RETRACTED: KRAS Activation in Gastric Adenocarcinoma Stimulates Epithelial-to-Mesenchymal Transition to Cancer Stem–Like Cells and Promotes Metastasis

There are nearly one million new gastric cancer cases and nearly 700,000 gastric cancer–related deaths worldwide per year, and thus gastric cancer accounts for almost 10% of all cancer deaths (1). Gastric adenocarcinomas comprise the vast majority of gastric cancers. The majority of patients with gastric adenocarcinoma present with locally advanced or metastatic disease. The response rate of gastric adenocarcinoma to multiagent chemotherapy can be 50% or greater, but nearly all patients develop chemotherapy resistance, and median survival is extended only to 10 to 12 months (2). Thus, new therapies are needed.

Genes encoding the Receptor Tyrosine Kinase (RTK)-RAS signaling pathway and the tumor-suppressor TP53 are altered in 60% and 50% of gastric adenocarcinomas, respectively (3). The RAS family of proteins (in humans, HRAS, KRAS, and NRAS) are small GTPases involved in cellular signal transduction supporting cell growth and survival (4). KRAS is amplified or mutated in 17% of gastric adenocarcinomas (3). Upon stimulation by upstream receptors, KRAS switches from an inactive, GDP-bound form to an active, GTP-bound form. This conformational change leads to its binding with RAF. KRAS recruits RAF to the membrane where is promotes RAF dimerization and activation. Activated RAF phosphorylates and activates MEK, and activated MEK phosphorylates and activates ERK.

There is some evidence that RTK-RAS signaling is important in the epithelial-to-mesenchymal transition (EMT) and maintenance of gastric cancer stem–like cells (CSC). CSCs, the existence of which is still somewhat controversial, share properties of normal stem cells such as the capacity for self-renewal and differentiation (5), and may be the source of metastases (6). Many of the phenotypic differences between CSCs and bulk tumor cells that lack stemness can be attributed to epigenetic changes caused by the EMT program (7). The CSC paradigm can explain how epigenetic changes can result in phenotypic diversity within tumor cells and lead to chemotherapy resistance. As most conventional chemotherapies do not reliably eradicate CSCs, treatment strategies that target these cells would both reverse chemotherapy resistance and prevent relapse. Some evidence linking RTK-RAS signaling to EMT and CSCs comes from Voon and colleagues, who treated Runx3^−/− p53^−/− murine gastric epithelial cells with TGFβ1 to induce EMT and found an increase in the EGFR/Ras gene expression signature (8). The addition of EGF or the increased expression of Kras led to increased sphere formation and colony formation in soft agar, suggesting that the EGFR/Ras pathway is involved in the promotion of EMT to generate CSCs. Although the role of the RTK-RAS pathway in EMT and CSCs has been more extensively studied in other types of cancer, there are relatively few studies specifically in gastric adenocarcinoma.

We have previously shown that oncogenic Kras can increase gastric tumorigenesis and metastasis in a genetically engineered mouse model (9). In gastric adenocarcinoma driven by Cdh1 and Trp53 loss in gastric parietal cells, 69% of mice developed diffuse-type gastric adenocarcinoma that metastasized to lymph nodes at 1 year (10). Combining that with oncogenic Kras (Kras^G12D) increased the penetrance of gastric adenocarcinoma development to 100% and reduced survival to 76 days. In these mice, both intestinal and diffuse primary tumors are observed throughout the stomach, as well as lymph node, lung, and liver metastases (identified by a YFP reporter). When Tcon mice are treated with a MEK inhibitor starting at 4 weeks of age, median survival increases from 76 to 95 days. In the present study, we investigate whether KRAS activation promotes EMT and acquisition of CSC phenotypes, including metastatic potential and chemoresistance.

AGS (RRID CVCL_0139) and NCI-N87 (RRID CVCL_1603) subsequently referred to as N87 are Lauren intestinal-type gastric adenocarcinoma cell lines, and KATOIII (RRID CVCL_0371), SNU-668 (RRID CVCL_5081), and MKN-45 (RRID CVCL_0434) are Lauren diffuse-type gastric adenocarcinoma cell lines. AGS, N87, and MKN-45 cells were obtained from the America Type Culture Collection (ATCC). The ATCC uses morphology, karyotyping, and PCR-based approaches to confirm the identity of human cell lines and to rule out both intraspecies and interspecies contamination. These include an assay to detect species-specific variants of the cytochrome C oxidase I gene (COI analysis) to rule out interspecies contamination and short tandem repeat profiling to distinguish between individual human cell lines and rule out intraspecies contamination. KATOIII and SNU-668 were obtained from the Korean Cell Line Bank (KCLB). KCLB uses DNA fingerprinting analysis, species verification testing, Mycoplasma contamination testing, and viral contamination testing. Cancer cell lines were actively passaged for less than 6 months from the time that they were received from the ATCC or KCLB, and United Kingdom Co-ordinating Committee on Cancer Research (UKCCCR) guidelines were followed (11). KATOIII cells were maintained in DMEM, and all others were maintained in RPMI 1640. All media were supplemented with 10% FBS, 100 U/mL penicillin and 100 mg/mL streptomycin, and l-glutamine 2 mmol/L ("regular media").

The immortalized human normal gastric epithelial cell line HFE145 was a gift from Dr. Hassan Ashktorab and Duane T. Smoot (Howard University, Washington, DC), and was maintained in RPMI 1640 (12). The HFE145 cell line was tested for Mycoplasma contamination using the ATCC Mycoplasma PCR testing service. It was not tested for cell line authentication.

5-Fluorouracil (5-FU) and cisplatin were purchased from US Biological and Enzo Life Sciences, respectively. The MEK inhibitor PD0325901 (S1036) was purchased from Selleckchem. Accutase (AT104) was purchased from Innovative Cell Technologies, Inc.

Cells were resuspended in DMEM-F12 containing 20 ng/mL of EGF, bFGF, N-2 (1X), and B27 (1X; “spheroid media”) and plated on Ultra-Low Attachment culture dishes (Corning Life Sciences) as previously described (13). Spheroids were collected after 5 to 7 days unless otherwise noted. Protein was extracted for analysis, or cells were dissociated with Accutase (Innovative Cell Technologies) and used for other experiments (14).

KRAS was silenced via lentiviral transduction of human KRAS shRNA (SC-35731-V; Santa Cruz Biotechnology) and mouse Kras shRNA (iV048022; abm Inc.). Scramble shRNA control (SC-108080; Santa Cruz Biotechnology) and GFP (sc-108084, Santa Cruz Biotechnology) constructs were also used. Maximal knockdown occurred 72 to 96 hours after transduction. KRAS^G12V and KRAS^WT were overexpressed using KRAS^G12V lentiviral activation particles (LVP1139-GP; GenTarget Inc.) and KRAS^WT lentiviral activation particles (LVP201104; abm Inc.), respectively, following the manufacturer's protocol. Control lentiviral activation particles (sc-437282; Santa Cruz Biotechnology) served as controls.

Activated RAS affinity precipitation assays were performed according to the manufacturer's protocol (9). Briefly, 250 mg of cell lysates were incubated with 10 mL of RAS-binding domain agarose beads (#14-278, Upstate, Millipore) overnight at 4°C. After washing 3 times with washing buffer (RIPA buffer with 1 mmol/L Na₃VO₄, 10 mg/mL leupeptin, 10 mg/mL aprotinin, and 25 mmol/L NaF), immunoprecipitation reactions were separated by 12% SDS-PAGE, and Ras was detected by Western blotting (antibody from Cell Signaling Technology, #3965), with β-actin (Sigma, A5441) as loading control.

Spheroid cells were dissociated using Accutase, and monolayer cells were collected with trypsin. To assay proliferation, 1 × 10⁴ cells were plated onto 96-well flat bottom plates and maintained in regular media overnight. An MTT Cell Proliferation Assay Kit (Colorimetric) was used to assess cell number by optical density after 3 days (15). Day 1 represents the time of cell plating. Data reflect the mean of six samples. Soft-agar colony formation from single cells was assayed as previously described (13). To measure migration and invasion, cells (2 × 10⁴ cells/well) were suspended in 0.2 mL of serum-free DMEM and loaded in the upper well of Transwell chambers (8-μm pore size; Corning); the lower well was filled with 0.8 mL of DMEM with serum. For the invasion assay, the upper wells of the chambers were precoated with BD Matrigel matrix (BD Biosciences) and 10 mg/mL growth factor; migration assays employed noncoated Transwell chambers. After incubation for 48 hours at 37°C, cells on the upper surface of the filter were removed with a cotton swab, and invading or migratory cells on the lower surface of the filter were fixed and stained with a Diff-Quick kit (Fisher Scientific) and photographed at 20x magnification. Invasiveness and migration were quantified as the average number of cells on phase-contrast microscopy among 5 microscopic fields per well.

For FACS, cells were dissociated using Accutase (Innovative Cell Technologies) and resuspended in PBS containing 0.5% BSA. Cells were stained with FITC-conjugated anti-CD44 (BD555478) or isotype control antibody (BD555742) and analyzed on a FACSCalibur (BD Biosciences) using Cell Quest software. CD44-positive cells were collected using a magnetic cell-sorting system (Miltenyi Biotech). Briefly, cells were dissociated using Accutase, stained with CD44-Micro Beads, and passed through an LS magnetic column that retains CD44-positive cells. CD44-positive cells were then eluted from the column after removal of the magnet and quantified by immunofluorescence using FITC-conjugated CD44 antibody (555478; BD Biosciences).

Samples were collected in RIPA buffer (Sigma) containing Complete Protease Inhibitor Cocktail (Roche Diagnostics), and protein concentration was determined by Bio-Rad Protein Assay. Western blot analysis was performed using the following antibodies: KRAS (sc-30), Pan-RAS (sc-32), ERK1 (sc-271270), and c-Myc (sc-40) from Santa Cruz Biotechnology; Sox2 (#2748, #3579), Oct-4 (#2750), Nanog (#4893), Slug (#9585), Snail (#3879), phospho-ERK1/2 (#9101), RAS (#3965), MEK1/2 (Ser217/221; #4694), phospho-MEK1/2 (Ser217/221; #9121, #9154), CD44 (#3578, #3570), E-cadherin (#14472), vimentin (#3932), and cleaved caspase-3 (#9661) from Cell Signaling Technology; N-cadherin (BD610920) and E-cadherin (BD610181) from BD Biosciences; Zeb1 (NBP-1-05987) from Novus Biologicals; and β-actin from Sigma.

All mouse protocols were approved by the MSKCC Institutional Animal Care and Use Committee. Atp4b-Cre; Cdh1^fl/fl; LSL-Kras^G12D; Trp53^fl/fl; Rosa26LSL^YFP/YFP triple conditional (Tcon) mice were generated as previously described (9). Tissues were dissected from indicated mice. Paraffin-embedded sections were deparaffinized, and sections were examined by IHC as described below.

For subcutaneous flank tumors, 5 × 10⁶ AGS cells previously transduced with sh.KRAS or sh.Scr were resuspended in 100 μL of Hank's balanced salt solution and injected subcutaneously into the right flank of athymic, 6- to 8-week-old male BALB/c nu/nu mice following isoflurane anesthesia. Mice were assigned into treatment groups (5 mice per group) when tumors reached 50 to 100 mm³ in volume, designated as day 0. Cisplatin (2 mg/kg) or carrier (PBS) was injected intraperitoneally once a week. Tumors were measured 3x per week for 2 weeks, and tumor volume (TV) was calculated as TV = length x (width)² x 0.52. After mice were sacrificed, tumors were excised and cut into thirds.

For induction of lung metastases, mice were injected via the tail vein with 2 × 10⁷ cells of monolayers and 1 × 10⁴ cells of spheroids grown from cells previously transduced with sh.KRAS or sh.Scr. After sacrifice at different time points, tumor colonies in the lungs were counted. Each tumor of all mouse groups was fixed in 10% buffered formalin for 24 hours, embedded in paraffin, and processed into 5 μm sections.

Tcon3077 and Tcon3944 gastric tumors in Tcon mice were harvested, and organoids were isolated as previously described (16). For in vitro culture, organoids were mixed with 50 μL of Matrigel (Cat. 354248, BD Bioscience) and plated in 24-well plates. After polymerization of Matrigel, cells were overlaid with DMEM/F12 supplemented with penicillin/streptomycin, 10 mmol/L HEPES, GlutaMAX, 1x B27, 1x N2 (Invitrogen), and 1.25 mmol/L N-acetylcysteine (Sigma-Aldrich), and the following growth factors, cofactors, and hormones were added: 0.05 μg/mL EGF, 0.1 μg/mL fibroblast growth factor-basic (FGF-Basic), 0.01 μmol/L gastrin I, 10 mmol/L nicotinamide, 10 μmol/L Y-27632, SB202190 (all from Sigma Aldrich), 1 μmol/L prostaglandin E2 (Tocris Bioscience), 0.5 μg/mL recombinant R-spondin 1, 0.1 μg/mL mNoggin, 0.1 μg /mL FGF-10 (all from PeproTech), and 0.1 μg /mL Wnt3A and 0.5 μmol/L A83-01 [a TGFβ kinase/activin receptor-like kinase (ALK 5) inhibitor; both from R&D Systems]. Organoids were passaged every week at a 1:5–1:8 split ratio by removing them from Matrigel using BD Cell Recovery Solution (BD Biosciences) following the manufacturer's instructions and transferring them to fresh Matrigel.

Spheroids were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100 in PBS. Following fixation, cells were incubated with antibodies against Sox2 (#4900; Cell Signaling Technology), E-cadherin (BD610181; BD Biosciences), p-MEK1/2 (#9121; Cell Signaling Technology), Slug (#9585; Cell Signaling Technology), and/ or CD44-FITC (555478; BD Biosciences) in a solution of PBS with 1% BSA and 0.1% Triton X-100 at 4°C overnight. Staining was visualized using anti-mouse AlexaFluor 488 (A11005; Life Technologies) and anti-rabbit AlexaFluor 594 (A11012; Life Technologies) with nuclear counterstaining using DAPI Solution, and imaging on an inverted confocal microscope. Images were processed using Imaris 7.6.

For IHC, formalin-fixed, paraffin-embedded sections were deparaffinized by xylene and rehydrated. Sections were either stained with hematoxylin and eosin (H&E) or immunostained using the VECTASTAIN Elite ABC Kit (Vector Laboratories Inc.) following the manufacturer's instructions and standard protocols (17). Antibodies included CD44 (#3570), p-MEK1/2 (#3579 and #9121), p-ERK1/2 (#9101; from BD Biosciences), and Snail (sc-271977, Santa Cruz Biotechnology).

We performed immunofluorescence as previously described (18). Antibodies used were as follows: anti-human CD44 (#3570), anti–p-MEK1/2 (#3579), and anti–p-EKR1/2 (#9101). Nuclei were counterstained using DAPI. Stained cells were visualized using an inverted confocal microscope, and images were processed using Imaris 7.6.

To evaluate expression of CD44, phospho-MEK1/2, and phospho-ERK1/2 in normal and stomach cancer tissues, commercially available paraffin-embedded tissue array slides containing 40 stomach cancer and 8 corresponding normal tissues (A209 II, ISU ABXIS, Seoul, Republic of Korea) were purchased. Sections were deparaffinized and then incubated with anti-human CD44 (#3570), anti–p-MEK1/2 (#9121), and anti–p-ERK1/2 (#9101, all from BD Biosciences) in a solution of PBS with 1% BSA and 0.1% Triton X-100 at 4°C overnight. Staining was visualized using anti-mouse Alexa Fluor 488 and anti-rabbit AlexaFluor 594, with nuclear counterstaining using DAPI and imaging on an inverted confocal microscope. Images were processed using Imaris 7.6.

Tissue microarrays (TMA) of tumors from 115 patients were also constructed using a precision tissue array instrument (FMUUH: Shanghai Outdo Biotech Co. Ltd.; MSKCC: Beecher Instruments, Inc.). All subjects provided written-informed consent for tumor analyses prospectively, and Institutional Review Board approval was obtained for this study. The study was conducted in accordance with the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS). These patients had adenocarcinoma arising in the stomach or gastroesophageal junction of Siewert type II or III and underwent radical gastrectomy between May 2013 and March 2014 at Fujian Medical University Union Hospital (Fujian, China). A representative core biopsy (2 mm diameter) was obtained from each tumor and embedded.

IHC for CD44 and p-MEK1/2 was performed as described above. The following information was collected from the FMUUH gastric database on each patient: age, gender, tumor invasion depth, number of lymph node metastases, and time to recurrence and death. The frequency of staining for each marker was scored as follows: ≤ 5% positive cells = 0, 6% to 25% positive cells = 1, 26% to 50% positive cells = 2, ≥ 51% positive cells = 3. If the total score (intensity x percentage score) was < 3, expression was defined as low; and if the score was ≥ 4, it was considered high.

Data are represented as mean ± SD unless otherwise noted. Groups were compared using Instant 3.10 software (GraphPad). P values were calculated using the Student t test. For comparisons among more than two groups, treatment groups were compared with the control group using one-way ANOVA with Bonferroni adjustment for multiple comparisons.

For human data analyses, continuous values are expressed as mean ± SD and analyzed using the Student t test. Categorical variables are analyzed using χ² or Fisher exact test. Overall survival curves were plotted by the Kaplan–Meier method and compared using the log-rank test. Cox proportional hazards regression modeling was used to examine the relationship between CD44 and p-MEK1/2 expression and survival while controlling for confounding covariates. Analyses were performed using IBM SPSS software for Windows version 21 (IBM). A P value less than 0.05 was considered statistically significant.

Tumor-derived organoids are an in vitro model that can recapitulate the pathophysiology of the original tumors along with preserving cellular heterogeneity and self-renewal capacity (19). We developed two organoid cultures from primary gastric tumors in Tcon mice, labeled Tcon3077 and TconY3944, as described in the Materials and Methods section. These tumor-derived organoids express oncogenic Kras^G12D. Kras knockdown in organoid cultures by transduced Kras shRNA was confirmed by Western blot analysis (Supplementary Fig. S1A) and led to a mild 9%–11% decrease in in vitro cell proliferation at 5 days (Supplementary Fig. S1B). Kras knockdown also decreased organoid size by 65% to 69%, disrupted normal organoid architecture (Fig. 1A), and decreased expression of EMT-related proteins including N-cadherin, Snail, and Slug (Fig. 1B and C). Kras shRNA also decreased migration and invasion of organoid cells by 53% to 74% (Fig. 1D; Supplementary Fig. S1C).

Because EMT can lead to the acquisition of CSC phenotypes (20), we next examined whether Kras knockdown affected expression of self-renewal transcription factors and formation of spheroids (an in vitro method for enriching for CSCs). Kras shRNA decreased expression of Sox2 protein (Fig. 1E) and decreased spheroid formation by 79% to 89% (Fig. 1F; Supplementary Fig. S1D). Thus, Kras knockdown appears to inhibit EMT and conversion to CSCs in these tumor-derived organoids.

If CSCs are the source of metastases, we reasoned that the number of CSCs should be higher in microscopic metastases compared with primary tumors or macroscopic metastases. Takaishi and colleagues tested several gastric cancer cell lines for CSC markers, and only CD44 expression was associated with tumor formation in immunodeficient mice and spheroid colony formation in vitro (21). We therefore examined primary gastric tumors, microscopic lung metastases, and macroscopic lung metastases from Tcon mice for expression of the gastric CSC marker CD44 and found that the number of CD44-expressing cells in microscopic lung metastases was 2.8-fold higher than in primary gastric tumors and 5.4-fold higher than in macroscopic lung metastases (Fig. 1G; Supplementary Fig. S1E). We also found higher levels of phosphorylated MEK, a downstream target of KRAS, in lung micrometastases compared with larger primary tumors and lung macrometastases (Fig. 1G; Supplementary Fig. S1F and S1G). These data suggest that oncogenic Kras in CSCs promotes metastasis in this model of gastric adenocarcinoma.

To further examine the role of CSCs and RTK-RAS activation in human gastric adenocarcinomas, we next assessed expression of CD44 and activation of MEK and ERK by immunostaining a commercially available TMA containing 40 human gastric adenocarcinomas and 8 human normal gastric tissues (A209 II, ISU ABXIS). Levels of CD44, phospho-MEK, and phospho-ERK levels were 3.1- to 3.5-fold higher in tumor tissue compared with normal tissue (Fig. 2A and B).

To determine if CD44 expression and MEK activation are prognostic factors for survival in patients with gastric adenocarcinoma, we immunostained a TMA of 115 gastric adenocarcinomas from patients undergoing curative-intent surgical resection at Fujian Medical University Union Hospital (Fujian, China; Fig. 2C). Clinical characteristics of these patients and pathologic characteristic of their tumors are shown in Supplementary Table S1. Patients whose tumors expressed high levels of CD44 or phospho-MEK had significantly worse overall survival (Fig. 2D and E). The worst overall survival was seen in the 32 patients with both high CD44 and high phospho-MEK expression in their tumors (Fig. 2F). On multivariate analysis, both increased expression of CD44 and phospho-MEK were independent prognostic factors for worse overall survival (Supplementary Table S1). Thus, increased tumor expression of the gastric CSC marker CD44 and activation of the RTK-RAS pathway as measured by MEK phosphorylation portend a worse prognosis in gastric adenocarcinoma patients undergoing surgical resection.

We next examined the functional consequences of (1) oncogenic KRAS activation in human gastric epithelial cells and (2) KRAS inhibition in gastric adenocarcinoma cell lines. For gastric epithelial cells, we used HFE-145 cells which are immortalized human, nonneoplastic gastric epithelial cells. The gastric adenocarcinoma cell lines examined included three human diffuse-type gastric adenocarcinoma cell lines (MKN-45, SNU-668, and KATOIII), two human intestinal-type gastric adenocarcinoma cell lines (AGS and NCI-N87), and two mouse gastric adenocarcinoma cell lines (Tcon3077 and Tcon3944; ref. 9). We performed targeted next-generation sequencing on our human gastric cancer cell lines and found that two gastric adenocarcinoma cell lines had activating KRAS mutations. AGS cells have a KRAS^G12D mutation, and SNU-668 cells have a KRAS^Q61K mutation. The other gastric adenocarcinoma cell lines (MKN-45, NCI-N87, and KATOIII) are KRAS wild-type. We first confirmed that KRAS activation was low in normal gastric epithelial (HFE-145) cells compared with gastric adenocarcinoma cell lines, as measured by coprecipitation of KRAS with its signaling target, RAF, and expression of phospho-MEK and phospho-ERK (Fig. 3A). Successful transduction of HFE-145 cells with oncogenic KRAS (*KRAS^G12V) or wild-type KRAS (KRAS^WT) was confirmed by Western blot; expression of KRAS and levels of phospho-MEK and phospho-ERK were increased relative to control cells (Supplementary Fig. S2A). KRAS^G12V mildly increased proliferation of HFE-145 cells in vitro (Supplementary Fig. S2B), significantly decreased expression of the epithelial marker E-cadherin, and significantly increased expression of the mesenchymal marker vimentin and the EMT transcription factor Slug (Fig. 3B). Expression of other EMT transcription factors including Snail and Zeb1 was not significantly changed. HFE-145 cells expressing oncogenic KRAS^G12V also displayed a more spindle cell morphology compared with control cells (Fig. 3C). Loss of E-cadherin expression and increased MEK phosphorylation in KRAS^G12V-expressing HFE145 cells was also apparent by immunofluorescence (Fig. 3D). Finally, KRAS^G12V increased migration and invasion by 15- to 17-fold compared with control HFE-145 cells (Fig. 3E). KRAS^WT transduction had an intermediate effect.

We next examined the effect of KRAS^G12V on CSC phenotypes and marker expression in HFE-145 cells. By FACS analysis, CD44 expression increased from 0.5% in HFE-145 monolayers to 4.4% in control HFE-145 spheroid cells and 6.4% in HFE spheroid cells expressing oncogenic KRAS^G12V (Supplementary Fig. S2C and S2D). KRAS^G12V also increased expression of self-renewal proteins such as Sox2 as assessed by immunofluorescence (Fig. 3F) and Western blot (Fig. 3G). Compared with control cells, HFE-145 cells expressing oncogenic KRAS^G12V formed 2.6-fold more spheroids (Fig. 3H). Again, KRAS^WT transduction had an intermediate effect. Together, these results show that expression of oncogenic KRAS in gastric epithelial cells promotes EMT and acquisition of CSC properties.

We assessed KRAS activation in gastric adenocarcinoma cells grown as monolayers and as spheroids via KRAS coprecipitation with RAF and phosphorylation of MEK and ERK (Fig. 4A). In all gastric adenocarcinoma cell lines, KRAS pathway activation was higher in cells grown as spheroids compared with cells grown as monolayers, as was expression of CD44 and Sox2 (Supplementary Fig. S3A).

In KATOIII cells, which express wild-type KRAS, transduction of oncogenic KRAS^G12V increased spheroid formation capacity (Fig. 4B) and expression of CD44 and Sox2 (Fig. 4C). Knockdown of KRAS using lentiviral shRNA in AGS and KATOIII cells resulted in loss of CD44 and Sox2 expression as measured by Western blot (Fig. 4D) and immunofluorescence (Supplementary Fig. S3B) as well as a 74% to 85% reduction in spheroid formation in both standard (Fig. 4E) and single-cell assays (Supplementary Fig. S3C).

We next performed a series of similar experiments using the MEK inhibitor PD0325901 rather than using KRAS shRNA. This MEK inhibitor reduced CD44 expression and Sox2 expression in AGS and KATOIII cells grown as spheroids as measured by Western blot analysis (Supplementary Fig. S4A) and by immunofluorescence (Supplementary Fig. S4B). PD0325901 also reduced spheroid formation by AGS and KATOIII cells at 5 days by 68% to 91% (Supplementary Fig. S4C), and the size of spheroids formed from single AGS and KATO III cells by 32% to 42% (Supplementary Fig. S4D).

We next examined AGS and KATO spheroid cells that were sorted into CD44(+) and CD44(-) expression by FACS. In these spheroids, 7% to 14% of cells are CD44(+) (data not shown). PD0325901 attenuated ERK phosphorylation and expression of Sox2 in CD44(+) cells (Supplementary Fig. S5A), and greatly reduced their ability to form spheroids (Supplementary Fig. S5B). MEK inhibition in CD44⁺ spheroid cells also decreased expression of EMT-related proteins, increased expression of the epithelial marker E-cadherin (Supplementary Fig. S5C), and dramatically reduced migration and invasion (Supplementary Fig. S5D).

As CSCs are generally resistant to chemotherapy (22), we examined the effects of KRAS inhibition on CSC sensitivity to 5-FU and cisplatin. AGS and KATOIII cells grown as monolayers were sensitive to these agents, whereas these same cells grown as spheroids were relatively resistant (Supplementary Fig. S6A). KRAS shRNA reversed chemotherapy resistance in AGS and KATOIII spheroid cells (Supplementary Fig. S6B). AGS flank xenografts were much more sensitive to cisplatin when KRAS was knocked down (Supplementary Fig. S6C). The effect of chemotherapy and inhibition of KRAS on AGS flank xenografts was examined for proliferation using Ki67 staining, for total apoptosis using cleaved caspase 3 staining, and for stemness using CD44 staining. Cisplatin combined with KRAS shRNA led to dramatic increases in apoptosis in CD44-positive cells compared with chemotherapy (Supplementary Fig. S6D).

We next examined the effect of reducing KRAS activity on EMT and metastasis. AGS and KATOIII cells were transduced with KRAS shRNA or a scramble control shRNA, and KRAS knockdown was confirmed by Western blot (Fig. 5A). KRAS knockdown significantly decreased expression of the mesenchymal marker N-cadherin, increased expression of the epithelial marker E-cadherin, and decreased expression of the EMT transcription factor Snail in both cell lines (Fig. 5A and B). Expression of other EMT transcription factors including Slug and Zeb1 was not significantly changed. KRAS knockdown also decreased migration and invasion by 85% to 90% (Fig. 5C) and reduced colony formation in soft agar by 86% to 87% (Fig. 5D) compared with control cells. Similar results were obtained using the MEK inhibitor PD0325901 (Supplementary Fig. S7A–S7C).

We next assessed the effect of KRAS knockdown on tumor growth and invasiveness in vivo by creating flank tumor xenografts using AGS spheroid cells transduced with KRAS shRNA or a scrambled control shRNA (Supplementary Fig. S7D). Flank tumors from control AGS cells grew faster than flank tumors from AGS cells with KRAS knockdown (Fig. 5E). Reducing KRAS activity also drastically inhibited invasion. Control AGS cells formed tumors with an infiltrating leading edge, whereas AGS cells with KRAS knockdown had a well-defined border between the tumor and normal surrounding tissues (Fig. 5E). KRAS shRNA reduced the number of infiltrating cells from 8.2 per mm² to 1.6 per mm² (Fig. 5F).

EMT and acquisition of CSC characteristics promote solid tumor metastasis (7). We examined the effect of KRAS knockdown on experimental lung metastases using a tail vein injection model in athymic nude mice. AGS cells grown as spheroids and injected into the tail vein of mice produced significantly more lung metastases than the same cells grown as monolayers (Fig. 6A). KRAS knockdown in cells grown as spheroids or as monolayers dramatically reduced the formation of lung metastases, leading to longer survival (Fig. 6A and B). Lung metastases from spheroid cells in which KRAS was knocked down had dramatically lower levels of CD44 and phospho-MEK compared with scramble control cells (Fig. 6C and D). Taken together, these data show that KRAS promotes EMT and metastasis in gastric adenocarcinoma cells.

This study demonstrates that RTK-RAS signaling promotes EMT in gastric adenocarcinoma cells, leading to the acquisition of CSC phenotypes and metastatic potential. Transduction of nontransformed gastric epithelial cells with oncogenic KRAS led to EMT, expression of the CSC marker CD44 and stem cell transcription factor Sox2, and increased spheroid formation. Gastric adenocarcinoma cell lines grown as spheroids had enrichment of CD44 expression and increased KRAS activity compared with monolayer cells. Inhibition of KRAS in gastric adenocarcinoma spheroid cells or CD44(+) cells using shRNA knockdown or pharmacologic inhibition diminished expression of Sox-2, reduced spheroid formation, and reversed chemotherapy resistance. KRAS inhibition in gastric adenocarcinoma spheroid cells also reduced expression of EMT markers including N-cadherin and Snail, greatly diminished migratory and invasive capabilities, attenuated the infiltrative nature of flank xenografts, and reduced experimental lung metastases. KRAS inhibition in organoids derived from gastric tumors in a mouse model of gastric adenocarcinoma with oncogenic Kras resulted in similar reductions in EMT and CSC phenotypes. Finally, tumor specimens from patients with gastric adenocarcinoma who underwent surgical resection of their gastric tumors revealed an association between high CD44 expression and RTK-Ras activity in their tumors and worse overall survival.

Although the relationship between KRAS activation and CSC function has not been extensively studied in gastric adenocarcinoma, there have been some studies in other gastrointestinal tumors. In colorectal cancer, Blaj and colleagues found that high MAPK activity promotes EMT and marks a progenitor cell subpopulation that was the predominant source of growing flank xenografts (23). Also, in colorectal cancer, Moon and colleagues showed that in cells carrying mutated APC, oncogenic KRAS increases expression of CSC markers (CD44, CD133, and CD166), spheroid formation, and the size of xenografts. In pancreatic CSCs, inhibition of KRAS led to downregulation of JNK signaling and loss of self-renewal and tumor-initiating capacity (24). In this study, we used spheroid formation and expression of CD44 as means of identifying gastric adenocarcinoma CSCs. These gastric adenocarcinoma spheroid cells or CD44(+) cells have increased activation of the KRAS pathway as measured by binding to RAF and phosphorylation of MEK and ERK. They also have elevated expression of the stem cell factor Slug along with increased colony formation in soft agar. Inhibition of KRAS in gastric adenocarcinoma cells using either shRNA or MEK inhibition blocked CSC phenotypes including spheroid formation, soft-agar colony formation, and chemotherapy resistance. Thus, KRAS activity may be a common pathway supporting CSC function in gastrointestinal cancers.

Our results also support a link between the EMT program, a naturally occurring transdifferentiation program long known to support the proliferation and metastasis of cancer cells (25), and the generation of CSCs. The phenotypic differences between CSCs and bulk tumor cells are predominately due to epigenetic changes which activate the EMT program (7). This link between the passage through EMT and the acquisition of stem-like properties is vital for cancer cells in order to metastasize and to survive chemotherapy, and this study highlights the role that KRAS plays in this process. In this study, we found that blockade of KRAS activity in gastric adenocarcinoma CSCs downregulates the EMT-related proteins Slug and N-cadherin and reduces migration, invasion, metastasis, and chemotherapy resistance. In Tcon mice, we found that there were more CD44(+) cells and higher phosphorylated MEK expression in micrometastases compared with macroscopic tumors, suggesting CD44(+) CSC with higher RTK-RAS activity may be the source of metastatic disease, but further studies are needed to prove this hypothesis. In addition to studies in Tcon mice, which develop primary gastric adenocarcinomas and metastases with oncogenic Kras, we used tumor-derived organoids from Tcon tumors as an in vitro model of Tcon tumors to preserve tumor heterogeneity and self-renewal capacity.”

Because RAS GTPases including KRAS are difficult to target directly with drugs because of structure-function considerations (26), we inhibited the KRAS pathway using an MEK inhibitor. Several MEK inhibitors including trametinib, cobimetinib, and binimetinib are currently FDA-approved for use in patients with BRAF-mutated melanoma (27). The MEK inhibitor used in this study, PD0325901, is currently in clinical trials for patients with various solid tumors. The inhibitory effects of PD0325901 on spheroid formation, soft-agar colony formation, EMT protein expression, migration, and invasion were similar to those of KRAS knockdown by shRNA.

Our results also suggest a potential strategy for targeting therapies that inhibit the RTK-RAS pathway to a subgroup of patients with gastric adenocarcinoma. Such therapies are urgently needed to overcome the challenge of chemotherapy resistance. The importance of appropriately selecting patients for targeted therapies is demonstrated by the success of trastuzumab plus chemotherapy in patients whose gastric adenocarcinoma overexpresses HER-2, which prolonged survival from 11 to 14 months (28), compared with the lack of benefit of combining cytotoxic chemotherapy with agents targeting the vascular endothelial growth factor A or EGF pathways in unselected patients (29–31). RTK-RAS pathway inhibition may only be effective for a subset of gastric adenocarcinomas with high RTK-RAS activity. Our finding that patients with increased tumor levels of CD44 and increased activation of the RAS pathway had significantly worse overall survival after resection of their tumors suggests that this may be a subgroup in which targeting the RTK-RAS pathway would be most beneficial.

Our use of tumor-derived organoids allowed us to investigate the relationship between KRAS activation and CSC properties in heterogeneous tumors in vitro (32). Tumor organoids can be manipulated under controlled conditions in ways that are not possible even in genetically engineered mouse models. We generated gastric cancer organoids from gastric tumors that developed in our Tcon genetically engineered mouse model (9). We performed many of the CSC and EMT assays on both gastric adenocarcinoma cell lines and organoids and confirmed that KRAS activation in gastric adenocarcinoma stimulates EMT to CSCs in both settings.

To the best of our knowledge, this is the first study to establish the importance of KRAS in gastric adenocarcinoma in terms of promoting EMT and the generation of CSCs and to demonstrate that inhibition of KRAS may be a strategy for blocking metastatic progression. Studies were performed using human and mouse gastric adenocarcinoma cells, nontransformed gastric epithelial cells, and organoids derived from a GEMM of gastric adenocarcinoma with oncogenic Kras in vitro, and the relevance of these studies was confirmed in analyses of mouse and human tumors. These studies provide rationale for studying inhibitors of the KRAS pathway to block metastasis and reverse chemotherapy resistance in gastric adenocarcinoma.

S. Ryeom has an ownership interest (including stock, patents, etc.) in Attis Lab. No potential conflicts of interest were disclosed by the other authors.

Conception and design: C. Yoon, J.-x. Lin, S.S. Yoon

Development of methodology: C. Yoon, J. Till, S.S. Yoon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Yoon, J. Till, S.-J. Cho, K.K. Chang, S.S. Yoon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Yoon, S.-J. Cho, J.-x. Lin, C.-m. Huang, S. Ryeom, S.S. Yoon

Writing, review, and/or revision of the manuscript: K.K. Chang, S. Ryeom, S.S. Yoon

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Yoon, J. Till, S.S. Yoon

Study supervision: S.S. Yoon

We would like to thank Drs. Hassan Ashktorab and Duane T. Smoot for providing the HFE-145 cell line. We also thank MSKCC senior editor Jessica Moore for reviewing this article. This study was funded by NIH grants 1R01 CA158301-01 (S.S. Yoon) and P30 CA008748 (S.S. Yoon) and by the DeGregorio Family Foundation Grant (S.S. Yoon and S. Ryeom).

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.