MicroRNA-21 induces loss of 15-hydroxyprostaglandin dehydrogenase in early gastric tubular adenocarcinoma

15-hydroxyprostaglandin dehydrogenase (15-PGDH), the rate-limiting enzyme in prostaglandin E2 degradation, is decreased in gastric cancers and microRNA (miR)-21 is one of the regulators. We investigated the expression and regulation of 15-PGDH in eary gastric carcinogenesis utilizing endoscopic submucosal dissection (ESD) and gastric cancer cell lines. Expression of 15-PGDH and cyclooxygenase-2 as well as the promoter methylation of 15-PGDH were evaluted. CRISPR, miR-21 transfection, proliferation and apoptosis assays were also done. We observed significant decreases in 15-PGDH expression but no promoter methylation was detected in any ESDs. 15-PGDH suppression by CRISPR induced enhanced growth kinetics. miR-21, which was detected in high level in gastric tumors from the TGCA data, caused increased proliferation, decreased apoptosis. miR-21 overexpression was confirmed with CISH and RT-PCR in the ESDs. Loss of 15-PGDH occurs at the very early stage of gastric adenocarcinoma by miR-21. H. pylori infection may affect miR-21 up regulation. Maintaining 15-PGDH enzyme activity could be a new strategic measure in preventing gastric cancer especially tubular adenocarcinoma.

In addition, we attempted to elucidate how 15-PGDH was regulated in early carcinogenesis. A few possible inhibitory mechanisms were reported previously, including promoter methylation in gastric cancers 12 and microRNA(miRNA) in cholangiocarcinomas 17 . In this study, methylation status was evaluated using methylation-specific polymerase chain reaction (MS-PCR), as well as pyrosequencing. Using the public data from The Cancer Genome Atlas (TCGA) 18 , we found a negative correlation between microRNA-21 (miR-21) and 15-PDGH. Along with assessing the expression profile of miR-21, functional analysis was conducted after transfecting miR-21 to gastric cancer cell lines.
Herein, we investigated the expression pattern and regulation of 15-PGDH using gastric cancer cell lines, as well as gastric ESD specimens.

15-PGDH expression was significantly decreased in early gastric tubular adenocarcinoma.
The expression of 15-PGDH, COX-2, and PGE 2 was evaluated via IHC using FFPE sections from 30 ESD specimens. Because these retrospectively collected specimens were very early lesions with tubular adenoma in the background, normal/atrophic and intestinal metaplastic mucosa were always present in the vicinity. The carcinoma and tubular adenoma area of these specimens consistently exhibited very low or absence of 15-PGDH, whereas COX-2 was intermediately or strongly expressed (Fig. 1a). The IHC staining scores of 15-PGDH were significantly lower than that of COX-2 (paired t-test, P < 0.0001).
We then compared the expression patterns of these proteins in adjacent normal/atrophic or intestinal metaplastic mucosa (Fig. 1b). 15-PGDH, which was well expressed in normal/atrophic and metaplastic foveolar epithelial cells, disappeared in dysplastic cells (tubular adenoma and/or adenocarcinoma). COX-2, on the other hand, was not expressed in normal/atrophic epithelial cells, but started to show positivity in metaplastic cells and revealed higher expression in dysplastic cells. PGE 2 exhibited a similar pattern as that of COX-2. The loss of 15-PGDH and high COX-2 and PGE 2 persisted in the carcinoma area ( Supplementary Fig. 1). This was further confirmed with snap frozen tumor/normal samples collected from 16 patients. The samples were collected with endoscopic forceps before performing ESD. All were confirmed to have early tubular adenocarcinomas in FFPE sections. 15-PGDH WB was lost or exhibited significantly decreased expression in all tumor tissues compared to the normal samples (Fig. 1c). Decreased (hydroxyprostaglandin dehydrogenase 15 (15-PGDH, HPGD) mRNA expression in tumor samples relative to that of normal samples was obvious (Fig. 1d), whereas prostaglandin-endoperoxide synthase 2 (COX-2, PTGS2) mRNA failed to present a difference between tumor and normal samples (Fig. 1e). On the other hand, the PGE 2 level was increased in the tumor samples (Fig. 1f).

15-PGDH and COX-2 expression pattern in gastric cancer cell lines.
The protein and mRNA expression of 15-PGDH and COX-2 was screened using various gastric cancer cell lines. Protein expression by WB (Fig. 2a) showed good correlation with mRNA expression by qRT-PCR (Fig. 2b,c). Based on these results, we could categorize gastric cancer cell lines in terms of 15-PGDH and COX-2 expression (Supplementary Table 1 Table 1 and Fig. 2d). This discrepancy with 15-PGDH expression and methylation was seen in patient samples as well.
None of the patient samples had enough methylation to account for the loss of 15-PGDH protein expression (Fig. 2e). This was also confirmed with MS-PCR (data not shown). Two cells lines (TMK-1 and SNU-601) were selected to determine if the de-methylation of 15-PGDH with 5-aza-dC treatment induced protein expression. Up to 50 µM of 5-aza-dC was used, but it had little or no effect on the promoter site de-methylation or protein expression ( Supplementary Fig. 2ab), although DNMT1 expression was decreased.

miR-21 is upregulated in gastric cancers.
In search for other possible mechanisms for 15-PGDH expression regulation, we determined the correlation of 194 miRNAs with 15-PGDH from the TCGA gastric cancer public database 18 . Three miRNAs (miR-181c, miR-769, and miR-21) were reversely correlated with the 15-PGDH (HPGD) gene (Fig. 4a). We considered miR-21 a good candidate for further study because miR-21 has been reported to have a binding site in 3′-UTR of 15-PGDH mRNA 17 .  (Fig. 5a). The mRNA expression alteration after miR-21 transfection correlated well with the protein, which was confirmed by qRT-PCR (Fig. 5b).
The effect miR-21 on cell proliferation and apoptosis was further studied using these six cell lines. The proliferation assay showed a significant increase in cell growth after 2 days of transfection in AGS, MKN-28, KATO III, and MKN-45 (Fig. 5c). In the apoptosis assay, KATO III and MKN-45 showed decreased early and late apoptosis (Fig. 5d). Colony formation assay using miR-21 transfected cells also confirmed increase cell growth ( Supplementary Fig. 4).
miR-21 CISH in patient tissue also confirmed that miR-21 was increased in tumor areas (Fig. 5e). The CISH score was higher in the tumor area than the adjacent normal area (2.3 ± 0.2 vs. 0.75 ± 0.2, paired t-test, P < 0.0001) and relative miR-21 expression measured by qRT-PCR was significantly higher in the tumor area  Fig. 5f). We performed miR-21 CISH on another patient set (retrospectively collected 30 cases of ESD). The CISH score was also higher in the tumor area (1.87 ± 1.07 vs. 0.33 ± 0.48, P < 0.0001, Supplementary Fig. 5).  Fig. 6A,B). This was done using the cohort from the previous study where we demonstrated that 15-PGDH expression was inhibited by H. pylori infection 10 . When H. pylori was infected to AGS, miR-21 mRNA expression increased (P = 0.008, Supplementary  Fig. 6C). Subgroup analysis with 16 ESD patients (13 H. pylori infected subjects) revealed that miR-21 CISH scores were higher in tumor tissues compared to normal tissues from H. pylori-infected patients, but not in patients with no H. pylori infection (P < 0.0001, Supplementary Fig. 6D). In addition, miR-21 was higher in the tumor tissues of H. pylori-infected patients than the tumor tissues of uninfected patients (P = 0.012, Supplementary Fig. 6D).

Discussion
In this study, we demonstrated that 15-PGDH expression is lost during early carcinogenesis of gastric cancer. Unlike diffuse-type gastric carcinomas, tubular adenocarcinomas or intestinal-type gastric carcinomas are believed to undergo an adenoma-carcinoma sequence similar to colorectal cancers, although specific alteration of genes involved in the process are quite different 19,20 . Loss of 15-PGDH promotes the earliest steps of colon carcinogenesis 21 and we observed this phenomenon in the early gastric carcinogenesis as well, using the ESD specimens. The expression of 15-PGDH was maintained until atrophy or intestinal metaplasia, but disappeared in dysplasia (from adenoma to carcinoma). On the other hand, COX-2 expression started to increase from metaplasia throughout dysplasia. Loss of 15-PGDH combined with COX-2 increase would result in PGE 2 accumulation and this was partly verified either by IHC or mass spectrometry. This observation advocates the role of 15-PGDH as a gatekeeping tumor suppressor. On the contrary, we observed persistent 15-PGDH expression in early signet ring cell carcinoma ( Supplementary Fig. 7), especially when the tumor was confined within the lamina propria. Signet ring cell carcinoma is a typical diffuse-type gastric carcinoma and it is believed to undergo a different pathway of carcinogenesis. One of the proposed mechanism of 15-PGDH down regulation in the literature is the promoter methylation. This was observed not only in advanced gastric cancer 12 , but in non-neoplastic gastric mucosa with H. pylori infection as well 10 . However, our results were to the contrary. No methylation was detected in all of the patient samples and in some of the gastric cancer cell lines with no 15-PGDH expression. Jang et al. also said that they did not observe promoter methylation in advanced gastric cancers 13 . It may be possible that methylation was responsible for 15-PGDH loss in a certain subset of patients, but miR-21 seemed more involved in 15-PGDH inhibition in early carcinogenesis according to our data. miR is a small non-coding RNA molecule, that functions in RNA silencing and post-transcriptional regulation of gene expression. In previous studies, Lu et al. 17 and Li et al. 22 found miR-21 binding sites in the 3′UTR of 15-PGDH mRNA using microRNA.org resource. In the present study, we showed that transfected miR-21 inhibited 15-PGDH, which in turn enhanced cell proliferation and reduced apoptosis. 15-PGDH suppression by CRISPR-Cas9 also increased proliferation, which is in agreement with the results from Li et al. who saw inhibition of cell proliferation by transfecting 15-PGDH 23 . However, miR-21 is a well-known oncomiR and overexpressed in many solid tumors. In addition, 15-PGDH is not the only possible target that miR-21 inhibits. Many studies have advocated that miR-21 enhances gastric cancer growth though PTEN 24,25 and this regulates sensitivity to certain drugs 26,27 . miR-21 has also been studied in other gastrointestinal cancers, such as cholangiocarcinoma 17 , and colorectal cancer 28,29 . Therefore, it is crucial to provide additional evidence that miR-21 binds to 15-PGDH mRNA by immunoprecipitation methods, such as RNA immunoprecipitation or crosslinking and immunoprecipitation 30 . In addition, it would be interesting to determine if PTEN is involved in 15-PGDH inhibition as well. Other possible mechanisms responsible for 15-PGDH loss would be COX-2. Liu et al. demonstrated that COX-2 down regulates 15-PGDH 31 . We also observed an inverse correlation with COX-2 and 15-PGDH in some gastric cancer cell lines, as well as in a few patient samples, but there were also cancer cell lines and cases that did not express COX-2 even at the absence of 15-PGDH, suggesting other regulatory mechanisms are in play.
H. pylori is considered a Group I carcinogen by the International Agency for Research on Cancer 32 and it is known to cause genetic and epigenetic changes 33 , as well as PGE 2 synthesis 34 . We have previously shown that H. pylori infection inhibits 15-PGDH expression and this was associated with epidermal growth factor receptor (EGFR) and Snail 10 . Zhao et al. also reported that H. pylori infection was strongly associated with 15-PGDH loss 35 . The data from the current study suggested that miR-21 might be also be involved in early carcinogenesis because of persistent H. pylori infection. However, this question was far beyond the scope of the present study because much more in-depth investigation is needed to unravel a direct correlation of H. pylori infection and miR-21.
In summary, we have shown that 15-PGDH is markedly lower during early carcinogenesis of gastric tubular adenocarcinoma and is regulated by miR-21, but not by methylation. We have also observed that H. pylori infection is correlated with an increase in miR-21, which requires further research. Maintaining 15-PGDH enzyme activity would be a new strategic measure in preventing tubular adenocarcinoma of the stomach.

Materials and Methods
Patients and tissue specimens. Three groups of human samples were used in this study. First, 30 ESD cases of well-differentiated adenocarcinomas were retrospectively collected between 2007 and 2008. All were very early lesions harboring tubular adenoma (low or high grade) and intestinal metaplasia in the background. Second, 20 more ESD cases were prospectively enrolled in 2014. Well-differentiated adenocarcinomas eligible for ESD were targeted and four forcep biopsies were taken from tumor and normal areas. Top frozen section slides were histologically evaluated for the proportion of tumor tissue. Four cases were excluded because of the lack of tumor tissue and 16 cases were used in this study. Lastly, we used formalin-fixed paraffin-embedded (FFPE) tissues from 26 patients enrolled in our previous study 10 . All patients had paired biopsy from the body and antrum. All studies using human specimens adhered to the guidelines established by the Declaration of Helsinki, and were approved by the institutional review board of the Asan Medical Center. We obtained informed consent from all human participants.  Western blotting (WB). For protein extraction, cells were lysed by RIPA buffer (Thermo) with a cocktail of protease and phosphatase. Protein lysates were separated on SDS-polyacrylamide gels and then transferred to PVDF membranes (Bio-Rad). Membranes were blocked in TBST buffer containing 5% bovine serum albumin (BSA) for 1 h at room temperature (RT) and probed with primary antibodies (supplementary materials) overnight. The membrane was then incubated in horseradish peroxidase-linked secondary antibodies (cell signaling) diluted 1:5000 in 5% BSA. Blots were developed using a Supersignal West pico chemiluminescent kit (Thermo).
Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis. qRT  Genomic data from the TCGA cohort. Genomic data from the TCGA gastric cancer cohort were downloaded from the TCGA data portal site (http://cancergenome.nih.gov/) and processed as described in previous studies 18 . We analyzed mRNA matched miRNA sequencing data.
The miR-21 and scramble mimic were transfected into gastric cancer cells using RNAiMAX (Invitrogen) according to the manufacturer's instructions.
Proliferation and apoptosis assay. miR-21 transfected cell lines were seeded at a density of 1 × 10 4 cells/ well in 96-well plates using an automated cell counter (Luna). Cell proliferation assays were performed according to the manufacturer's instructions (WST-1 reagent, ROCHE). The absorbance of solution was measured according to the manufacturer's protocol using an ELISA reader. The apoptosis assay was analyzed by flow cytometry, which measures cells positively stained using the Annexin V-FITC apoptosis detection kit (BD Pharmingen) according to the manufacturer's instructions. Flow cytometry was performed on BD FACSCanto ™ (BD Biosciences). Data from a total of 10,000 events were analyzed using BD FACSCanto ™ clinical software v2.4 (BD Biosciences).