Role of miR‐218‐GREM1 axis in epithelial‐mesenchymal transition of oral squamous cell carcinoma: An in vivo and vitro study based on microarray data

Abstract Oral squamous cell carcinoma (OSCC) is a prevalent cancer that develops in the head and neck area and has high annual mortality despite optimal treatment. microRNA‐218 (miR‐218) is a tumour inhibiting non‐coding RNA that has been reported to suppress the cell proliferation and invasion in various cancers. Thus, our study aims to determine the mechanism underlying the inhibitory role of miR‐218 in OSCC. We conducted a bioinformatics analysis to screen differentially expressed genes in OSCC and their potential upstream miRNAs. After collection of surgical OSCC tissues, we detected GREM1 expression by immunohistochemistry, RT‐qPCR and Western blot analysis, and miR‐218 expression by RT‐qPCR. The target relationship between miR‐218 and GREM1 was assessed by dual‐luciferase reporter gene assay. After loss‐ and gain‐of‐function experiments, OSCC cell proliferation, migration and invasion were determined by MTT assay, scratch test and Transwell assay, respectively. Expression of TGF‐β1, Smad4, p21, E‐cadherin, Vimentin and Snail was measured by RT‐qPCR and Western blot analysis. Finally, effects of miR‐218 and GREM1 on tumour formation and liver metastasis were evaluated in xenograft tumour‐bearing nude mice. GREM1 was up‐regulated, and miR‐218 was down‐regulated in OSCC tissues, and GREM1 was confirmed to be the target gene of miR‐218. Furthermore, after up‐regulating miR‐218 or silencing GREM1, OSCC cell proliferation, migration and invasion were reduced. In addition, expression of TGF‐β signalling pathway‐related genes was diminished by overexpressing miR‐218 or down‐regulating GREM1. Finally, up‐regulated miR‐218 or down‐regulated GREM1 reduced tumour growth and liver metastasis in vivo. Taken together, our findings suggest that the overexpression of miR‐218 may inhibit OSCC progression by inactivating the GREM1‐dependent TGF‐β signalling pathway.


| INTRODUC TI ON
Squamous cell carcinoma is the most prevalent cancer of the head and neck, with oral squamous cell carcinoma (OSCC) accounting for 90% of the cases. 1 OSCC is the eight most common cancer in the USA, with over 48 000 new OSCC cases diagnosed annually. 2 The survival rate for OSCC patients is only about 50%. 3 Despite expenditure of significant efforts in improving the diagnostic and therapeutic approaches of OSCC, the mortality rate in OSCC patients remains unchanged in recent years. 4 Moreover, OSCC patients will often suffer from functional and cosmetic defects following treatment. 5 Lifestyle, environment and genetic background are important factors associated with the development of OSCC, 6 and the main events related to OSCC-induced mortality are metastasis and recurrence. Therefore, it is important to identify biological factors and regulatory molecular pathways that might lead to improved treatment for OSCC. 7 The abnormal expression of microRNAs (miRNAs) can serve as tumour or oncogene suppressors, capable of regulating various biological processes such as development, differentiation, apoptosis and cell proliferation. 8 Multiple types of cancer, including OSCC, have been observed to have dysregulated expression of miRNAs. 7 miR-218, a vertebrate-specific intronic miRNA that is expressed along with its host gene tumour suppressor gene SLIT2/3 plays a tumour suppressive role by targeting numerous oncogenes associated with cell proliferation, apoptosis and invasion. 9 According to prediction results obtained from Bioinformatic websites, Gremlin1 (GREM1) could be a target gene of miR-218, which led us to hypothesize that miR-218 can modulate GREM1 expression in OSCC.
GREM1 is an antagonist of bone morphogenetic protein, which plays a key role in multiple biological processes, including cancer biology. 10 In particular, GREM1 plays an important role in certain malignancies by antagonizing bone morphogenetic proteins and regulating angiogenesis. 11 In addition, Kim et al identified GREM1 as an OSCCrelated gene. 12 Previous studies have suggested that GREM1 can inhibit BMP, 13 which is a member of TGF-β family [14][15][16][17] and have also suggested that interference with GREM1 expression may result in decreased TGF-β expression. 18 Insofar as GREM1 could potentially affect TGF-β signal pathway, we conducted the present study with the main objective of exploring the effects of miR-218 on OSCC and its mechanism involving GREM1 and the TGF-β signalling pathway.

| Ethical statement
This study was conducted in strict accordance with the Helsinki declaration and approved by the medical ethics committee of Linyi

| Gene expression omnibus (GEO) expression dataset retrieval and differential analysis
In the GEO database (https://www.ncbi.nlm.nih.gov/geo/), four expression data sets related to OSCC were obtained, namely GSE10121 (35 OSCC cases and 6 controls), GSE30784 (167 OSCC cases and 45 controls), GSE37991 (40 OSCC cases and 40 controls) and GSE74530 (6 OSCC cases). The R 'limma' package was used to analyse the difference between the OSCC group and the normal control group of the four expression data sets, and the differential gene expression dendrogram was constructed using the 'pheatmap' package. A Venn diagram of differential genes in the four expression data sets was constructed to determine the intersection of the differential genes.

| Analysis of known gene retrieval and gene interaction in OSCC
DigSee (http://210.107.182.61/geneS earch/) is a text retrieval search engine to provide evidentiary sentences describing that 'genes' are involved in the development of 'disease' through 'biological events'.
The database was used to retrieve the known associated genes of OSCC, the first ten of which were selected for subsequent analysis. Using STRING database (https://string-db.org/), we analysed the interaction of the 10 genes and the above intersection genes in the expression data set. After using Cytoscape software (ver3.7.1) for visualization of the interaction, we constructed a gene interaction network diagram.

| Study subjects
A total of 65 resected specimens (OSCC tissues and adjacent normal tissues) were collected from the patients with pathologically confirmed OSCC who underwent surgery resection at Linyi People's Hospital. The patients consisted of 43 males and 22 females (range 34-78 years). The OSCC were categorized according to the American Joint Committee on Cancer tumour-node-metastasis (AJCC TNM) classification in 2010: stage I/II, n = 26; stage III, n = 39; high differentiation, n = 25; moderate and low differentiation, n = 40; liver metastases nodes, n = 29; non-LNM, n = 36. All specimens were fixed with 10% formaldehyde, embedded with paraffin and cut into 8 μm-thick sections for subsequent use.

| Immunohistochemistry
The tissue sections were baked at 60°C for 1 hour, dewaxed with conventional xylene and hydrated with gradient ethanol. Next, the sections underwent incubation in phosphate-buffered saline (PBS) containing 0.5% Triton-X at room temperature for 20 minutes. After  After subsequently culturing the cells in a complete culture medium for 24-48 hours, the cells were collected, and their RNA and protein were extracted for analysis. The primer sequence for miR-218 mimic was: 5′UUGUGCU UGAUCUAACCAUGU3′, 21 and that for miR-218 inhibitor was: 5′AACACGA ACUAGAUUGGUACA3′.

| Reverse transcription quantitative polymerase chain reaction (RT-qPCR)
Total RNA from tissues and cells was extracted by TRIzol (Invitrogen, Carlsbad, CA, USA) and a nanodrop2000 ultraviolet spectrophotom-  Table 1). ABI7500 quantitative PCR instrument (7500; Applied Biosystems Inc) was applied to conduct RT-qPCR detection. The ratio of target gene expression between the experiment group and the control group was determined using the 2 −ΔΔCt method, with β-actin used as the internal control.

| Western blot analysis
The total protein of cells to be detected was extracted using Radio

| Transwell assay
The apical chamber of the Transwell chambers was coated with Matrigel diluted in pre-cooled serum-free DMEM medium and incubated in an incubator at 37°C for 4-5 hours. After the solidification of Matrigel, 100 μL serum-free medium was used to dilute the transfected cells to make a suspension containing 1 × 10 6 cells/mL, followed by inoculation. The basolateral chamber was added with 500 μL DMEM containing 20% FBS. Each group was set three duplicate wells. After incubation at 37°C for 24 hours with 5% CO 2 , the Transwell chamber was fixed with 5% glutaraldehyde at 4°C and then stained with 0.1% crystal violet for 5 minutes. After removal of the surface adhering cells with cotton balls, the cells were observed with an inverted fluorescence microscope (Nikon TE2000, Tokyo, Japan). The mean number of cells in three visual fields was measured.

| Scratch test
The transfected cells were incubated in an incubator at 37°C for 24 hours with 5% CO 2 , whereupon the monolayer cells were wounded by scratching with a sterile 10 μL pipette tip. Next, the cells were added with serum-free medium and incubated for 24 hours.
Cell migration at 0 and 24 hours was observed using the inverted microscope. Three sites were selected in each group for photography and measurement of the relative migration distance between cells on either side of the scratch. The distance difference was divided by two to obtain relative migration distance, and cell migration rate was calculated as the relative migration distance/the distance from the scratch margin to centerline at 0 hours. The experiment was conducted in triplicate.
Then, the single cell suspension was stained by trypan blue to count the numbers of living cells.

| Liver metastasis in nude mice
Cells were treated as described above, and lentivirus expressing shRNAs targeting human GREM1 transcripts was generated using the Mission PLKO.   Figure 1E). The results revealed 25 genes at the intersection from all 4 expression data sets, which were selected as candidate genes. DigSee database was used to retrieve the known genes associated with OSCC, and the top 10 were selected for subsequent analysis ( Table 2). We analysed the genetic interactions between the 10 known retrieved genes and the 25 candidate genes obtained from the expression data set ( Figure 1F). The results showed that, among the 25 candidate differentially expressed genes, GREM1 gene was in the most core position and had an interaction with the ten known genes associated with OSCC ( Figure 1G To elaborate further the mechanism of GREM1 gene in OSCC, 5 databases were used to predict miRNAs mediating GREM1 and the intersections of the predicted results were determined ( Figure 1H).

| Up-regulated GREM1 and down-regulated miR-218 are found in OSCC
Immunohistochemistry showed that the positive staining of GREM1 had a brown yellow appearance. Immunohistochemistry and Western blot analyses showed that GREM1 protein expression was much higher in OSCC tissues when compared to adjacent normal tissues (P < 0.02, Figure 2A in OSCC tissues compared with that in adjacent normal tissues, while the expression of GREM1 was high in the tumour tissues ( Figure 2D; n = 65). We further analysed the correlation between miR-218 and GREM1 expression, finding that miR-218 was negatively correlated with GREM1 expression (P < 0.05) ( Figure 2E). In addition, the analysis of the relationship between clinicopathological features of patients with OSCC and the expression of miR-218 and GREM1 showed that the expressions of miR-218 and GREM1 were closely related to LNM and TNM stage (both P < 0.05) but had no relationship with patient age or gender, or the size of OSCC tumour and degree of differentiation (Table 3). These findings suggest that OSCC tissues had increased expression of GREM1 and decreased miR-218 expression.

| GREM1 may be target gene of miR-218
According to the bioinformatics online prediction tool microRNA.org, GREM1 might be a potential target gene of miR-218 ( Figure 3A). Dualluciferase report gene assay ( Figure 3B) was used to confirm the targeting relationship between miR-218 and GREM1, showing that, compared with that in the NC group, the luciferase activity of pGREM1-Wt was markedly decreased in the miR-218 mimic group (P < 0.05), while the luciferase activity of pGREM1-Mut did not differ between the two groups (P > 0.05). Therefore, GREM1 was the potential target gene of miR-218, which could negatively regulate its expression.

| Overexpressed miR-218 may inhibit the EMT progression via the TGF-β signalling pathway
To verify further the impact of GREM1 knockdown on OSCC behaviour, we tested the knockdown efficiency of GREM1, finding relatively higher knockdown efficiency of si-GREM1-1 ( Figure 4A

| Up-regulated miR-218 or silencing GREM1 suppresses the OSCC cell invasion, migration and proliferation
Transwell assay and scratch test were conducted to elucidate the in the miR-218 inhibitor group, but was markedly reduced in the miR-218 mimic and si-GREM1 groups (all P < 0.05). There were no prominent differences in the miR-218 inhibitor + si-GREM1 group, the blank group and the NC group (P > 0.05; Figure 5C; Figure   S2E). The aforementioned findings imply that overexpressed miR-218 and silenced GREM1 could suppress the invasion, migration and proliferation in OSCC cells.

| Overexpressed miR-218 inhibits tumour growth and liver metastasis in nude mice with OSCC
The effect of miR-218 on tumour growth and liver metastasis in agomir group and the si-GREM1 group was lower compared with the blank group, while the number of liver metastasis nodes in the miR-218 antagomir group was higher than that in the blank and NC groups (all P < 0.05). Furthermore, the number of liver metastasis nodes did not differ among the miR-218 antagomir + si-GREM1, blank and NC groups (all P > 0.05; Figure 6D). The findings show that overexpressed miR-218 or silencing GREM1 led to the inhibition of tumour growth and liver metastasis of nude mice with OSCC.

| D ISCUSS I ON
OSCC is the most prevalent malignancy of the head and neck region, and its occurrence rate has increased in recent years. 27 However, current treatment modalities for OSCC, which include extensive surgery, radiotherapy, chemotherapy or concurrent chemo-radiation, have low efficacy in patients with advanced OSCC, due to the increased occurrence of tumour recurrence and metastasis. 9 Therefore, identifying the underlying factors involved in the recurrence of OSCC is pivotal to reduce its postoperative recurrence and obtained better survival. 28  expression, which conceivably inhibits cell proliferation and promotes cell apoptosis, 31 thus indicating that miR-218 can inhibit cancer occurrence and progression mainly by suppressing cancer cell proliferation and invasion through the targeting of cancer genes. 32 Another study demonstrated that reduced miR-218 expression in gastric cancer was associated with advanced clinical stage, LNM and absence of accurate prognosis, while the expression of miR-218 in metastatic cells had the ability to suppress migration, invasion and metastasis formation in vitro and in vivo. 33 As a member of the DAN family of bone morphogenetic protein (BMP) antagonists, GREM1 is involved in the regulation of numerous cell functions in developing and adult tissues. 34 Interestingly, cancer-associated fibroblasts (CAFs) of human basal cell carcinomas have high levels of GREM1 expression, which promote the proliferation of cultured BCC cells. 35 Based on the biology prediction website microRNA.org and present results of the dual-luciferase reporter gene assay, GREM1 was identified as the target gene of miR-218, which could negatively regulate the expression of GREM1.
Our study also demonstrated that the overexpression of miR-218 or silencing GREM1 could suppress the invasion, migration and proliferation of OSCC cells by inhibiting the expressions of TGF-β, Vimentin and Snail, while promoting the expressions of Smad, p21 and E-cadherin. Tatarano et al similarly observed that miR-218 expression could influence down-regulated oncogenic genes and up-regulated tumour suppressive genes. 36 GREM1 plays a suppressive role by binding to BMP dimers, thus preventing their interaction with BMP receptors, as well as inhibiting BMP secretion and improving extracellular BMP endocytosis. We note that BMPs are members of the TGF-β super-family. 37 As a multifunctional cytokine, TGF-β1 can regulate a complicated signalling and inflammatory network and has dual effects in tumour development. 38 Moreover, networks regulated by Smad-independent TGF-β also can be found in the well-characterized Smad signalling pathway. The activation of signalling pathways leads to the activation of transcriptional regulators such as Snail, which is a crucial regulator of EMT, and can suppress the gene expression of E-cadherin, which is a key feature of the epithelial phenotype. 39 Through tumour formation in nude mice, we also found that the overexpression of miR-218 resulted in the inhibition of tumour growth and liver metastasis. Tian et al reported that overexpression of miR-218 in oesophageal squamous cell carcinoma cell lines suppressed cell proliferation, colony formation, migration and invasion and also inhibited tumour growth in nude mice. 40

| CON CLUS ION
The key finding from this study is our demonstration that miR-218 could inhibit OSCC proliferation, migration, invasion, EMT and liver metastasis by inactivating the TGF-β signalling pathway via GREM1 down-regulation. Thus, miR-218 can serve as a cancer suppressor in OSCC and has potential as a predictor for the prognosis of OSCC ( Figure 7). These findings may present this pathway as a promising target for development of novel OSCC therapies. However, the specific underlying mechanism of miR-218 in OSCC shall require further elucidation in larger surgical specimens and in alternate cell lines.

ACK N OWLED G EM ENTS
The authors sincerely appreciate all members participated in this work.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no competing interests.