Introduction

The tumor stroma, including the GC stroma, consists of the extracellular matrix and various cells including fibroblasts and inflammatory cells. Among them, cancer-associated fibroblasts (CAFs), play a critical role in promoting tumor growth, invasion and angiogenesis1. CAFs are usually considered to be derived from normal fibroblasts (NFs) that are in close contact with tumor cells2. Elevated expression of pro-inflammatory genes is essential to maintain the characteristics of CAFs3. On the other hand, neoplastic cells can induce the transformation of fibroblast cells to adapt to the tumor microenvironment and co-evolve with tumor cells to foster malignancy2.

CAFs and neoplastic cells communicate via growth factors and inflammation factors4. Inflammatory signaling pathways are emerging as highly promising targets to block the communication between tumor cells and CAFs5. Among them, cyclooxygenase-2 (COX-2)/prostaglandin E2 (PGE2) signaling, which was discovered in transformed epithelial cells6, contributes uniquely to tumorigenesis through increasing the population of tumor initiating cells (TICs)7. Overexpression of COX-2 is sufficient to induce tumor formation in transgenic mouse models8, and has been proposed to exert its pro-tumor effects through the production of PGE29. PGE2 promotes tumor development by stimulating angiogenesis, inhibiting apoptosis, increasing invasion, and modulating inflammatory and immune responses10. In addition to tumor cells, elevated stromal PGE2 levels could be ascribed to fibroblasts11,12. Although the role of COX-2/PGE2 signaling in tumorigenesis has been studied extensively, the mechanism underlying the communication between tumor cells and fibroblasts remains unclear.

Increased PGE2 levels induce the expression of the inflammatory factor interleukin (IL)-6 in fibroblasts13,14. IL-6 is an important mediator involved in the crosstalk between CAFs and tumor cells3,15. Contrary to PGE2, CAF-producing IL-6 is likely to play a role mainly in the cancer cell niche3,15,16,17,18. The presence of tumor cells enhances the release of IL-6 by CAFs16. Recent findings indicate that IL-6 can also increase PGE2 production19,20, suggesting that IL-6 signaling is essential for the activation of the COX-2/PGE2 pathway in cancer cells16. Therefore, a positive PGE2-IL-6 feedback loop is likely to be involved in the crosstalk between the tumor and its stroma.

Dysregulation of microRNAs (miRNAs) is associated with tumorigenesis21 and fibroblast reprogramming in cancer22. Most studies, including those from our group, have focused on miRNAs in tumor cells, their aberrant expression and their target genes23,24, whereas miRNAs that do not change in tumor tissues are usually ignored. Recent reports suggest that miRNAs are aberrantly expressed in CAFs compared to NFs in endometrial, bladder, breast and ovarian cancers22,25,26,27. The deregulation of miRNAs contributes to the pro-tumor ability of CAF through altering the expression of their target genes. However, tumor tissues consist of different cell types, of which CAFs occupy only a small percentage. Deregulation of gene or protein in these cells is usually missed in whole-tissue analysis. Genetic mutations in CAFs are rare, implying that an alternative way of gene regulation may be active28,29. A genome-wide analysis of stromal cells from breast cancer suggests that epigenetic changes are likely to be the main regulatory mechanism in CAFs30. Therefore, it is possible that altered miRNA expression in CAFs may fine-tune PGE2 signaling to orchestrate the interaction between CAFs and neoplastic cells.

In the present study, we identify miRNAs that are deregulated in CAFs and investigate the mechanisms underlying their dysregulation as well as their role in tumorigenesis.

Results

miR-149 targets IL-6 mRNA and is downregulated in GC CAFs

To identify miRNAs that may play a role in tumorigenesis, the miR profiles of 28 pairs of gastric tumors and para-tumors were analyzed by microarray (Supplementary information, Figure S1A). The roles of several significantly deregulated miRNAs in gastric tumors have been described20,21. Here, we focused on miRNAs whose expression did not differ between tumor and para-tumor tissues. CAFs play an important role in tumorigenesis, and miRNAs can reprogram NFs into CAFs22. The conversion of NFs to CAFs occurs through a mechanism similar to that of fibrosis31, in which miRNAs, such as miR-21 and miR-29, are indispensable32,33. To identify CAF-specific deregulated miRNAs, we searched published datasets for miRNAs deregulated in CAFs in breast cancer27 and cardiomyocyte fibrosis34, and compared them with miRNAs unchanged in our array data (Supplementary information, Figure S1B). The three groups shared one common miR, miR-149, which is dramatically downregulated in fibrosis and CAFs.

A search of target predicting tools (TargetScan, Microrna.org, PicTar and RNAhybrid)35 for potential target genes of miR-149 identified IL-6, which is CAF specific and associated with tumorigenesis16,17,18 (Figure 1A). To validate the predicted role of IL-6 as a target gene of miR-149, wild-type and mutant 3′-UTR fragments of IL-6 were cloned into the pMIR-REPORT luciferase vector (Supplementary information, Figure S2). The constructs were co-transfected with miR-149 mimics or control into fibroblasts followed by luciferase reporter assay. The relative luciferase activity of pMIR/IL-6-3′-UTRwt, but not that of pMIR/IL-6-3′-UTRmut, was significantly suppressed by miR-149 mimics. By contrast, anti-miR-149 enhanced the relative luciferase activity of pMIR/IL-6-3′-UTRwt, but not that of pMIR/IL-6-3′-UTRmut (Figure 1B).

Figure 1
figure 1

miR-149 targets IL-6 in gastric stroma fibroblasts and regulates FAP expression via IL-6. (A) Schematic graph of 3′-UTR of IL-6 with the putative binding sites of miR-149. The minimum free energy (mfe) required for RNA hybridization was predicted by RNAhybrid software (mfe: −19.5 kcal/mol). (B) Effect of miR-149 mimics and miR-149 inhibitor on IL-6 expression. Luciferase activity in fibroblasts co-transfected with reporter vector with wild-type or mutant IL-6 3′-UTR and miR-149 mimics or inhibitors. (C) The miR-149 expression levels in 5 CAFs and 5 NFs established from gastric tumor tissues and matched para-tumor tissues were quantified by qRT-PCR. (D) Concentrations of IL-6 in the media of cultured CAF and NF cell lines were analyzed by ELISA. (E) The levels of miR-149 expression correlate inversely with IL-6 expression in CAFs and NFs. (F) The FAP expression levels in CAF or NF transfected with controls, miR-149 or anti-miR-149 (CAFNC, CAFmiR-149, NFanti-miR-149 and NFanti-NC, respectively) as analyzed by flow cytometry. (G) The relative FAP mRNA levels in CAFNC, CAFmiR-149, NFanti-miR-149 and NFanti-NC were detected by qRT-PCR. (H) Concentration of IL-6 in the media of cultured CAFNC, CAFmiR-149, NFanti-miR-149 and NFanti-NC. (I, J) Relative FAP levels in CAFmiR-149and NFanti-miR-149 in the presence or absence of IL-6 or IL-6 Ab were analyzed by flow cytometry (I) and qRT-PCR (J). All data represent means ± SD of three independent experiments. *P < 0.05, **P < 0.01, student's t-test.

Full size image

Based on the results indicating that miR-149 targets IL-6 mRNA, we investigated the relationship between miR-149 and IL-6 and their role in GC. To confirm the downregulation of miR-149 in CAFs, five primary CAF and NF cell lines were established from gastric malignant tissues and matched normal tissues, respectively. Quantitative analysis showed that miR-149 expression levels were markedly lower in GC CAFs than in NFs (Figure 1C).

The secretion of IL-6 was compared between CAFs and NFs, which showed higher levels of IL-6 secretion from CAFs than from NFs (Figure 1D). Moreover, IL-6 secretion levels were inversely correlated with miR-149 levels (Figure 1E). Taken together, these results suggest that miR-149 targets IL-6 mRNA and inhibits IL-6 production in CAFs.

miR-149 inhibits the activation of fibroblasts by reducing IL-6 expression

Fibroblast-derived IL-6 is critical for CAF function16,17, which prompted us to examine whether miR-149 modulates CAFs through IL-6. We used the cell surface marker fibroblast activation protein (FAP) to evaluate the activation of fibroblasts. After confirming that FAP expression levels were higher in CAFs than in NFs by flow cytometry and mRNA quantification (Supplementary information, Figure S3A and S3B), the effect of miR-149 on FAP expression was examined by introducing miR-149 mimics and miR-149 inhibitors into CAFs and NFs, respectively. FAP expression was significantly downregulated by miR-149 mimics in CAFs and upregulated by miR-149 inhibitors in NFs (Figure 1F and 1G) similar to IL-6 secretion (Figure 1H). Addition of IL-6 or an IL-6 neutralizing antibody reversed the downregulation of FAP expression by miR-149 mimics and its upregulation by miR-149 inhibitors, respectively (Figure 1I and 1J). These results strongly support the notion that miR-149 plays a role in maintaining NFs and repressing the function of CAFs via the regulation of IL-6 expression.

miR-149 is critical for the tumor-promoting ability of fibroblasts

CAFs promote cancer cell proliferation, migration and invasion1,31. To determine whether miR-149 regulates the tumor-promoting ability of fibroblasts, we investigated the effects of conditioned media from CAFs or NFs with manipulated levels of miR-149 on GC cells (Figure 2A). As shown in Figure 2B-2E, miR-149 mimics significantly suppressed the stimulatory effect of CAFs on GC cell proliferation, colony forming ability, migration and invasion; conversely, miR-149 inhibitors conferred NFs the enhancing effects.

Figure 2
figure 2

miR-149 influences the tumor-promoting ability of fibroblast. (A) Schematic chart of the evaluation of gastric cancer cell line SGC-7901 cultured in different conditioned medium (CM) as indicated. (B) Effect of miR-149 on GC cell proliferation measured by the WST assay. SGC-7901 cells were cultured in the conditioned mediums from CAF transfected with either miR-149 mimics or NC mimics, or NF transfected with either anti-miR-149 or anti-NC at a final concentration of 100 nM. (C) Effect of miR-149 on GC cell proliferation evaluated by colony formation assay. SGC-7901 cells were cultured in the indicated conditioned media and the number of colonies was counted 14 days after seeding. (D) Representative images of migratory or invasive cells on the membrane (magnification 100×). (E) Average number of migratory or invasive cells counted from D. (F) Tumors derived from SGC-7901 cells mixed with different fibroblasts (CAF transfected with agomir-149 or agomir-NC, NF transfected with antagomir-149 or antagomir-NC) and injected subcutaneously into nude mice. Tumors were isolated 28 days after injection. (G) Growth kinetics of tumors in nude mice. Tumor diameters were measured and the volumes were calculated every 7 days. (H) Average weight of tumors from nude mice. (I) Representative images of tumor samples stained with Ki-67 antibody from nude mice (magnification 200×). (J) Relative proliferation index from I. All data are means ± SD of three independent experiments. *P < 0.05, **P < 0.01, student's t-test.

Full size image

To further examine the role of miR-149 in vivo, nude mice were inoculated with SGC-7901 GC cells together with CAFs or NFs transfected with agomir-149 or antagomir-149 or their respective controls. As shown in Figure 2F-2H, miR-149 mimics significantly inhibited tumor growth and tumor burden in the group co-injected with GC cells and CAFs, whereas miR-149 inhibitors significantly increased tumor growth and tumor burden in the group co-injected with GC cells and NFs. The proliferation status of cells in tumors was verified by analyzing the proliferation marker Ki-67. The proliferation rate of cells positive for Ki-67 was dramatically reduced by miR-149, whereas it was substantially enhanced by the miR-149 inhibitor (Figure 2I and 2J). Taken together, our results indicate that miR-149 is an important negative regulator of the pro-tumorigenesis activity of CAFs both in vitro and in vivo.

CAFs enhance epithelial-to-mesenchymal transition and stem-like properties of GC cells in a miR-149-IL-6-dependent manner

Since CAFs stimulate epithelial-to-mesenchymal transition (EMT), which confers metastatic and self-renewal potential to cancer cells36, and stem cell compartmentation during tumorigenesis37, we postulated that miR-149 levels in CAFs might influence the EMT of GC cells. To prove this hypothesis, GC cells were treated with conditioned media from CAFs and NFs, and the expression of the epithelial cell marker E-cadherin and the mesenchymal cell markers N-cadherin, Vimentin and Snail was analyzed in GC cells. Conditioned medium from CAFs, but not that from NFs, induced the change of EMT markers at the mRNA level(Figure 3A). An IL-6 neutralizing antibody blocked the CAF conditioned medium-induced the change of expression of EMT markers, whereas IL-6 conferred the NF conditioned medium the ability to induce EMT, suggesting that IL-6 is crucial to this process (Figure 3A). Next, we grew SGC-7901 cells in different conditioned media from CAFs transfected with either miR-149 mimics or control and examined the expression of EMT markers at both mRNA and protein levels. miR-149 suppressed the CAF-induced EMT of GC cells, and the suppression was rescued by IL-6 (Figure 3B and 3C). The results of the culture of SGC-7901 cells using conditioned media from NFs transfected with either miR-149 inhibitors or control further supported the role of miR-149 and IL-6 in EMT induced by CAFs (Figure 3D and 3E). Furthermore, these results were confirmed by immunofluorescence detection of EMT markers in SGC-7901 cells cultured in different conditioned media (Figure 3F and 3G).

Figure 3
figure 3

miR-149 inhibits the EMT of GC cells through IL-6. (A) Relative mRNA levels of EMT markers (E-cadherin, N-cadherin, Vimentin and Snail) in SGC-7901 cultured in the CM from CAF or NF in the presence or absence of IL-6 or IL-6 antibody as indicated were quantified by qRT-PCR. (B) Relative mRNA levels of EMT markers in SGC-7901 cultured in the CM from CAFNC or CAFmiR-149 in the presence or absence of IL-6 as indicated were quantified by qRT-PCR. (C) Protein levels of EMT markers in SGC-7901 cultured as in B were analyzed by western blotting. (D) Relative mRNA expression of EMT markers in SGC-7901 cultured in the CM from NFanti-NC or NFanti-miR-149in the presence or absence of IL-6 antibody as indicated were quantified by qRT-PCR. (E) Protein levels of EMT markers in SGC-7901 cultured as in D were detected by western blotting. (F, G) Immunofluorescence analyses were carried out in SGC7901 cultured in the CM as indicated in B or D with antibodies for E-cadherin (green), Vimentin (red), and DAPI (blue), (magnification 400×). All data are means ± SD of three independent experiments. *P < 0.05, **P < 0.01, student's t-test.

Full size image

Cancer stem cells (CSCs) are considered as TICs that drive tumor formation, and the role of tumor stromal IL-6 in promoting the “stemness” of cancer cells has been established38. GC TICs are enriched in aldehyde dehydrogenase high39 and side population (SP) cells40. We therefore investigated whether CAFs could increase the CSC-enriched population. We cultured GC cells with different conditioned media from CAFs and NFs transfected with miR-149 mimics, miR-149 inhibitors or their respective controls, and then quantified the ALDH-positive or SP cells in GC cells by flow cytometry. The proportions of cells with high levels of ALDHhigh and SP cells were modulated by different conditioned media. As shown in Figure 4A, introduction of miR-149 into CAFs reduced the ALDHhigh proportion of cancer cells cultured in conditioned medium from CAFs, and this was rescued by IL-6 supplementation. By contrast, anti-miR-149 boosted the ALDHhigh proportion of cancer cells cultured in conditioned medium from NFs, and this was reversed by IL-6 neutralizing antibody (Figure 4C). A similar effect on the distribution of SP cancer cells was observed (Figure 4B and 4D). We also examined the effects of conditioned media from fibroblasts with manipulated levels of miR-149 on sphere formation of GC cells. As shown in Figure 4E and 4F, high miR-149 levels in CAFs had a negative effect on GC cell sphere formation and growth, and this was rescued by IL-6 supplementation, whereas the presence of anti-miR-149 in NFs had an enhancing effect that could be suppressed by an IL-6 neutralizing antibody. These results indicate that miR-149 in CAFs or NFs has an indirect effect on the proliferation of CSC or CSC-like cells.

Figure 4
figure 4

miR-149 inhibits the stem-like characteristics of GC cells through IL-6. (A) Aldeflour assay by flow cytometry of ALDHhigh cells in SGC-7901 cells cultured in CM from CAFNC or CAFmiR-149 in the presence or absence of IL-6 as indicated. Representative flow cytometric analysis was shown. (B) Side population (SP) cells in SGC-7901 cells cultured as in A were measured by flow cytometry. SGC-7901 cells were stained with Hoechst 33342 dye in the presence (upper panel) or absence (bottom panel) of verapamil. Representative flow cytometric analysis was shown. (C) ALDHhigh cells in SGC-7901 cells cultured in the CM from NFanti-NC or NFanti-miR-149 in the presence or absence of IL-6 antibody as indicated by flow cytometric analysis. (D) SP cells in SGC-7901 cells cultured in the CM as in C and analyzed as in B. (E) Representative images of spheroid body from SGC-7901 cells cultured as in A or C (magnification 100×). (F) The rates of sphere formation were quantified. All data are means ± SD of three independent experiments. *P < 0.05, **P < 0.01, student's t-test.

Full size image

PGE2 induces the epigenetic silencing of miR-149 in fibroblasts

DNA hypermethylation of many miRNAs is correlated with reduced gene expression in tumor tissues41. We identified an H3K27ac-enriched area42 in a GC-rich region within −2 500 to 1 500 bp upstream of the miR-149 transcriptional start site using the MethPrimer-Tools and Databases (http://www.urogene.org/methprimer) and the ENCODE project data on the UCSC genome browser (http://goo.gl/6tDJ3v, Supplementary information, Figure S4A). We hypothesized that the downregulation of miR-149 in CAFs could be associated with hypermethylation of this enhancer region. Methylation specific PCR (MSP) and bisulfite sequencing PCR (BSP) primers overriding this region were designed to detect the methylation status in CAFs and NFs. MSP analysis showed that this region was hypermethylated in CAFs compared to NFs (Supplementary information, Figure S4B). BSP analysis results showed that the average rate of CpG sites methylation was significantly higher in CAFs (51.4% ± 8.82%) than in NFs (21.8% ± 6.11%; Figure 5A and Supplementary information, Figure S4C). These results suggest that hypermethylation of the miR-149 promoter is associated with the CAF phenotype and may contribute to the activation of fibroblasts.

Figure 5
figure 5

H. pylori-induced COX-2/PGE2 activation enhances epigenetic silencing of miR-149. (A) Methylation levels in the CpG sites rich region of miR-149 are higher in CAFs. Bisulfite sequencing was performed in 10 clones for each cell line. Open circle: unmethylated CpG site; filled circle: methylated CpG site. The ratios of methylation were summarized in the right panel. (B) miR-149 mRNA levels in CAF or NF treated or untreated with PGE2 or 5-Aza were analyzed by qRT-PCR. 5-Aza derepresses the PGE2-induced miR-149 repression. (C) Schematic graph of the 3′-UTR of EP2 with the putative binding site of miR-149. The mfe predicted by RNAhybrid is −16.3 kcal/mol. (D) Relative luciferase activity in fibroblasts co-transfected with reporter vector with wild-type or mutant 3′-UTR of EP2 and miR-149 mimics, inhibitors or controls as indicated. (E) Sketch map of PGE2 activating IL-6 through EP2 by eliminating miR-149, a common inhibitory factor of IL-6 and EP2. miR-149 targets IL-6 and EP2 for suppression in NFs (left panel). Through binding to EP2, PGE2 induces the hypermethylation and suppression of miR-149 that leads to the derepression of IL-6 and EP2 (right panel). (F) The PGE2 produced by SGC-7901 and GES-1 cells with or without H. pylori infection was measured by ELISA. (G, H) COX-2 mRNA and protein levels in SGC-7901 and GES-1 cells as in F. mRNA and protein levels were analyzed by qRT-PCR and western blotting, respectively. (I) The PGE2 produced by SGC-7901 and GES-1 cells with or without H. pylori infection and in the presence or absence of NS-398 (50 μmol/L, an inhibitor of COX-2) was measured by ELISA. (J, K) COX-2 mRNA and protein levels in SGC-7901 and GES-1 cells as in I. mRNA and protein levels were analyzed by qRT-PCR and western blotting, respectively. All data are means ± SD of three independent experiments. *P < 0.05, **P < 0.01, student's t-test.

Full size image

PGE2 signaling is critical for IL-6 production in the tumor stroma and plays important roles in cancer and inflammation7,13. In addition, PGE2 promotes global methylation in tumor cells43. We postulated that PGE2 might be the signal triggering miR-149 methylation in fibroblasts. To test this possibility, we analyzed miR-149 expression in NFs in response to PGE2 treatment and confirmed that PGE2 induced DNA methylation of miR-149 (Supplementary information, Figure S5) and downregulated miR-149 expression, while an inhibitor of DNA methylation, 5-Aza, abolished this effect (Figure 5B).

PGE2 receptor, PTGER2, is also a potential target of miR-149

We found that the PGE2 receptor, prostaglandin E receptor 2 (PTGER2, subtype EP2), also contains a seed match for miR-149 on its 3′-UTR (Figure 5C), and PGE2 can induce IL-6 expression in fibroblasts through EP244. We therefore, cloned the wild-type or mutant 3′-UTR fragment of EP2 into the pMIR-REPORT luciferase vector (Supplementary information, Figure S6), and transfected the constructs together with miR-149 mimics or miR-149 inhibitor into fibroblasts, followed by luciferase reporter assay. The relative luciferase activity of pMIR/EP2-3′-UTRwt, but not that of pMIR/EP2-3′-UTRmut, was significantly suppressed by miR-149 mimics compared to NC mimics. By contrast, miR-149 inhibitors enhanced the activity of wild-type but not that of the mutant reporter (Figure 5D). This indicates that EP2 could be another potential target of miR-149. We therefore compared the EP2 expression levels in CAFs and NFs and found that it was higher in CAFs than in NFs (Supplementary information, Figure S7A) and the EP2 expression might be regulated by miR-149 (Supplementary information, Figure S7B). Furthermore, we also noticed higher EP2 expression level in tumor stroma compared to para-tumor normal stroma in the clinical specimens (Supplementary information, Figure S8).

Taken together, our data suggest a pro-tumor signaling mechanism by which stromal PGE2 induces IL-6 expression through EP2 by suppressing a shared inhibitory factor, miR-149 (Figure 5E). miR-149 targets IL-6 and EP2, downregulating their expression in fibroblasts. PGE2 binds to EP2 and induces the hypermethylation and suppression of miR-149, reversing the downregulation of IL-6 and EP2 expression.

H. pylori infection activates COX-2/PGE2 signaling in gastric epithelial cells

H. pylori infection may cause gastritis with a chronically inflamed stroma and the infection is recognized as the major cause of stomach cancer45. In addition, the COX-2/PGE2 pathway is activated upon H. pylori infection46,47. Thus, we hypothesized that H. pylori infection might elicit the malignant transformation of gastric epithelial cells, leading to a pro-inflammatory response and consequently the dysfunction of miRNAs in fibroblasts, which in turn promotes tumor development. We infected cells with H. pylori and found that it induced PGE2 secretion more significantly in the gastric epithelial cell line GES-1 cells than in SGC-7901 cells (Figure 5F). Similarly, H. pylori infection induced COX-2 expression at both protein and mRNA levels more significantly in GES-1 (Figure 5G and 5H). We used an inhibitor of COX-2, NS-398, and confirmed that it could block PGE2 secretion in both cell lines (Figure 5I). In addition, COX-2 expression at mRNA and protein levels was also inhibited by NS-398 in both GES-1 and SGC-7901 cells with or without H. pylori infection (Figure 5J and 5K). These results indicate that H. pylori infection can activate the COX-2/PGE2 pathway to a greater extent in gastric epithelial cells than in GC cells.

H. pylori infection induces hypermethylation and downregulation of mmu-miR-149 in murine fibroblasts in vivo

H. pylori infection has been shown to be able to induce methylation-mediated silencing of miRNAs, which could be risk factors for GC48,49. Our above results raised the possibility that H. pylori infection might activate COX-2/PGE2 signaling which in turn suppresses miR-149 expression through aberrant DNA methylation. To recapitulate human GC development in an animal model and to investigate whether H. pylori infection affects the hypermethylation and expression of the miR-149 gene in vivo, we performed intragastric administration of H. pylori in mice. The development of gastritis and GC in mice was verified by hematoxylin and eosin staining (Supplementary information, Figure S9B). Since the mouse IL-6 gene also contains a seed match for Mir-149 (mmu-miR-149; Figure 6A), we performed a similar assay to that described in Figure 1 to confirm that miR-149 also targets IL-6 in the mouse (Figure 6B and Supplementary information, Figure S9A). Fibroblasts from the stomach of control- and H. pylori-infected mice were isolated and cultured (Figure 6C). In those cells, Fap expression levels were higher in mouse CAFs (mCAFs) and mouse gastritis-associated fibroblasts (mGAFs) than in mouse NFs (mNFs; Supplementary information, Figure S9C). In addition, miR-149 levels were lower and IL-6 secretion levels were higher in both mCAFs and mGAFs than in mNFs (Figure 6D and 6E).

Figure 6
figure 6

miR-149 is epigenetically silenced in fibroblasts from H. pylori-infected mouse stomach. (A) Schematic graph of 3′-UTR of IL-6 with the putative binding sites of mmu-miR-149. The mfe predicted by RNAhybrid is −16.5 kcal/mol. (B) Luciferase activity in cells co-transfected with reporter vector with wild-type or mutant IL-6 3′-UTR and mmu-miR-149 mimics or inhibitors as indicated. (C) Scheme for mice infected with H. pylori. Mouse normal fibroblasts (mNFs), mouse gastritis-associated fibroblasts (mGAFs) and mouse cancer-associated fibroblasts (mCAFs), were isolated from the control- and H. pylori-infected stomach of mice and cultured. (D) The mmu-miR-149 expression levels in mNFs, mGAFs and mCAFs were determined by qRT-PCR. (E) IL-6 secreted into media by mNFs, mGAFs and mCAFs was measured by ELISA. (F, G) Methylation levels in the CpG sites of miR-149 are higher in mCAF and mGAF than in mNF. Bisulfite sequencing was performed in 5 clones from each cell line. The rates of methylation were summarized in G. All data are means ± SD of three independent experiments. *P < 0.05, **P < 0.01, student's t-test.

Full size image

Subsequently, we analyzed the mouse ENCODE data and identified a CpG sites rich region between −2 200 and −1 750 bp upstream of the mmu-miR-149 that is conserved in humans (http://goo.gl/mqQu79, Supplementary information, Figure S9D). BSP analysis showed that the DNA methylation levels around that location were significantly higher in mCAFs (45% ± 3.7%) and mGAFs (47.3% 2.1%) than in mNFs (8.3% ± 2.9%; Figure 6F and 6G, Supplementary information, Figure S9E). These results suggest that H. pylori infection is likely to promote the pro-tumor properties of stromal fibroblasts in vivo by silencing mmu-mir-149 and stimulating the production of IL-6.

Discussion

Although CAFs are not malignant themselves, they play a key role in the crosstalk between tumor cells and their microenvironment to support tumor growth1,37. In the present study, we identified miR-149 as a critical factor for the transformation of NFs into CAFs in GC. We showed that H. pylori infection activates the COX-2/PGE2 pathway, leading to hypermethylation of the miR-149 promoter and repression of miR-149 in human and mouse fibroblasts in vitro and in vivo. Through the regulation of IL-6 expression, suppression of miR-149 promotes the pro-tumor activity of CAFs by inducing EMT and stem-like traits in GC cells.

miRNAs have been implicated in the activation of fibroblasts, and they have been proposed as an emerging bridge linking H. pylori infection to GC25,50. The up- and downregulation of miR-149 expression is associated with several types of cancer51,52,53,54,55. The existence of miRNA profiles specific to different cancers and even to certain subpopulations of cells in the same cancer is possible. We showed that miR-149 is significantly downregulated in CAFs from human GC tissues and H. pylori-infected mice (Figures 1 and 6), similar to the levels reported in breast CAFs27. This suggests the existence of a common mechanism underlying the regulation of CAFs in different cancers, which is further supported by the enhancement of IL-6 production by CAFs demonstrated in the present study (Figure 2) and in other reports17,18,19,20,21.

The cytokine IL-6 is required for the development of inflammation-associated cancers16,56. Increasing evidence strongly indicates that IL-6 is a common inflammatory factor secreted by CAFs from different sources3,15,16,17,18,56,57,58. A recent study showed that CAFs are activated in incipient neoplasia and contribute to tumor-promoting inflammation in an NF-κB-dependent manner3. In this process, NFκB acts as an essential positive regulator of IL-6. In the present study, we showed that miR-149 is downregulated in CAFs and acts as a negative regulator of IL-6 expression (Figure 1). Blocking the IL-6 signal using a neutralizing antibody has a similar negative effect to miR-149 upregulation on the activation of fibroblasts and the tumor-promoting ability of CAFs, whereas the inhibitory effect of miR-149 can be reversed by supplementation of IL-6 (Figures 1 and 2). Our results indicate that miR-149 plays an important role in the activation of NFs and the maintenance of CAFs via the regulation of IL-6 expression, suggesting that epigenetic changes may guide the co-evolution of malignant cells and CAFs. During the preparation of this manuscript, miR-149 was also found to be downregulated in osteoarthritis chondrocytes and it is correlated with increased expression of inflammatory cytokines including IL-659. Moreover, IL-6 is an important regulator of miRNA expression60. Therefore, IL-6 might be a pivot of miR-mediated inflammation in tumor microenvironment.

Tumor cells affect the secretion of IL-6 by stromal cells in the tumor microenvironment through mesenchymal stem cells (MSCs)-derived PGE27. The MSC-derived PGE2 operates in an autocrine manner to induce the production of IL-6, which together with PGE2 acts in a paracrine manner on tumor cells to induce EMT and increase the CSC population7. In the present study, PGE2 production was induced upon H. pylori infection in gastric epithelial cells (Figure 5F-5I). The nature of H. pylori-positive GC suggests that tumor cells originate from malignant transformed gastric epithelial cells, and H. pylori infection has been shown to induce the generation of CAFs in mice15. Therefore, our results support the notion that neoplastic PGE2 could be a trigger or early signal for the initiation of the transformation of NFs to CAFs10.

DNA hypermethylation of many miRNAs is correlated with reduced gene expression in tumor tissues61. DNA hypermethylation of miR-149 was detected in colorectal cancer62. In the present study, we showed the hypermethylation of a GC-rich region upstream of the miR-149 transcriptional start site in CAFs from GC tissues, but not in NFs (Figure 5). Coincidently, PGE2 can induce the hypermethylation of miR-149 in fibroblasts (Figure S5)43. Surprisingly, we also found that the PGE2 receptor, EP2, is a target of miR-149. This indicates the existence of a positive feedback loop in CAFs in which PGE2 triggers the hypermethylation of miR-149. The consequent silencing of miR-149 expression stimulates PGE2 signaling through the derepression of EP2. In the present study, the relationship between PGE2 and miR-149 methylation (Figure 7) strengthens the argument that epigenetic modification is essential for the generation and maintenance of the CAF phenotype30.

Figure 7
figure 7

Schematic graph depicting the involvement of miR-149 in the interaction between H. pylori-infected gastric epithelial cells and gastric cancer-associated fibroblasts. H. pylori infection induces the secretion of PGE2 by gastric epithelial cells. Stromal PGE2 silences miR-149 probably through the hypermethylation of miR-149. The silence of miR-149 removes the suppression of its target genes, IL-6 and EP2, leading to elevated level of IL-6 that acts on fibroblasts to stimulate the transformation of NF into CAF. Meanwhile, IL-6 induces more aggressive behaviors of malignant gastric epithelial cells.

Full size image

We investigated the role of H. pylori infection and miR-149 in a mouse model of gastritis and GC. Significantly lower miR-149 levels and correspondingly higher IL-6 secretion levels were detected in both mCAFs and mGAFs than in mNFs (Figure 6). Furthermore, we showed that H. pylori infection could activate the COX-2/PGE2 pathway in gastric epithelial cells. PGE2 secretion and COX-2 expression are not apparently induced by H. pylori infection in GC SGC-7901 cells, and can be suppressed by the COX-2 inhibitor NS-398. This could be explained by the constitutive activation of the COX-2/PGE2 signaling pathway in malignant epithelial cells.

Tumor development is dependent on the crosstalk between tumor cells and the tumor stroma. An understanding of how tumor cells communicate with the stroma will lead to the discovery of innovative avenues of therapeutic intervention. In the present study, we showed that miR-149 modulates PGE2 and IL-6 signaling in the crosstalk between tumor cells and the tumor stroma. These findings shed light on the mechanism of action of the PGE2-IL-6 activation loop and further highlight the significance of this loop in tumorigenesis. We propose that H. pylori infection-induced PGE2 secretion from epithelial cells or from tumor cells promotes the methylation of miR-149 and the subsequent induction of the expression of its target gene (IL-6) in stromal fibroblasts. The role of miRNAs as a sensor signal triggered upon H. pylori infection in the mucosa and epithelium has been proposed49. However, our results support a role for miR-149 in stromal cells as a potential amplifier signal triggered by H. pylori infection. Therefore, detection or intervention of miR-149 expression in the tumor stroma could be of prognostic or therapeutic value in GC.

Materials and Methods

Cell lines and primary cell isolation

GC cell line SGC-7901 and the immortalized gastric epithelial cell line GES-1 were purchased from Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, and were cultured at 37 °C with RPMI-1640 medium containing 10% fetal calf serum. Fibroblasts were isolated from five independent GC patients undergoing radical gastrostomy at the Department of Surgery, Ruijin Hospital, Shanghai Jiao Tong University. None of the patients received preoperative treatment. The primary fibroblast isolation and culture were described in our previous study63. The tumor tissues and non-tumor tissues were collected, and after cleaning and disinfection, the tissues were minced into organoids of ∼1 mm3 and seeded into petri dishes. These conditions produced a homogenous fibroblastic cell population after about 7 days of culture. The doubling time of fibroblasts is about 2-3 days. We harvested cells passaged for up to 10 population doublings for subsequent experiments in order to maintain the characters and the phenotype of primary cells.

Reagents

Recombinant IL-6 was purchased from PeproTech. FAP antibody was purchased from Santa Cruz Biotech; E-cadherin, N-cadherin, Vimentin, Snail and COX-2 antibodies were purchased from Abcam; EP2 antibody was purchased from BBI; Human IL-6 mAb and mouse IL-6 mAb were purchased from R&D Systems (Minneapolis, MN, USA).

RNA isolation and qPCR

Total RNA was isolated using Trizol regent (Invitrogen, Carlsbab, CA, USA) or mirVanaTM miR Isolation Kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. The mRNA expression level was measured by qPCR according to TaqMan Gene Expression Assays protocol (Applied Biosystems). The primers used were: FAP: 5′-AATGAGAGCACTCACACTGAAG-3′ and 5′-CCGATCAGGTGATAAGCCGTAAT-3′; E-cadherin: 5′-CCCACCACGTACAAGGGTC-3′ and 5′-CTGGGGTATTGGGGGCATC-3′; N-cadherin: 5′-CAACTTGCCAGAAAACTCCAGG-3′ and 5′-ATGAAACCGGGCTATCTGCTC-3′; Vimentin: 5′-AAAACACCCTGCAATCTTTCAGA-3′ and 5′-CACTTTGCGTTCAAGGTCAAGAC-3; Snail: 5′-CCCCAATCGGAAGCCTAACT-3′ and 5′-GGTCGTAGGGCTGCTGGAA-3′; COX-2: 5′-TAAGTGCGATTGTACCCGGAC-3′ and 5′-TTTGTAGCCATAGTCAGCATTGT-3′; Fap (Mus musculus): 5′-GGCTGGGGCTAAGAATCCG-3′ and 5′-GCATACTCGTTCACTGGACAC-3. The expression levels of miRNAs were measured by miRNAs qPCR Kit (GenePharma, Shanghai, China) according to the manufacturer's instructions.

Transfection of miR mimics

Oligonucleotides for hsa-miR-149 mimics (miR-149), negative control (miR-NC), hsa-miR-149 inhibitors (anti-miR-149), inhibitor negative control (anti-miR-NC) and mmu-miR-149 mimics, inhibitors or controls were purchased from GenePharma (Shanghai, China). Transfection of cells with oligonucleotides was performed with Lipofectamine 2000 (Invitrogen) at a final concentration of 100 nM. For the cells used in tumor xenograft mouse model, chemically modified mimic (agomir) or antisense (antagomir) oligonucleotides, which were also from GenePharma, for has-miR-149 was applied.

Luciferase activity assay

The fragment of wild-type (wt) 3′-UTR of IL-6 containing the putative miR-149-binding sites or mutant (mut) 3′-UTR of IL-6 were synthesized and cloned into the pMIR-Report luciferase vector (Ambion) containing Firefly, and they were named IL-6-3′-UTRwt Luc and IL-6-3′-UTRmutLuc. Cells were seeded into the 24-well plates 24 h before transfection. A total of 500 ng of IL-6-3′-UTRwt Luc or IL-6-3′-UTRmut Luc were transfected into each well together with 20 ng of pRL-TK vector (Promega) and 60 pmol of oligonucleotides. Cells were harvested 48 h later and Firefly and Renilla luciferase activities were measured by Dual-luciferase reporter assay (Promega). This method was also performed to verify that miR-149 targets the 3′-UTR of EP2 and mmu-mir-149 targets the 3′-UTR of IL-6 (Mus musculus).

ELISA

The levels of cytokines IL-6 and PGE2 in supernatants were detected by ELISA kit (R&D Systems) according to the instructions of manufacturer.

Cell proliferation assay

Cell proliferation was assessed by WST (water-soluble tetrazolium salt) assay using a Cell Counting Kit-8 (Dojindo, Japan) according to the manufacturer's instructions. Cells (2 × 103 cells/well) were seeded into 96-well plates. The cell proliferation was examined every 24 h for 5 days. The number of viable cells was assessed by measurement of the absorbance at 450 nm.

Soft agar colony formation assay

Cells were resuspended with 0.3% soft agar in RPMI1640 containing 10% FBS or conditioned medium and layered onto 0.6% solidified agar in RPMI1640 containing 10% FBS in 6-well plates (1 × 103 cells/well). The plates were incubated at 37 °C in humidified atmosphere of 5% CO2 for 2 weeks, the stained with 0.02% crystal violet for 1 h (Sigma, USA). The colonies were counted in 5 random fields under microscope. Colonies containing at least 50 cells were counted.

Cell migration and invasion assay

Migration of cells was performed using QCM 24-Well Colorimetric Migration Assay Kit (Millipore) according to the manufacturer's instructions. For the invasion assay, Cell Invasion Assay Kit (Millipore) was used. Cells (2 × 104) in 200 μl serum-free medium were added to the upper chambers and cultured for 48 h. Non-migrating or non-invading cells were removed with cotton swabs, cells that migrated or invaded to the bottom of the membrane were stained with the cell stain buffer and counted under microscope and photographed. Three independent experiments were performed for the same conditions.

Tumor xenograft model and tumorigenicity assay

SGC-7901 cells co-cultured with fibroblasts were subcutaneously injected into 4-week-old male nude mice (Institute of Zoology, Chinese Academy of Sciences). The mice were checked every 7 days, and the tumor nodules were measured with a caliper. Mice were euthanized 28 days after injection, the tumors were removed and weighed and tumor volume was evaluated. Tumor growth curves were calculated. All tumor grafts were excised, weighed, harvested, fixed and embedded. The Ki-67 antibody (Dako, dilution 1:100) was used to determine the expression of Ki-67 by immunohistochemistry. The proliferative index score was measured by the rate of Ki-67-positive cells.

SP analysis

The cells were removed, pelleted by centrifugation, washed with PBS (phosphate-buffered saline), and suspended at 1 × 106 cells/ml in PBS supplemented with 4% FBS and incubated at 37 °C for 60 min in the dark with the DNA binding dye, Hoechest 33342 (5 μg/ml, Sigma, USA), either alone or in the presence of verapamil (500 μM, Sigma, USA). After incubation, 1 μg/ml propidium iodide was added and then the cell suspension was filtered through a 40 μm cell strainer to obtain single suspension cells, and the cells were kept at 4 °C in the dark before FACS. Analyses and sorting were performed using BD FACS Aria II (BD, New Jersey, USA).

Examination of ALDH activity by Aldeflour reagent

Aldeflour assay kit (Stemcell Technologies, Vancouver, Canada) was used to detect the ALDH activity of cells. Cancer cells (1 × 106 cells/ml) were suspended in Aldeflour assay buffer and incubated with the ALDH substrate (BODIPY-aminoacetaldehyde, BAAA) for 45 min at 37 °C. As a negative control, 1.5 mM diethylaminobenzaldehyde (DEAB), an ALDH specific inhibitor was added. FACS Calibur flow cytometer was used for cell analysis.

Confocal immunofluorescence

The PathScan EMT Duplex IF Kit (Cell Signaling Technology) was adopted to detect the expression of the EMT markers: E-cadherin (Alexa Flour488) and Vimentin (Alexa Fluor555). Cells were fixed in 4% formaldehyde buffer for 15 min at room temperature. Cells were then blocked in blocking buffer for 60 min and incubated with the primary antibodies overnight at 4 °C. After being rinsed three times with PBS, cells were incubated with detection cocktail buffer at room temperature in the dark, covered with slides with Antifade Reagent (with DAPI) and subjected to confocal microscopy analyses (LSM510, Carl Zeiss, Jena, Germany).

DNA isolation and methylation analysis

Genomic DNA from tissue samples was purified using DNAzol (Invitrogen). Sodium bisulfite conversion was conducted using the Qiagen Epitect Bisulfite Kit. The methylation status was examined by MSP or BSP. The MSP products were confirmed by agarose gel electrophoresis. The BSP products were subcloned into pMD19-T Vector (Takara) according to the manufacturer's instructions.

Mice and sample preparation

Following previously reported procedures64,65,66, BALB/c mice (Institute of Zoology, Chinese Academy of Sciences) were divided into three groups: Group 1 and Group 2 of 20 mice each as the H. pylori-infected groups, Group 3 as the normal control group. Mice in Group 1 were inoculated with 0.1 ml suspension of H. pylori SS1 strain (1.0 × 109 CFU/ml) intragastrically every other day from age of 5 weeks, and were given distilled drinking water. Group 2 were inoculated with H. pylori as Group 1, and were treated with N-methyl-N-nitrosourea (MNU, Sigma-Aldrich) mixed in drinking water (60 ppm, freshly prepared 3 times per week) for 10 weeks; Group 3 were given distilled water, but not treated with MNU and H. pylori. All mice were sacrificed 40 weeks after treatment.

The mouse stomach was resected and cut open along the greater curvature. Half of them were harvested, fixed and embedded. H&E staining was performed to detect the H. pylori colonization and tumor occurrence. The other half was used for primary fibroblast culture.

Statistical analysis

Results were summarized as means ± SD. Student's t-test and one-way analysis of variance were used to analyze the data and the significance level was set at P < 0.05.