Abstract

   MicroRNAs (miRNAs) play important roles in regulating tumour
   development and progression. Here we show that miR-647 is repressed in
   gastric cancer (GC), and associated with GC metastasis. Moreover, we
   identify that miR-647 can suppress GC cell migration and invasion in
   vitro. Mechanistically, we confirm miR-647 directly binds to the 3'
   untranslated regions of SRF mRNA, and SRF binds to the CArG box located
   at the MYH9 promoter. CCG-1423, an inhibitor of RhoA/SRF-mediated gene
   transcription, inhibits the expression of MYH9, especially in SRF
   downregulated cells. Overexpression of miR-647 inhibits MGC 80-3 cells'
   metastasis in orthotropic GC models, but increasing SRF expression in
   these cells reverses this change. Importantly, we found the synergistic
   inhibition effect of CCG-1423 and agomir-647, an engineered miRNA
   mimic, on cancer metastasis in orthotropic GC models. Our study
   demonstrates that miR-647 functions as a tumor metastasis suppressor in
   GC by targeting SRF/MYH9 axis.

   Keywords: Gastric cancer, Metastasis, MicroRNA-647, MYH9, SRF.

Introduction

   Gastric cancer (GC) is the third leading cause of cancer-related death
   worldwide and causes more than 723,000 deaths every year [64]^1.
   Despite recent therapeutic advances, the mortality of GC remains high
   due to its silent nature, late clinical presentation and high levels of
   biological and genetic heterogeneity [65]^2. GC is a heterogeneous
   disease that evolves in various genetic and epigenetic alterations
   [66]^2^-[67]^9, including transcriptional changes of noncoding RNAs.

   MicroRNAs (miRNAs), a class of noncoding RNAs, regulates protein
   expression through incomplete base pairing with the 3' untranslated
   regions (3'UTR) of target messenger RNAs (mRNAs) [68]^10. Accumulating
   evidences suggest that miRNAs play an important role in GC development
   and progression. Though certain miRNAs, such as miR-21 [69]^11,
   miR-148a [70]^12, miR-218 [71]^13 and miR-29c [72]^14, have been
   confirmed to act as oncogenes or tumor suppressor genes in GC, there
   are many other miRNAs suggested to be dysregulated by microRNA arrays,
   of which the detailed molecular mechanisms in GC are still not well
   elucidated. MicroRNA-647 (miR-647) has been reported to be a predictive
   biomarker for prostate cancer recurrence and a prognostic factor for
   Taxol-sensitive ovarian cancer patients [73]^15^, [74]^16. It was also
   identified as one of the 7 population-differentiated-miRNAs that are
   currently implicated as cancer biomarkers or diagnostics [75]^17.
   Although miR-647 has been reported to be dysregulated in GC [76]^18^,
   [77]^19, the molecular mechanism of miR-647 in cancer metastasis was
   not fully uncovered.

   Here we report that miR-647 expression is repressed in GC. In vitro and
   in vivo restorations of miR-647 via exogenous transfection are
   sufficient to inhibit both GC cell migration and invasion. SRF, a
   transcription factor known to promote tumor metastasis, and MYH9, a
   well-known cytoskeleton, are direct and indirect downstream targets of
   miR-647, respectively. MiR-647/SRF/MYH9 axis is also strongly
   correlated with GC metastasis in patients. Importantly, we demonstrate
   that agomir-647, an engineered miRNA mimic, and CCG-1423, an inhibitor
   of the Rho/MKL1/SRF signaling [78]^20, synergistically inhibit cancer
   metastasis in orthotropic GC models. Taken together, our results
   provide an explanation of the metastatic mechanism of GC, and might be
   novel therapeutic targets or prognostic markers for GC.

Materials and Methods

Patients and Specimens

   Two independent cohorts involving 199 GC patients were enrolled in this
   study. In cohort 1, fresh GC samples and adjacent noncancerous tissues
   were collected from 109 pathologically verified GC patients who
   underwent gastrectomy between November 2012 and November 2013 in
   Nanfang Hospital, Southern Medical University (Guangzhou, China). None
   of the patients received chemotherapy or radiotherapy before
   gastrectomy. The GC diagnosis was made by endoscopic biopsy combined
   with histopathological information after surgery. Diagnosis was
   confirmed by two pathologists in Nanfang hospital according to American
   Joint Committee on Cancer (AJCC) TNM Staging Classification for
   Carcinoma of the Stomach (7th ed., 2010) [79]^21. Most of cohort 1
   patients were followed up for 3 years. This study was approved by the
   Ethics Committee of Nanfang hospital (Guangzhou, China). For cohort 2,
   a gastric adenocarcinoma tissue microarray (TMA, HStm-Ade180Sur-05;
   Shanghai Outdo Biotech) was obtained from the National Engineering
   Center For Biochip at Shanghai. The TMA was constructed with 90 paired
   formalin-fixed, paraffin-embedded gastric adenocarcinoma tissues and
   their corresponding adjacent normal tissues. Cohort 2 patients were
   followed up for 7.8 years. The corresponding clinicopathological
   information was listed in Table [80]S1.

Cell Lines and Cell Culture

   Human GC cell lines (MGC 80-3, MKN45, SNU-5, AGS and KATOIII) were
   purchased from the Cell Resource Center, Shanghai Institute of
   Biochemistry and Cell Biology at the Chinese Academy of Sciences. Cells
   were cultured in RPMI-1640 medium (MGC 80-3 and MKN45), F12 (AGS) or
   Dulbecco's Modified Eagle's Media (KATOIII and SNU-5) containing 10%
   fetal bovine serum (FBS) at 37°C in an atmosphere of 5% CO[2].

In Situ Hybridization (ISH)

   All the paraffin sections were examined with locked nucleic acid (LNA)
   based ISH using DIG-labeled miRCURY miRNA probes (Exiqon, Vedbaek,
   Denmark). The protocol for detection of miRNAs has been previously
   published [81]^22.

Statistical Analysis

   Statistical analysis was performed using GraphPad Prism software 5.0
   (GraphPad Software, Inc., San Diego, CA, USA) and SPSS software
   (Version 21.0; Abbott Laboratories, Chicago, IL). Data were shown as
   mean ± standard error of mean unless otherwise noted. The Student t
   test was used to detect significance of data from qPCR experiments,
   migration and invasion assays, luciferase reporter assay, ChIP assay
   and in vivo metastatic nodules assay. Moreover, it was used to detect
   the relationship between miR-647 expression and clinicopathologic
   characteristics. Pathway enrichment analysis of miR-647 target genes
   was based on Fisher's exact test using a 2×2 contingency table.
   Mann-Whitney U test was used to analyze SRF protein expression between
   normal and GC tissues detected by IHC. The associations between miR-647
   expression detected by ISH and proteins expression detected by IHC were
   analyzed by the χ^2 test. Survival curves were calculated using
   Kaplan-Meier and log-rank tests. The effects of variables on survival
   were determined by univariate and multivariate Cox proportional hazards
   modeling. Correlation between SRF and MYH9 proteins in the TMA was
   analyzed by Spearman rank correlation, while correlation between SRF
   and MYH9 mRNA data extracted from TCGA was analyzed by Pearson
   correlation. A p value less than .05 was considered statistically
   significant.

   All supplementary methods are available in [82]supplementary material.

Results

MiR-647 is downregulated in GC tissues and associated with GC metastasis

   To identify the expression of miR-647 in GC, we first detected its
   expression in 109 human GC tissues by qPCR. The result showed that
   miR-647 was frequently downregulated in GC tissues (T) compared to
   normal gastric mucosa (N) (p <0.0001, Figure [83]1A). Moreover, 73.4%
   (80 of 109) of the GC had at least 2-fold reduced expression of miR-647
   compared with their corresponding nontumorous tissues (Figure [84]1B),
   which was supported by data from [85]GSE36968 (Figure [86]S1A). To
   determine whether miR-647 expression levels were related to GC
   progression, we analyzed the association between miR-647 and
   clinicopathologic status in 109 GC patients (Table [87]S1-2). As shown
   in Figure [88]1C and Table [89]S2, statistical analysis represented a
   strong correlation between miR-647 expression and TNM stage (p=0.0020),
   lymph-node metastasis (p=0.0046) and local invasion (p=0.0062).
   Collectively, these findings strongly suggested that miR-647 was
   downregulated in GC and negatively associated with GC metastasis.

Figure 1.

   [90]Figure 1
   [91]Open in a new tab

   MiR-647 is downregulated in gastric cancer tissues and associated with
   the migration and invasion of gastric cancer cells. (A) Expression of
   miR-647 was determined by qPCR in 109 human gastric cancer tissues,
   which was normalized against an endogenous U6 RNA control. ***p<0.001.
   (B) Relative miR-647 levels of GC and normal tissues measured by qPCR
   were shown using pie chart and waterfall plot. The fold change of
   relative miR-647 expression (T/N) >2 or <1/2 was defined as
   significant. For waterfall plot, the fold change of relative miR-647
   expression (log[2][T/N]) >1 or <-1 was defined as significant. (C) The
   relationship between miR-647 expression and clinical stages, lymph
   node-metastasis or local invasion. *p<0.05; **p<0.01. (D) Expression of
   miR-647 in 5 normal human gastric mucosa tissues, 5 GC cell lines and 3
   colorectal cancer cell lines was analyzed by qPCR. (E and F) Expression
   of miR-647 in transfected MGC 80-3, MKN45 and AGS cell lines were
   analyzed by qPCR. ***p<0.001. (G and H) The 3D migration and invasion
   changes of different MGC 80-3 cells were tested using transwell chamber
   migration assay and invasion assay. Cell 2D migration was tested using
   monolayer wound healing assay. ***p<0.001.

   To detect the function of miR-647 in GC cell invasion and migration, we
   first determined the expression of miR-647 in 5 normal human gastric
   tissues from 109 gastric cancer patients, 5 human gastric cancer cell
   lines (MGC 80-3, MKN45, SNU-5, AGS and KATOIII) and 3 human colorectal
   cancer cell lines (SW480, SW620 and Caco2) available in our lab by
   qPCR. Since miR-647 had been identified as one of the 7 miRNAs that
   were implicated as cancer biomarkers or diagnostics for multiple
   cancers, including colorectal cancer [92]^17, the human colorectal
   cancer cell lines were used as positive controls. The results showed
   miR-647 expression was relatively lower in GC cell lines compared with
   normal tissues (Figure [93]S2A). Since miR-647 expression was lowest in
   MGC 80-3 among 5 GC cell lines, followed by MKN45 (Figure [94]1D), and
   AGS had relatively higher miR-647 expression level and lentivirus
   transfection efficiency, we chose MGC 80-3, MKN45 and AGS to establish
   miR-647 stably overexpressing cell lines by lentivirus infection.
   Successful overexpression or knockdown of miR-647 was confirmed by qPCR
   (Figure [95]1E-F). Intriguingly, overexpression of miR-647
   significantly suppressed the migratory and invasive abilities of MGC
   80-3 and AGS cells (Figure [96]1G-H and S2D). Simultaneously, knockdown
   of miR-647 expression in these LV-miR-647 GC cells by transiently
   transfecting miR-647 inhibitors could restore the migration and
   invasion of GC cells (Figure [97]1G-H and S2D).

MiR-647, associated with Rho Signaling Pathways, negatively regulates SRF and
MYH9 expression in GC cells

   To identify how miR-647 was involved in GC metastasis, we firstly used
   4 common miRNA target prediction engines (miRanda, TargetScan, miRDB,
   and miRWalk) to search for putative protein-coding gene targets of
   miR-647
   ([98]http://zmf.umm.uni-heidelberg.de/apps/zmf/mirwalk/micrornapredicte
   dtarget.html), and 7724 potential targets were found (Figure [99]2A).
   Since miR-647 significantly inhibited GC migration and invasion, we
   selected 26 genes, listed out in Figure [100]2A and Table [101]S3,
   associated with GC metastasis in PubMed. Additionally, we performed
   pathway enrichment analysis of miR-647 target genes predicted in the
   miRBase Targets database ([102]http://microrna.sanger.ac.uk/targets)
   using miRGator online software ([103]http://genome.ewha.ac.kr miRGator)
   (Table [104]S4) [105]^23^, [106]^24. Specifically, the analysis showed
   that miR-647 was associated with Rho cell motility signaling pathway
   (Figure [107]2B, p=0.0134, BioCarta) and Role of MAL in Rho-mediated
   activation of SRF (Figure [108]2B, p=0.0348, BioCarta), which were
   critical pathways synergistically promoting cancer cell migration and
   invasion [109]^25. Because Rho-mediated pathways are closely associated
   with cancer metastasis and frequently activated in GC [110]^7^,
   [111]^26, we selected out 5 candidate genes (SRF, ROCK1, ZAK, ELMO2,
   and MYH9) from Figure [112]2A (red marked) for further research, which
   were reported to be up-regulated in GC or associated with Rho-ROCK and
   Rho-MRTF/SRF pathways [113]^12^, [114]^27^-[115]^30.

Figure 2.

   [116]Figure 2
   [117]Open in a new tab

   MiR-647 involves in Rho-associated signal pathway and negatively
   regulates SRF and MYH9 in gastric cancer cell lines. (A) Bioinformatics
   prediction of potential miR-647 targets by 4 common miRNA databases.
   Part of the genes associated with cell motility were showed and
   Rho-associated genes were marked as red. (B) Pathway enrichment
   analysis of miR-647 target genes predicted in the miRBase Targets
   database ([118]http://microrna.sanger.ac.uk/targets). Analysis was
   performed using version 4 of the database and miRGator. Cell
   motility-associated signal pathways were listed. (C) Expression of
   Rho-associated genes were analyzed in MGC 80-3 and AGS by qPCR. U6
   served as an internal control. ***p<0.001. (D-E) Expression of SRF and
   MYH9 proteins in 5 normal gastric mucosa tissues and 5 gastric cancer
   cell lines was analyzed by western blot and followed grayscale value
   analysis. GAPDH served as an internal control. (F) Western blot
   analysis was used to detect the expression level of SRF and MYH9 in MGC
   80-3 and MKN45 cells after infection with miR-647 expressing or control
   lentivirus. GAPDH served as an internal control. (G) Western blot
   analysis was used to detect the expression changes of SRF and MYH9 in
   MGC 80-3-LV-miR-647 and MKN45-LV-miR-647 cells after infection with
   miR-647 inhibitors or its negative control (INC). GAPDH served as an
   internal control.

   Since MGC 80-3 and MKN45 expressed miR-647 at relatively low level, AGS
   and SNU-5 expressed miR-647 at relatively high level (Figure [119]1D),
   we transfected MGC 80-3 and MKN45 with miR-647 mimics, AGS and SNU-5
   with miR-647 inhibitors (Figure [120]S2E). Then, we carried out qPCR to
   detect these genes in transfected cells. The results revealed that the
   expression levels of SRF mRNA were most downregulated in MGC 80-3 and
   MKN45 cells treated with miR-647 mimics, followed by MYH9 and ZAK mRNAs
   (Figure [121]2C and S2F). Consistently, these three mRNAs were
   upregulated in AGS and SNU-5 cells treated with miR-647 inhibitors
   (Figure [122]2C and S2G). Because the fold changes of SRF and MYH9
   mRNAs in miR-647-dysregulated GC cells are the highest, we focused on
   these two genes and carried out western blot and qPCR to confirm their
   relationship with miR-647. As showed in Figure [123]2D-E and S2B-C, the
   expression levels of SRF and MYH9 were both positively associated with
   miR-647 expression between GC cell lines and normal tissues. Moreover,
   we detected SRF and MYH9 proteins in GC cell lines stably
   overexpressing miR-647 (MGC 80-3-LV-miR-647, MKN45-LV-miR-647 and
   KATOIII-LV-miR-647) by western blot. The results showed SRF and MYH9
   levels were significantly downregulated in the case of miR-647
   overexpression (Figure [124]2F, S2H and S2J). In addition, when we
   transfected MGC 80-3-LV-miR-647, MKN45-LV-miR-647 and SNU-5 with
   miR-647 inhibitors, the expression of SRF and MYH9 proteins
   significantly restored (Figure [125]2G, S2I and S2J). Taken together,
   these results supported that miR-647 was negatively associated with SRF
   and MYH9 expression levels.

MiR-647 inhibits the migration and invasion of GC cell lines by directly
targeting to SRF mRNA, and indirectly regulating MYH9 expression

   We performed luciferase reporter assays to determine whether miR-647
   directly interacts with the 3'UTR of SRF and MYH9 mRNAs. Using a
   computer-based sequence analysis (TargetScan and miRanda), we acquired
   potential miR-647-targeting sequences of SRF and MYH9. As shown in
   Figure [126]3A, three potential binding sites of miR-647 were located
   in 3'UTR of SRF, while one potential binding site in 3'UTR of MYH9.
   Then, we subcloned two 235-bp fragments of the SRF 3'UTR (Figure
   [127]3A; position 19-253 and 1152-1386, marked as WT1 and WT2) that
   included all the three predicted miR-647 recognition sites (Figure
   [128]3A; position 122-128, 150-155 and 1260-1271), and their
   corresponding mutant counterparts (Figure [129]3A; Mut1-3) into a
   luciferase reporter plasmid (pGV306). For testing whether SRF was a
   direct target of miR-647, MGC 80-3 and AGS cells were cotransfected
   with miR-647 mimics (or a negative control) and luciferase vectors
   containing wild type SRF-3'UTR (WT1 and WT2) (or luciferase vectors
   containing mutant-type SRF-3'UTR (Mut1-3)). Similarly, for determining
   whether miR-647 directly interacted with the MYH9 mRNA, we subcloned a
   203-bp fragment of the MYH9 3'UTR (Figure [130]3A; position 732-934,
   marked as WT3) that included the predicted miR-647 recognition site
   (Figure [131]3A; position 832-837), and inserted it or its
   corresponding mutant counterpart (Mut4) into pGV306. Then, MGC 80-3 and
   AGS cells were cotransfected with miR-647 mimics (or a negative
   control) and luciferase vector containing wild type MYH9-3'UTR (WT3)
   (or luciferase vector containing mutant-type MYH9-3'UTR (Mut4)). As
   shown in Figure [132]3B, in cells transfected with WT SRF-3'UTR (WT1 or
   WT2) vectors and the miR-647 mimics, a significant decrease in
   luciferase activity was observed compared with cells cotransfected
   SRF-3'UTR vectors and negative controls. However, the luciferase
   activity of mutant-type SRF-3'UTR vectors (Mut1 and Mut3), excepting
   for Mut2, did not change following cotransfection with the miR-647
   mimics, indicating miR-647 could directly interact with two sequences
   (GCAGCCA and UGAGUGCAGCCA) within SRF 3'UTR (Figure [133]3B). In GC
   cells (MGC 80-3 and AGS) transfected with WT MYH9-3'UTR (WT3) vector
   and the miR-647 mimics, no significant change in luciferase activity
   was observed compared with WT3 vector and negative controls (Figure
   [134]3C). And the luciferase activity of mutant-type MYH9-3'UTR vectors
   (Mut4) also did not change following cotransfection with the miR-647
   mimics (Figure [135]3C). Collectively, these findings suggested miR-647
   could directly target to the 3'UTR of SRF, but not the 3'UTR of MYH9.

Figure 3.

   [136]Figure 3
   [137]Open in a new tab

   MiR-647 inhibits the migration and invasion of gastric cancer cell
   lines by directly targeting to SRF mRNA, which indirectly suppresses
   MYH9 expression. (A) Potential miR-647 binding sequences in the 3'UTR
   of SRF and MYH9 mRNAs predicted by TargetScan and miRanda databases. (B
   and C) Relative luciferase activity was analyzed after the described
   reporter plasmids (pGV306-3'UTR) or mock reporter plasmid (pGV306) were
   transfected into LV-miR-647 or LV-NC stably expressed MGC 80-3 and AGS
   (MGC 80-3-LV-miR-647 and AGS-LV-miR-647). The firefly luciferase
   activity was normalized to Renilla luciferase activity. ***p<0.001. (D
   to E) Dual-luciferase reporter assay was used to confirm SRF targeted
   to the promoter of MYH9 gene. ***p<0.001. (F) ChIP-qPCR analysis of SRF
   binding to the MYH9 promoter region in MGC 80-3 and AGS cells. RNA
   polymerase II (RNAPII) was used as a positive control. The fold
   enrichment over the IgG control is represented (mean ± SD; *p<0.05;
   **p<0.01; ***p<0.001). (G) Illustration of SRF protein targeting to the
   CArG box of MYH9 promoter. (H) Western blot analysis was used to detect
   the expression changes of SRF and MYH9 proteins followed by
   transfecting miR-647-overexpressed gastric cancer cell lines with SRF
   plasmid (SRF) or MYH9 plasmid (MYH9). GAPDH served as an internal
   control. (I) Cell 3D migration and invasion were tested using transwell
   chamber migration assay and invasion assay. AGS-LV-miR-647 transfected
   with SRF plasmid (SRF), MYH9 plasmid (MYH9) and vectors were used. Each
   bar represents the mean ± SD. ***p<0.001.

   To understand why MYH9 expression was indirectly associated with
   miR-647, we analyzed the 1-kb region directly upstream of MYH9 and
   predicted potential transcriptional factors (TFs) by using
   bioinformatics analysis websites (Proscan and JASPAR) (Figure [138]3D).
   We generated the full-length MYH9 promoter construct (luciferase
   plasmid construct) pGL3-MYH9-1300 (wild-type) and five truncation
   constructs (pGL3-MYH9-1000, pGL3-MYH9-903, pGL3-MYH9-753, pGL3-MYH9-528
   and pGL3-MYH9-328) according to the locating sites of predicted TFs
   (Figure [139]3D). Intriguingly, dual-luciferase reporter assay showed
   that position -589 to -439, which potentially contained SRF binding
   site (CCATATATGG), was found to significantly increase the relative
   luciferase activity (Figure [140]3D). SRF was a ubiquitously expressed,
   pleiotropic transcription factor that activated genes by targeting to
   CArG box, which was a 10-bp CC (A+T-rich)[6]GG DNA sequence [141]^31.
   To further confirm this binding site, we constructed pGL3-MYH9-903-mut,
   which contained CArG box-mutant sequence (AACGACCAGAG). Luciferase
   reporter assay showed a significant decrease in reporter activity of
   MGC 80-3 cells cotransfected with pGV311-SRF and pGL3-MYH9-903-mut
   (Figure [142]3E), compared with MGC 80-3 cells cotransfected with
   pGV311-SRF and pGL3-MYH9-903. ChIP-qPCR analysis was further performed
   on MGC 80-3 and AGS cells. RNA polymerase II (RNAPII), which binds to
   the GAPDH promoter, was used as a positive control. The fold enrichment
   over the IgG control was represented. MYH9 promoter sequences were
   specifically enriched by anti-SRF and anti-MRTFA antibodies (a
   coactivator of SRF; Myocardin-related transcription factor A), but not
   by negative control antibody IgG in both MGC 80-3 and AGS cells (Figure
   [143]3F and Figure [144]S3A). These data suggested that MYH9 was a
   direct target of transcriptional factor SRF in GC cells (Figure
   [145]3G). To detect whether SRF-MYH9 axis contributed to
   miR-647-associated GC metastasis, we transfected MGC
   80-3/AGS-LV-miR-647 with SRF or MYH9 plasmids, followed by puromycin or
   G418 selections. Western blot analysis showed that upregulation of SRF
   in MGC 80-3/AGS-LV-miR-647 restored SRF and MYH9 expression, while
   overexpression of MYH9 in these cells only contributed to MYH9
   upregulation (Figure [146]3H). Cell 3D migration and invasion assays
   suggested that the migratory and invasive abilities of MGC
   80-3/AGS-LV-miR-647 restored after being transfected with either SRF or
   MYH9 plasmids (Figure [147]3I and S3B). Taken together, miR-647 could
   inhibit the migration and invasion of GC cell lines by targeting
   SRF/MYH9 signaling pathway.

MiR-647 is associated with poor prognosis and negatively related with SRF and
MYH9 expression in GC patients

   To determine whether there was an association between the expression
   levels of miR-647 and its target SRF in human GC tissues, we firstly
   analyzed the data from GEO database. Analysis of [148]GSE36968 showed
   that miR-647 expression tended to be negatively associated with SRF
   mRNA (r[s]=-0.2982; p=0.2293; Figure [149]S1B) and MYH9 mRNA
   (r[s]=-0.5273; p=0.0245; Figure [150]S1B), although the correlation
   between miR-647 and SRF mRNA was not significant. To further verify the
   correlations, we detected the SRF and MYH9 expression of GC tissues and
   their matched normal mucosa in 52 GC patients from cohort 1 by qPCR.
   The results showed that both SRF and MYH9 mRNAs were significantly
   upregulated in GC tissues (Figure [151]S1C), which was supported by
   [152]GSE63288 (Figure [153]S1D), and their expressions were negatively
   associated with miR-647 expression detected by qPCR and IHC (Figure
   [154]S1E-F). Since our studies in vitro suggested miR-647 indirectly
   regulated MYH9 expression, the negative correlation between miR-647 and
   MYH9 mRNA could be attributed to dysregulation of SRF protein by
   miR-647. To confirm the relation between miR-647 expression and SRF
   protein levels, we further detected the SRF protein expression in 109
   GC patients and their corresponding nontumorous tissues by IHC. The
   results showed that SRF protein was significantly upregulated in GC
   (Figure [155]4A and [156]4B), and high expression of SRF protein was
   more likely to be seen in GC with low levels of miR-647 (data from
   qPCR) (Figure [157]4C), suggesting that the upregulation of SRF protein
   mainly resulted from repression of miR-647 in GC. Furthermore, we
   detected the expression of miR-647 by ISH (Figure [158]4D). It was also
   found that miR-647 levels were negatively correlated with the levels of
   SRF and MYH9 protein (Figure [159]4E and [160]4F). Low miR-647
   expression was significantly associated with higher histological grade
   (p=0.028), deeper local invasion (p=0.001), lymph-node metastasis
   (p=0.000) and advanced clinical stage (p=0.024) in Table [161]S5.
   Kaplan-Meier analysis revealed that low-level expression of miR-647 was
   associated with poor overall survival in GC patients (p=0.024; Figure
   [162]4G). A multivariate Cox regression analysis indicated that
   low-level expression of miR-647 was an independent prognostic factor
   for predicting poor survival of GC (Table [163]1).

Figure 4.

   [164]Figure 4
   [165]Open in a new tab

   Expression of miR-647 is inversely correlated with the expression
   levels of SRF and MYH9 in gastric cancer samples. (A)
   Immunohistochemical staining of SRF protein in 109 paired of gastric
   cancer and their corresponding nontumorous tissues. Original
   magnification, 100×. (B) Statistical analysis of SRF expression
   according to the scoring (chi-square test; ***p<0.001.). (C)
   Correlation between SRF expression and miR-647 levels in the 109
   gastric cancer tissue samples. ***p<0.001. (D) miR-647 expression
   patterns in human gastric cancer tissues and normal adjacent epithelial
   tissues as determined by in situ hybridization. Original magnification,
   100× and 200×. DIG labeled LNA probes for U6 and scramble were used as
   positive and negative controls, respectively. (E and F) Expression
   associations of miR-647 with SRF and MYH9 expression in 109 GC
   patients. Two representative cases are shown (E). Original
   magnification, 200×. The percentage of specimens showing low or high
   miR-647 expression in relation to the expression levels of SRF and MYH9
   (F); *p<0.05. (G) Kaplan-Meier survival analysis of miR-647 expression
   in patients with gastric cancer (log-rank test).

Table 1.

   Univariate and Multivariate Analysis of Different Prognostic Parameters
   for GC Patients
   Cohort 1 Cohort 2
   Variables HR (95% CI)^a p Value^b HR (95% CI)^a p Value^b
   Univariate Analysis
   Gender (male vs female) 1.120 (0.654-1.919) 0.680 1.034 (0.603-1.773)
   0.903
   Age (years) (<59 vs ≥59) 1.073 (0.629-1.830) 0.795 1.211 (0.706-2.077)
   0.486
   Tumor size (cm) (<5.5 vs ≥5.5) 1.247 (0.732-2.126) 0.417 1.616
   (0.968-2.699) 0.067
   Location (C/B vs A/W)^c 0.909 (0.528-1.567) 0.733 0.637 (0.377-1.077)
   0.092
   Histological grade (W/M vs P/U)^d 0.526 (0.310-0.891) 0.017 2.465
   (1.367-4.447) 0.003
   pT status (T1/T2 vs T3/T4) 1.172 (0.620-2.215) 0.626 3.726
   (1.347-10.308) 0.011
   pN status (N0 vs N1-3) 2.516 (1.190-5.320) 0.016 2.199 (1.139-4.249)
   0.019
   pM status (M0 vs M1) 2.101 (1.086-4.065) 0.027 2.829 (1.375-5.822)
   0.005
   AJCC stage (I/II vs III/IV) 2.057 (1.106-3.824) 0.023 3.462
   (1.936-6.191) 0.000
   miR-647 expression (low vs high) 0.451 (0.233-0.871) 0.018
   SRF expression (low vs high) 2.090 (1.141-3.826) 0.017 2.426
   (1.327-4.435) 0.004
   MYH9 expression (low vs high) 1.972 (1.077-3.608) 0.028 2.497
   (1.345-4.634) 0.004
   Multivariate Analysis I
   Histological grade (W/M vs P/U)^d 0.450 (0.263-0.772) 0.004
   AJCC stage (I/II vs III/IV) 1.883 (1.009-3.514) 0.047
   miR-647 expression (low vs high) 0.388 (0.198-0.762) 0.006
   Multivariate Analysis II
   Histological grade (W/M vs P/U)^d 0.453 (0.245-0.728) 0.004 1.915
   (1.054-3.479) 0.033
   AJCC stage (I/II vs III/IV) 1.788 (0.956-3.344) 0.069 2.881
   (1.597-5.198) 0.000
   SRF expression (low vs high) 2.321 (1.245-4.327) 0.008 1.922
   (1.046-3.533) 0.035
   Multivariate Analysis III
   Histological grade (W/M vs P/U)^d 0.517 (0.304-0.877) 0.014 1.940
   (1.065-3.535) 0.030
   AJCC stage (I/II vs III/IV) 1.831 (0.979-3.422) 0.058 2.886
   (1.596-5.220) 0.000
   MYH9 expression (low vs high) 1.910 (1.037-3.518) 0.038 2.120
   (1.136-3.956) 0.018
   [166]Open in a new tab

   ^aHazard ratios (HRs) and 95% confidence intervals (CIs) were
   calculated using univariate or multivariate Cox proportional hazards
   regression in SPSS.

   ^bp values were calculated using univariate or multivariate Cox
   proportional hazards regression in SPSS. p values < 0.05 were
   considered to indicate statistical significance

   ^cC, cardia; B, body; A, antrum; W, whole

   ^dW, well; M, moderate; P, poor; U, undifferentiated

Activated SRF/MYH9 axis promotes GC metastasis and is associated with poor
prognosis in GC

   Since miR-647 could inhibit GC cell metastasis by SRF/MYH9 axis in
   vitro, we further detected the activity of this axis in GC. Firstly, we
   transfected MGC 80-3 and MKN45 cells with 3 SRF siRNAs and found SRF
   siRNA2 as the most effective sequence for knocking down SRF expression
   in GC cell lines by western blot (Figure [167]S4). Then, we constructed
   a plasmid containing this sequence, marked as shRNA2, and transfected
   MGC 80-3 cells with it to obtain SRF knocked-down GC cells (MGC
   80-3-SRF KD). Western blot and qPCR analysis showed that MYH9
   expression in MGC 80-3-SRF KD significantly downregulated (Figure
   [168]5A). In addition, we further transfected MGC 80-3-SRF KD with a
   SRF plasmid, and found that both SRF and MYH9 expression were restored
   (Figure [169]5B), which further supported MYH9 was a downstream gene of
   SRF.

Figure 5.

   [170]Figure 5
   [171]Open in a new tab

   SRF/MYH9 signaling increases gastric cancer cell lines invasion and its
   activity is associated with poor survival in gastric cancer patients.
   (A) Western blot and qPCR analyses were used to detect the expression
   changes of SRF and MYH9 in MGC 80-3 cells infected with SRF siRNA2
   plasmid or vector control, which were selected with neomycin. (B)
   Western blot and qPCR analysis were used to detect SRF and MYH9
   expression in the SRF knocked down MGC 80-3 cells (MGC 80-3-SRF-KD)
   infected with SRF plasmid or control. ***p<0.001. (C) Cell 3D invasion
   was tested using transwell chamber invasion assay (Matrigel-coated
   membrane). MGC 80-3 transfected with SRF siRNA2 plasmid, SRF plasmid
   and their corresponding vector controls were detected. CCG-1423
   (7.5μmol/L) and its control Dimethyl sulfoxide (DMSO) were used to
   detect their influences on gastric cancer cell invasion. ***p<0.001;
   **p<0.01. (D) Western blot and qPCR analysis were used to detect the
   expression changes of SRF and MYH9 in the above MGC 80-3 cells. (E)
   Subcellular localization of SRF and MYH9 proteins in gastric cancer
   cell lines (MGC 80-3 and MKN45) and gastric cancer tissues were
   determined by confocal immunofluorescence microscopy. (F to I) The
   coordinate expression of SRF and MYH9 in GC patients. TMA analysis of
   SRF and MYH9 in primary human GC tissues (n=90) and adjacent normal
   tissues (F). Original magnification, 35× and 200×. Kaplan-Meier
   survival analysis of SRF and MYH9 expression (G) in patients with
   gastric cancer (log-rank test) were performed. Spearman rank
   correlation analysis was performed based on the IHC expression score of
   SRF and MYH9 from the TMA (H). Linear regression analysis of SRF and
   MYH9 was performed based on the RNA-seq data from TCGA (I).

   To test whether SRF/MYH9 axis regulated by Rho signaling pathway, we
   treated GC cells with CCG-1423, an inhibitor of RhoA/SRF-mediated gene
   transcription [172]^32. Firstly, we detected its IC50 in MGC 80-3 and
   AGS cells. The results showed that the IC50 of CCG-1423 in MGC 80-3 was
   5.22 uM and in AGS was 5.08 μM (Figure [173]S5A). In addition, before
   the function assays, GC cells (MGC 80-3 and AGS) were exposed to
   CCG-1423 with a linear concentration gradient, and 7.5uM CCG-1423 was
   found to be a better concentration due to its less toxic compared with
   9uM in MGC 80-3 (Figure [174]S5B). Intriguingly, we found that CCG-1423
   effectively inhibited GC cells' invasion and MYH9 expression (Figure
   [175]5C-D and S5C), which suggested that SRF/MYH9 axis in GC was
   mediated by RhoA signaling pathway. In addition, the dysregulation of
   SRF expression by directly targeting to SRF gene could change GC cells'
   invasion. We also detected SRF and MYH9 expressions in GC cell lines
   (MKN45 and MGC 80-3) and GC tissues by confocal immunofluorescence
   microscopy. The results showed both proteins broadly located in
   cytoplasm and nucleus of GC cells (Figure [176]5E).

   To further investigate the clinicopathological and prognostic
   significance of SRF and MYH9 in GC patients, we analyzed two GC cohorts
   by IHC, which included paired samples of normal gastric mucosa and GC
   tissues from 109 patients in our hospital (cohort 1) and National
   Engineering Center for Biochip in Shanghai (TMA; Figure [177]5F; cohort
   2). High SRF expression was positively correlated with higher
   histological grade (p=0.031), deeper local invasion (p=0.016),
   lymph-node metastasis (p=0.000), distant metastasis (p=0.017) and
   advanced clinical stage (p=0.009) in cohort 1 (p<0.05; Table [178]S6),
   and positively correlated with lymph-node metastasis (p=0.032) in
   cohort 2 (p<0.05; Table [179]S7). Meanwhile, high MYH9 expression was
   positively related with lymph-node metastasis (p=0.000) and distant
   metastasis (p=0.042) in cohort 1 (p<0.05; Table [180]S6), and
   positively related with lymph-node metastasis (p=0.040), advanced
   clinical stage (p=0.043) and tumor location at antrum or whole stomach
   (p=0.038) in cohort 2 (p<0.05; Table [181]S7). Kaplan-Meier analysis
   showed that high levels of SRF and MYH9 expression were correlated with
   poor overall survival in both cohorts (Figure [182]5G and S6A-B). And
   miR-647, SRF and MYH9 expression levels were correlated with
   disease-free survival in cohort 1 (Figure [183]S6C-E), respectively.
   Furthermore, multivariate Cox regression analysis revealed SRF and MYH9
   expression to be an independent prognostic factor for poor survival
   (Table [184]1). Moreover, the expression of MYH9 protein was positively
   correlated with the expression of SRF protein (Figure [185]5H and
   S6F-G), which was further supported by analysis of these two mRNAs'
   expression data extracted from TCGA and GEO database (Figure [186]5I
   and S1Ga-c), and data detected by qPCR in 52 GC patients involved in
   cohort 1 (Figure [187]S1Gd).

Ectopic expression of miR-647/SRF regulates tumor growth and metastasis by
orthotropic implantation of GC in vivo

   To further explore the role of miR-647 on tumor metastasis in vivo, MGC
   80-3 cells stably expressed LV-NC/vector, LV-miR-647/vector and
   LV-miR-647/SRF were transplanted subcutaneously into nude mice, and the
   growth of resultant primary tumors then was monitored. After growth to
   an average size of 1 cm^3, tumors were orthotopically sutured to the
   stomach of nude mice (Figure [188]S7) [189]^33. Mice were sacrificed at
   the humane end point at two months after operation. The results showed
   metastatic lesions in the intestine, liver and abdominal wall (Figure
   [190]6A). Among these three groups, the tumor volumes of
   LV-miR-647/vector group were significantly smaller than the control
   group (LV-NC/vector group), and the tumor volumes of LV-miR-647/SRF
   group were relatively bigger than the LV-miR-647/vector group (Figure
   [191]6B). Additionally, in the group of mice bearing tumors of MGC
   80-3-LV-NC/vector cells (control group), 100% (7 of 7), 85.7% (6 of 7)
   and 71.4% (5 of 7) of mice had intestinal, hepatic and abdominal wall
   metastases, respectively (Figure [192]6C). However, in the group of
   mice bearing tumors of MGC 80-3-LV-miR-647/vector cells, 75% (6 of 8),
   25% (2 of 8) and 25% (2 of 8) of mice had intestinal, hepatic and
   abdominal wall metastases, respectively (Figure [193]6C). As expected,
   in the group of mice bearing tumors of MGC 80-3-LV-miR-647/SRF cells,
   100% (8 of 8), 87.5% (7 of 8) and 75% (6 of 8) of mice had intestinal,
   hepatic and abdominal wall metastases, respectively (Figure [194]6C).

Figure 6.

   [195]Figure 6
   [196]Open in a new tab

   Agomir-647 and CCG-1423 synergistically inhibit tumor growth and
   metastasis of gastric cancer in vivo. (A-C) In vivo gastric cancer
   tumourigenesis analysis in the xenograft nude mouse models using
   gastric cancer cells (LV-NC/vector, LV-miR-647/vector and
   LV-miR-647/SRF). (A) White-light images of orthotopic tumors,
   intestinal, hepatic and peritoneal metastases of mice. Yellow arrows
   indicated metastatic nodules. Primary tumor volumes (B) and numbers of
   metastatic intestinal, hepatic and peritoneal nodules (C) per group
   were analyzed. (D-F) CCG-1423 (0.15 mg/kg/d, intraperitoneally) or
   vehicle alone (DMSO) was used to treat orthotopic-transplant nude-mouse
   for 4 weeks (2 procedures). Primary tumor volumes (E) and numbers of
   metastatic intestinal, hepatic and peritoneal nodules (F) per group
   were analyzed. (G-I) CCG-1423 (2 weeks per procedure, 2 procedures in
   all) and miR-647 agomir (80mg/kg/3d, tail vein injection, 5 times per
   procedure, 2 procedures) were combined to treat orthotopic-transplant
   nude-mice. Primary tumor volumes (H) and numbers of metastatic
   intestinal, hepatic and peritoneal nodules (I) per group were analyzed.
   *p<0.05, **p<0.01, ***p<0.001. a, primary tumor; b, intestinal
   metastatic nodules; c, hepatic metastatic nodules; d, peritoneal
   metastatic nodules.

   Lastly, the miR-647 expression in primary tumors of three groups was
   detected by qPCR, and the SRF and MYH9 expression in primary tumors and
   their metastatic nodules (4 intestinal metastases, 2 hepatic metastases
   and 2 abdominal wall metastases per group) were detected by IHC (Figure
   [197]S8A-B and S9A-B). The qPCR data showed that LV-miR-647/vector
   group and LV-miR-647/SRF group had significantly higher levels of
   miR-647 expression, and there was no obviously different between these
   changes in vivo and the data from cell lines used for orthotropic
   implantation (Figure [198]S8A). Meanwhile, the IHC data showed SRF and
   MYH9 expression was negatively correlated with miR-647 expression
   (Figure [199]S8B). Although the SRF expression levels in randomly
   chosen metastatic nodules had no significantly difference between
   LV-miR-647/vector group and LV-miR-647/SRF group, there was expected
   tendency in two groups. Besides, all primary tumors and metastatic
   lesions were confirmed by HE staining (Figure [200]S8G). These results
   further suggested that miR-647 inhibited tumor metastasis in vivo by
   downregulating SRF expression.

Agomir-647 and CCG-1423 synergistically inhibited GC metastasis in vivo

   Since CCG-1423 (a SRF inhibitor) [201]^32^, [202]^34 could downregulate
   MYH9 expression in our study, we carried out the CCG-1423 treatment
   study in orthotopic-transplant nude-mouse models of GC [203]^33. Two
   weeks after orthotopic-transplant surgery, the mice were treated with
   CCG-1423 (0.15 mg/kg/d, intraperitoneally) for 2 weeks [204]^34. Then,
   the same procedure was repeated at six weeks after surgery. The mice
   were sacrificed at 2 months after surgery. As shown in Figure
   [205]6D-F, the tumor volumes of CCG-1423-treated group were smaller
   than the control group, and the metastatic nodules were also
   significantly less in CCG-1423-treated group. The qPCR data showed that
   the CCG-1423-treated group and its control group had no significantly
   different levels of miR-647 expression, and this expression was not
   obviously different with the expression of cell lines used for
   orthotropic implantation (Figure [206]S8C). Meanwhile, the IHC data
   showed SRF expression was no significantly different between two
   groups, but MYH9 expression was obviously downregulated in
   CCG-1423-treated group (Figure [207]S8D and S9B).

   Agomir, an engineered miRNA mimic, is well confirmed to be effective in
   mice diseases models [208]^35^-[209]^37. Since CCG-1423 and miR-647
   inhibited GC migration and invasion by different targets of
   Rho/SRF/MYH9 axis (Figure [210]S10), we further detected the
   synergistic inhibitory effect of CCG-1423 and agomir-647 on GC
   metastasis. The detailed procedure for constructing
   orthotopic-transplant nude-mouse models and drug administration were
   shown in Figure [211]S7. As shown in Figure [212]6G-I, compared with
   DMSO-NC, CCG-1423-NC, DMSO-agomir-647 groups, the primary tumors
   volumes and numbers of metastatic nodules in combined treatment group
   (CCG-1423, 0.15 mg/kg/d, intraperitoneally; agomir-647 80mg/kg/3d, tail
   vein injection) were smaller and less, respectively (Figure [213]6G-I).
   Meanwhile, the qPCR data showed that miR-647 expression in DMSO-NC and
   CCG-1423-NC groups was no obviously different with its expression in
   cell lines used for orthotropic implantation, and miR-647 expression in
   primary tumors in DMSO-agomir-647 and CCG-1423-agomir-647 groups were
   significantly upregulated compared with its expression in DMSO-NC and
   CCG-1423-NC groups (Figure [214]S8E), respectively. We also detected
   SRF and MYH9 protein expression in primary tumors and metastatic
   nodules. Firstly, SRF expression in CCG-1423-treated groups
   (CCG-1423-NC and CCG-1423-agomir-647 groups) was not significantly
   changed compared with their own controls (DMSO-NC and DMSO- agomir-647
   groups), but SRF expression in agomir-647-treated groups
   (DMSO-agomir-647 and CCG-1423-agomir-647 groups) was significantly
   downregulated compared with their own controls (DMSO-NC and CCG-1423-NC
   groups) (Figure [215]S8F and S9B). In addition, MYH9 expression in
   agomir-647-treated or CCG-1423-treated groups was downregulated
   compared with their respective controls. MYH9 expression was not
   significantly different between CCG-1423-NC and CCG-1423-agomir-647
   groups. To be sure, all primary tumor, intestinal metastasis, hepatic
   metastasis and peritoneal metastasis were confirmed by HE staining
   before qPCR and IHC studies (Figure [216]S8G). The expression levels of
   SRF and MYH9 were evaluated by IHC staining (Figure [217]S8D, S8F and
   S9B).

Discussion

   Though microRNAs' roles in cancer have been extensively studied, the
   significance and specific mechanism of miR-647 in cancer remains
   unknown. MiR-647 has been confirmed to be upregulated in
   Taxol-resistant ovarian cancer cells, and high expression of miR-647
   was significantly associated with longer survival in chemosensitive
   ovarian cancer patients [218]^16. In prostate cancer, miR-647 has been
   reported to be one of the three genes in the predictive biomarker panel
   negatively associated with recurrence [219]^15. Besides, miR-647 is one
   of the 7 globally population-differentiated miRNAs associated with 37
   diseases including GC, though the function of miR-647 is still unknown
   [220]^17. Meanwhile, relevant studies seemed to have different results.
   For example, the expression of miR-647 was found to be no difference
   between localized stage I non-small cell lung cancer patients with
   recurrence and those without recurrence [221]^23. Even due to its
   stable expression patterns across cases as well as the two case groups,
   miR-647 was suggested to be a potential reference miRNA for future
   stage I non-small cell lung cancer studies. Therefore, different
   cancers, even different stages, may have different miRNA expression
   profiles. Our study demonstrates that miR-647 is frequently
   downregulated in human GC, and functions as a tumor suppressor via
   SRF/MYH9 axis in GC.

   In this study, we first performed qPCR in human GC tissues and
   corresponding normal gastric mucosa to identify miR-647 was associated
   with GC metastasis. Thereafter, using a series of in vitro and in vivo
   assays, we uncovered that miR-647 acted as an important invasion and
   metastasis suppressor in GC. Although miR-647 expression has been
   reported previously in other human cancers [222]^15^-[223]^17^,
   [224]^19^, [225]^23, ours is the first study that provides a novel and
   comprehensive insight into the functional role of miR-647 in GC
   metastasis and Rho-mediated signal pathways (Figure [226]S10).

   We confirmed that miR-647 was frequently downregulated in GC and
   significantly associated with local invasion, advanced clinical stage
   and lymph node-metastasis. After validating miR-647 acts as a
   metastasis-associated miRNA in vitro and in vivo, we performed a
   bioinformatics search for potential mRNAs targeted by miR-647 and
   carried out pathway enrichment analysis for miR-647-associated signal
   pathways and finally focused on Rho-mediated signal pathways due to its
   consentaneous function in cancer metastasis. Intriguingly, by qPCR and
   Luciferase reporter assays, SRF and MYH9, which are involved in
   Rho-mediated signal pathways and predicted to be potential miR-647
   targets, were confirmed to be direct and indirect downstream targets of
   miR-647 in GC cells, respectively.

   We found miR-647 directly targeted SRF mRNA and thus regulated
   Rho-mediated signaling pathways in GC cells. Rho signaling is strongly
   implicated in cancer metastasis [227]^38, and this is generally thought
   to reflect direct effects on the cytoskeleton [228]^39. In addition to
   the direct cytoplasmic role of Rho effectors in cytoskeletal
   regulation, Rho GTPases also control activity of MAL/MRTF-A and
   MKL2/MRTF-B, which are transcriptional co-activators of SRF [229]^40^,
   [230]^41. SRF is an important regulator of genes through binding to the
   sequence CC(A/T)[6]GG termed the CArG box [231]^42, which is involved
   in immediate early and tissue-specific gene expression, proliferation,
   differentiation, apoptotic and cancer metastasis [232]^25^, [233]^28^,
   [234]^42. Recently, SRF has been reported to be upregulated in GC and
   modulate the epithelial to mesenchymal transition by regulating
   miR-199a-5p expression [235]^28. However, little is known about the
   mechanism why SRF is overexpressed in GC tissues. In this study, we
   found miR-647, which was significantly downregulated in GC, could lead
   to the loss of inhibition of SRF via 3'UTR of SRF mRNA. The expression
   of miR-647 is also negatively associated SRF levels in GC tissue
   samples. In addition, our study also confirmed that MYH9 was a directly
   target of SRF in GC cells. MYH9, known as non-muscle myosin heavy chain
   IIA, was a known effector of Rho-ROCK signaling pathway and frequently
   found to be upregulated in GC [236]^27^, [237]^43. In our study, forced
   change of miR-647 expression not only led to SRF but also MYH9
   dysregulation. Overexpression of MYH9 in LV-miR-647-infected GC cells
   could increase cells' migration and invasion. Thus, MYH9 was a critical
   downstream molecule of SRF in GC cells.

   Currently, it is urgent to uncover a treatment strategy for GC
   metastasis, especially peritoneal metastasis. Our in vitro results
   showed that miR-647/SRF/MYH9 axis could effectively inhibit GC
   metastasis. MicroRNAs are thought to be promising next generation
   therapeutic targets in human diseases [238]^44^, [239]^45. Moreover,
   agomir is an engineered miRNA mimic, and has been well confirmed to be
   effective in mice diseases models [240]^35^-[241]^37. CCG-1423, an
   inhibitor of Rho/MKL1/SRF signaling, has been used to improve glucose
   uptake and tolerance in insulin-resistant mice [242]^34. We further
   used the combined treatment strategy of agomiR-647 and CCG-1423 in
   vivo, which suppressed GC invasion and migration in different targets
   of miR-647/SRF/MYH9 axis. The proliferation and metastasis of
   orthotopic-transplant nude-mouse models of GC were dramatically
   inhibited. This synergistically inhibitory effect in vivo strengthened
   the potential clinically significance of miR-647/SRF/MYH9 axis in GC
   metastasis.

   Our study has a few caveats. Since one miRNA often targets to many
   genes to regulate cell behavior, other molecules might also contribute
   to the invasion and metastasis suppressing effect of miR-647. ZAK, one
   of the predicted genes, was a factor of aggressiveness in cancer cells
   [243]^46. Our study found that the expression of ZAK mRNA was not
   significantly upregulated in AGS cells treated with miR-647 inhibitor,
   though its significant upregulation in MGC 80-3 cells treated with
   miR-647 mimics. Moreover, though miR-647 is negatively correlated with
   overall survival of GC patients, the potential clinical application of
   miR-647 remains a subject for further study. Whether it can be detected
   in patients' body fluids and serve as a diagnostic indicator are still
   unknown.

   In conclusion, miR-647 is a potent tumor metastasis suppressor in the
   GC cells, and its migration and invasion inhibitory effects are, in
   part, mediated through its downstream target gene, SRF. MiR-647 may
   have a therapeutic potential to suppress GC metastasis by inhibiting
   Rho associated SRF/MYH9 signaling pathway.

Acknowledgments