected murine gastric epithelium and with the activation of YAP signaling upon cholinergic neural stimulation of the gastric tissue (15, 68). In addition, RUNX3, which is not a member of the canonical Hippo pathway, reportedly regulates the role the YAP–TEAD complex plays in gastric tumorigenesis (69). Thus, YAP/TAZ activation may result from various carcinogenic stimuli rather than from genomic mutations, which might explain why YAP/TAZ has been overlooked in previous studies mainly focused on genomic mutations (2–6).

Together, our model provides solid genetic evidence that YAP/TAZ activation can initiate gastric tumoriogenesis in vivo. We also demonstrate that this signaling cascade is enriched in a subset of human gastric cancers, strengthening its clinical importance. Therefore, we propose further investigation of YAP/TAZ in future translational studies. Because the YAP/TAZ–active gastric cancer subgroup of patients showed worse prognoses, YAP/TAZ activity may be a useful marker for predicting clinical outcomes. Our results also have led us to investigate the therapeutic impact of YAP/TAZ or MYC blockade in patients showing upregulation of this signaling cascade.

No potential conflicts of interest were disclosed.

Conception and design: W. Choi, G.Y. Koh, D.-S. Lim

Development of methodology: W. Choi, J. Kim, D.-S. Lim

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W. Choi, J. Kim, J. Park, D.T. Smoot, S.-Y. Kim, C. Choi, D.-S. Lim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W. Choi, C. Choi, D.-S. Lim

Writing, review, and/or revision of the manuscript: W. Choi, D. Hwang, G.Y. Koh, D.-S. Lim

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Hwang, J.-H. Kim, H. Ashktorab, C. Choi, D.-S. Lim

Study supervision: D.-H. Lee, G.Y. Koh, D.-S. Lim

We thank Dr. Young-Yun Kong (Seoul National University) for providing Lgr5-EGFP-IRES-creERT2 mice for this study. DDX17 plasmid was a generous gift from Dr. Didier Auboeuf (INSERM, France). JQ1 for in vivo treatment was kindly provided by Dr. Hee-Dong Park (LG Chem, Korea). The biospecimens and data used for this study were provided by the Biobank of Chonnam National University Hwasun Hospital, a member of the Korea Biobank Network. This study was supported by a grant from the National Creative Research Initiatives Center Program (NRF-2010-0018277 to D-S. Lim), Korean Foundation for Cancer Research (KFCR-2016-004 to W. Choi), Individual Basic Science & Engineering Research Program (NRF-2016R1D1A1B03935764 to D.-H. Lee), and Institute for Basic Science funded by the Ministry of Science and ICT, Republic of Korea (IBS-R025-D1 to G.Y. Koh).

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.