Circular RNA 0001789 sponges miR-140-3p and regulates PAK2 to promote the progression of gastric cancer

Normal gastric mucosa tissue samples of healthy controls (n = 50) and gastric mucosa tissues of patients with GC (n = 70) were collected at The First Affiliated Hospital of Xiamen University (Xiamen, China). The controls were healthy individuals with matched ages and sex to those of the patients with GC. Patients who underwent chemotherapy or radiotherapy were excluded from the study. All the enrolled participants had provided informed consent. The study protocols were approved by the Research Ethics Committee of The First Affiliated Hospital of Xiamen University, and the study was carried out following the Declaration of Helsinki.

To identify the differentially expressed circRNAs between GC tissues and normal gastric mucosa tissues, a published circRNA microarray dataset (GSE83521) was extracted, which contained 50 gastric mucosa samples from healthy controls and 70 samples from patients with GC. Cufflinks v2.2.1 was used to derive expression (fragments per kilobase million) values, and differential gene expression was performed with the Cuffdiff package. False Discovery Rate (FDR)-adjusted P-value following Benjamini-Hochberg correction for multiple comparisons was used to define differential expression. CircRNAs with FDR-adjusted P < 0.05 and log2 fold-change  > 1 were considered as significantly differentially expressed.

All cell lines were purchased from American Type Culture Collection. Primary antibodies against E-cadherin, N-cadherin, vimentin, VEGF-A, PAK2, Ki67, CD63, tumor susceptibility gene 101, calnexin and CD34 were purchased from Abcam. A chemiluminescence imaging kit was obtained from Beyotime Institute of Biotechnology. RPMI-1640, Ham’s F-12 K and MEM media, as well as fetal bovine serum (FBS) were obtained from Gibco (Thermo Fisher Scientific, Inc.). Short hairpin RNA (shRNA) (sh-circ#1, sh-circ#2 and sh-circ#3) and sh-negative control (NC), as well as pcDNA-PAK2, vector-exo, circ_0001789-exo, empty vector (vector), miRNA controls (miR-NC), and miR-430-3p mimic or inhibitor (miR-430-3p or anti-miR-430-3p) were purchased from Shanghai GeneChem Co., Ltd. Lipofectamine™ 2000 transfection reagent was obtained from Thermo Fisher Scientific, Inc. Goat anti-rabbit and mouse horseradish peroxidase (HRP)-conjugated IgG antibodies were purchased from Santa Cruz Biotechnology, Inc. 3-Methyladenine was purchased from Sigma-Aldrich (Merck KGaA). Pierce™ Magnetic RNA-Protein Pull-Down Kit was obtained from Thermo Fisher Scientific, Inc. Primers were synthesized by Integrated DNA Technologies, Inc.

GC cell lines (AGS, HGC27, MKN-45 and MKN-74), primary human umbilical vein endothelial cells (HUVECs) and the human gastric mucosal epithelial cell line (GES-1) were maintained in RPMI-1640, Ham’s F-12 K or MEM media (Gibco; Thermo Fisher Scientific, Inc.) supplemented with 10% FBS, penicillin/streptomycin (1:100; Sigma-Aldrich; Merck KGaA), 4 mM l-glutamine and 0.19% HEPES (Sigma-Aldrich; Merck KGaA) in a 37 °C incubator with 5% CO₂. The usage of primary human cells in the present study was approved by the Research Ethics Committee at The First Affiliated Hospital of Xiamen University.

The shRNA sequences (sh-circ#1, sh-circ#2 and sh-circ#3) were synthesized by Shanghai GeneChem Co., Ltd., and the sequences are summarized in Additional file 2: Table S1. The shRNA sequences were cloned into the pLKO.1-Puro vector for shRNA-mediated gene silencing. To overexpress circ_0001789, the sequence of circ_0001789 was synthesized by Integrated DNA Technologies, Inc. and cloned into the pLCDH-ciR vector (Guangzhou RiboBio Co., Ltd.). The transfection of of circ_0001789 shRNA or expression vector was conducted as follows [20]: Initially, cells were seeded in a 6-well plate for 24 h and allowed to reach 60% confluence. The PLUS^™ reagent of the Lipofectamine^™ 2000 transfection kit was mixed with 3 μg vector in 150 μl serum-free medium at room temperature. Next, the mixtures were combined with an equal volume of serum-free medium containing 6 μl Lipofectamine^™ 2000. After 30 min incubation, the transfection solution was added to the cells, and complete medium with FBS was added 12 h after transfection for an additional 48 h culture before experimental analysis.

The proliferation of the GC cells subjected to different treatments was measured by Cell Counting Kit (CCK)-8 assay. Briefly, GC cells were seeded in 96-well plates at a density of 1 × 10⁴ cells per well, and cultured for 24, 48 and 72 h. CCK-8 solution (1:50 dilution) was added into the cell culture at the indicated time points and incubated for 3 h at 37 °C. Finally, the absorbance at 450 nm was determined with a EnSpire Multi-label Plate Reader (PerkinElmer, Inc.).

Cells (0.25 × 10⁵) seeded in 6-well plates were transfected with biotinylated control or circRNA probe (100 nM) using Lipofectamine™ 2000. After 48 h, cells lysates (from 1 × 10⁶ cells) were collected by using IP Lysis Buffer (Beyotime Biotechnology), and 10% of the lysates was employed as the input. The remaining lysates were incubated with M-280 streptavidin magnetic beads (Sigma-Aldrich) at 4 °C overnight. A magnetic bar was used to precipitate the magnetic beads. After 4 washes with a high-salt wash buffer, both the input and the precipitated samples from the pull-down were purified with TRIzol^® reagent (Invitrogen). The enriched miRNA was then detected by reverse transcription-quantitative PCR (RT-qPCR) analysis.

The sequence containing the wild-type (WT) binding site or the mutated (MUT) binding site was cloned into the firefly luciferase PmirGLO reporter vector (Promega Corporation). The reporter plasmid and Renilla luciferase (hRlucneo) control plasmid were co-transfected into cells in the presence of miRNA mimic or miR-NC in a 12-well plate using Lipofectamine^™ 2000 reagent. At 48 h post-transfection, the relative luciferase activities were measured using Luciferase Reporter Gene Assay System (PerkinElmer, Inc.) on GloMax^® Discover Microplate Reader (Promega Corporation).

Cell lysates (from 1 × 10⁶ cells) collected by 1 mL IP lysis buffer were incubated with 100 μl protein-A magnetic beads (Sigma) coated with 10 μg control IgG or anti-AGO2 antibody at 4 °C overnight. The magnetic beads were precipitated using a magnetic bar, and the precipitated samples were washed 3 times with washing buffer. The eluted samples were purified with TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.). The relative level of precipitated molecules was detected using RT-qPCR analysis and normalized to that in the input samples.

A total amount of 10 µg protein sample was subjected to 10 or 12% SDS-PAGE, followed by the transfer to PVDF membranes (MilliporeSigma). After blocking with 5% skimmed milk, the membrane was incubated with primary antibodies against E-cadherin, N-cadherin, vimentin VEGF-A and actin (1:1000, all from Abcam) at 4 °C. After further probing with HRP-linked secondary antibodies at room temperature for 1 h, the protein bands were visualized using an enhanced chemiluminescence kit (Santa Cruz Biotechnology, Inc.) and photographed on a gel imaging system (Bio-Rad Laboratories, Inc.). Densitometry analysis was performed with Image J software [21].

The animal experiments were approved by the Institutional Animal Care and Use Committee of the First Affiliated Hospital of Xiamen University (Xiamen, China). NU/NU mice (6 weeks old) were obtained from the Experimental Animal Center of Xiamen University. For establishing the xenograft model, 1 × 10⁶ AGS cells transfected with sh-NC and sh-circ-0001789 were injected subcutaneously on the back of the mice. Tumor growth was monitored using a Vernier caliper every 7 days, and tumor volume (V) was calculated as V = length x (width)²/2 [22]. If the tumor xenograft exceeded 2,000 mm³, the mice would be euthanized immediately. For euthanasia, a chamber was connected to a carbon dioxide cylinder, and the flow rate was adjusted to displace 40% of the cage volume per min. Mice were placed into the euthanizing chamber for 10 min until no movement was observed. Animal death was confirmed by subsequent cervical dislocation.

Paraffin-embedded tissues were cut into 4-5 μm sections using a microtome. After de-paraffinization and hydration, antigen retrieval was conducted by boiling the slides in citrate buffer for 90 s, followed by the incubation with 3% hydrogen peroxide for 10 min. After 3 washes in TBST buffer, the sections were blocked for 1 h in TBST with 5% normal goat serum, and then probed with primary antibodies (1:500) overnight at 4 °C. Following washing with TBST buffer, the sections was soaked with 1–3 drops of SignalStain® Boost Detection Reagent (Cell Signaling Technologies, Inc.) and incubated in a humidified chamber for 30 min at room temperature. The signal was developed using 400 μl SignalStain® substrate (Cell Signaling Technologies, Inc.) for 5 min. The sections were mounted with coverslip using mounting medium (Cell Signaling Technologies, Inc.) before being imaged under a bright-field microscope [23, 24].

Transfected cells were collected after 48 h and re-suspended in serum-free DMEM. The Transwell inserts were pre-coated with 1% Matrigel at 37 °C for 1 h. Approximately 5 × 10⁵ cells were inoculated into the upper chamber, while 500 μl 10% serum-containing medium was added to the lower chamber. After 18 h, cells on the membrane were fixed with 4% paraformaldehyde for 10 min and stained with 0.5% crystal violet (Sigma-Aldrich) for 20 min. Cells were photographed under a microscope (Olympus Corporation). Transwell migration assay was performed in Transwell inserts without Matrigel coating, and the other procedures were the same as those described above for the invasion assay.

Total RNA from GC tissues and cells was extracted using RNeasy Mini Kit (Qiagen GmbH) or mirVana miRNA Isolation Kit (Thermo Fisher Scientific, Inc.). Exosomes were purified from serum samples or cell cultures using Total Exosome Isolation Kit (Invitrogen; Thermo Fisher Scientific, Inc.), and total RNA was purified from exosome samples using TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc). Total RNA (5 μg) was used for RT into cDNA using PrimeScript^™ RT Master Mix (Takara Bio, Inc.). qPCR analysis was performed in triplicate using Tiangen Master Mix SYBR Green RT-PCR SuperMix (Tiangen Biotech Co., Ltd.) on a CFX96^™ Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.). Actin and U6 small nuclear RNA were used as internal references to normalize mRNA and miRNA, respectively. The sequences of the primers used in the present study are summarized in Additional file 3: Table S2.

RNase R (Takara Bio, Inc.) was applied to digest the RNA samples in order to evaluate their stability. Total RNA samples were divided equally into two portions: One was digested with RNase R (RNase R + group), while the other one was used as a control (mock group). The two samples were incubated at 37 °C for 25 min. The relative quantity of linear RAB11 family interacting protein 1 (RAB11FIP1) mRNA and circ_0001789 in each sample was detected by RT-qPCR.

For the RNA stability assay, transcription was blocked by 3 μg/ml actinomycin D (Sigma-Aldrich; Merck KGaA) for the indicated duration, and RNA samples were collected with TRIzol®. The stability of circ_0001789 and RAB11FIP1 mRNA was analyzed by RT-qPCR by comparison with the samples before treatment (0 h time point).

An in vitro angiogenesis assay kit (Abcam) was used to evaluate angiogenic potential. In brief, 40 μl extracellular matrix solution was added to a pre-chilled 96-well culture plate and incubated at 37 °C for 20 min. HUVECs (2.0 × 10⁴) were seeded in a coated plate and incubated with exosomes from different cell lines for 48 h. Endothelial cell tube formation was observed with an inverted microscope in a bright field [25]. Image J Angiogenesis Analyzer (National Institutes of Health) was used for quantification of the network structure.

RNAScope kit (Invitrogen; Thermo Fisher Scientific, Inc.) was used to perform FISH according to the manufacturer’s instructions. Briefly, cells cultured on a coverslip were fixed with 4% paraformaldehyde and permeabilized with 0.05% Triton-X100 for 15 min. Next, 50 nM circ_0001789 probe with cyanine 3 fluorescent dye (Guangzhou RiboBio Co., Ltd.) in hybridization buffer was applied for 3 h incubation at 50 °C. Upon washing with TBST buffer, mounting medium containing DAPI (Vector Laboratories, Inc.) was used to mount the cell samples on the slides. The expression intensity and localization of circ_0001789 was observed with a fluorescence microscope.

H&E staining was performed in lung tissues using H&E Stain Kit (Abcam). Deparaffinized/hydrated sections were incubated in a sufficient volume of hematoxylin solution (Mayer’s) to completely cover the tissue section for 5 min. The section was then rinsed twice with distilled water and incubated with Bluing Reagent for 60 s. Upon washing, the section was dehydrated in absolute alcohol and stained with eosin Y-solution for 3 min. The sections were rinsed in absolute ethanol 3 times, and images were collected under an inverted microscope.

Data are expressed as the mean ± SD, and were analyzed with GraphPad Prism 9.0 (GraphPad Software, Inc.). Statistical significance between two groups was assessed with paired two-tailed Student’s t-test, while comparisons among multiple groups were performed with one-way or two-way ANOVA followed by Tukey’s post hoc test. χ² test was applied to investigate the association between GC and the circ_0001789 expression level. Kaplan Meier Curve and log-rank test were used to compare the cumulative survival rates in patients. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001.

To identify the differentially expressed circRNAs between GC tissues and normal gastric mucosa tissues, a published circRNA microarray dataset (GSE83521) was extracted. Differential expression analysis (log2 fold-change > 1, adjusted P < 0.05) showed that a number of circRNAs were downregulated (as shown in blue) and upregulated (as represented in red) in GC samples (Fig. 1A). Circ_0001789 was identified as one of the most significantly upregulated circRNAs in GC samples (Fig. 1A, B). Since the functional role of circ_0001789 in GC progression is unknown, this circRNA was selected for further investigation. To confirm its expression in GC tumors, gastric mucosa tissue samples were collected from healthy controls (n = 50), while GC tissues were collected from patients with GC (n = 70). RT-qPCR validated that circ_0001789 was highly expressed in GC tissues compared with its expression in normal gastric mucosa samples (Fig. 1C). Compared with that of samples from patients without metastasis, circ_0001789 showed a relatively higher expression level in patients with metastasis (Fig. 1D). To analyze the association of circ_000178 level and patient prognosis, patients with GC were divided into high (n = 35) and low (n = 35) expression groups based on the median expression value of circ_000178 in the 70 patients with GC. The survival rate in patients with GC exhibiting high circ_0001789 expression levels was worse than that of patients with GC and low expression of circ_0001789, as showed by Kaplan-Meier analysis of the survival curve represented in Fig. 1E.

The expression level of circ_0001789 was significantly higher in GC cell lines (AGS, HGC27, MKN-45 and MKN-74) than in normal gastric epithelial cells (GES-1) (Fig. 1F). The association between circ_0001789 level and the clinical features of patients with GC was also examined. As shown in Table 1, a high level of circ_0001789 expression was significantly associated with advanced TNM stage and distal node metastasis. However, there was no significant difference in patient’s age, sex or tumor size between the circ_0001789 low and high-expression groups.

As a circRNA, sequence alignment showed that circ_0001789 was formed by the back splicing of exon2 of RAB11FIP1 (Fig. 1G). To verify its circular structure, transcription was first blocked by actinomycin D, and the RNA decay rate of RAB11FIP1 and circ_0001789 was analyzed. RT-qPCR analysis showed that the RAB11FIP1 mRNA level decreased rapidly following actinomycin D treatment, while circ_0001789 remained relatively stable upon transcription inhibition (Fig. 1H). In addition, RNase R digestion markedly reduced the level of RAB11FIP1 mRNA in the AGS and HGC27 cell lines, whereas it showed little effect on circ_0001789 (Fig. 1I). These data suggested that circ_0001789 possessed a closed-loop structure that was stable and resistant to RNase R degradation.

As exosome-derived circRNAs are involved in cell–cell communication, it was hypothesized that circ_0001789 participates in cell-to-cell communication via exosomes. Exosomes were isolated from the serum of normal individuals and patients with GC, and their morphology was examined using transmission electron microscopy (Fig. 2A) and the enumerated diameter (Fig. 2B), which revealed that the exosomes were ~ 100 nm in diameter. RNA was purified the from the exosome samples of healthy controls and patients with GC, and a relatively higher expression of circ_0001789 was observed in the serum samples of patients with GC (Fig. 2C). In addition, the increased expression of exosome circ_0001789 could be a diagnostic indicator, as the area under the curve value of the receiver operating characteristic curve was > 0.8 (Fig. 2D). As shown in Fig. 2E, western blot analysis indicated that the expression level of the exosome marker CD63 was higher in GC cell lines than in GES-1 cells.

Since MKN-45 cells expressed high levels of exosomal markers and exosomal circ_0001789 (Fig. 2E, F), MKN-45 cells were used as donor cells of exosomes, while AGS and HGC27 cells were used as recipient cells. MKN-45 cells were transfected with control and circ_0001789 plasmids, and the circ_0001789 plasmid significantly increased the cellular and exosomal levels of circ_0001789 (Fig. 2G).

Exosomes from MKN-45 cells transfected with control vector (vector-exo) and circ_0001789 plasmid (circ_0001789-exo) were collected and used to treat AGS and HGC27 cells. Circ_0001789-exo incubation significantly increased the cellular level of circ_0001789 in AGS and HGC27 cells (Fig. 2H). In addition, circ_0001789-exo incubation promoted the proliferation of AGS and HGC27 cells, as revealed by the results of CCK-8 proliferation assay (Fig. 2I). Transwell migration and invasion assays were performed, which demonstrated that circ_0001789-exo incubation also enhanced the migratory and invasive capabilities of AGS and HGC27 cells (Fig. 2J, K). Furthermore, tube formation assay in HUVECs suggested that exosome-derived circ_0001789 promoted angiogenesis (Fig. 2L). Changes in epithelial-mesenchymal transition (EMT) biomarkers were also examined, and upregulation of N-cadherin, vimentin and VEGF-A was observed, as well as downregulation of E-cadherin in AGS and HGC27 cells upon incubation with circ_0001789-exo (Fig. 2M). These data suggested that exosomal circ_0001789 mediated cell–cell communication to promote the progression of GC.

To verify the biological function of circ_0001789 in GC progression, circ_0001789 was silenced in two GC cell lines (AGS and HGC27) with high expression of circ_0001789 by using shRNAs. As depicted in Fig. 3A, the shRNAs targeting circ_0001789 (sh-circ_0001789#1, sh-circ_0001789#2 and sh-circ_0001789#3) significantly reduced circ_0001789 expression. Among them, sh-circ_0001789#1 showed the strongest knockdown effect and was therefore selected for subsequent experiments.

Notably, knockdown of circ_0001789 showed no effect on RAB11FIP1 mRNA expression (Fig. 3B). CCK-8 proliferation assay revealed that knockdown of circ_0001789 suppressed the proliferation of AGS and HGC27 cells (Fig. 3C). Silencing circ_0001789 also impaired the migration and invasion abilities of AGS and HGC27 cells (Fig. 3D, E). Exosomes from AGS and HGC27 cells transfected with sh-NC or sh-circ_0001789 were also isolated. Incubation of HUVECs with exosomes from sh-circ_0001789-treated cells led to impaired tube formation ability in comparison with that observed in cells treated with exosomes derived from control cells (Fig. 3F). Western blot results showed downregulation of N-cadherin, vimentin and VEGF-A, and upregulation of E-cadherin upon circ_0001789 silencing in AGS and HGC27 cells (Fig. 3G). Silencing experiments were also performed using sh-circ_0001789#2 (Additional file 1: Fig. S1), which showed consistent results with sh-circ_0001789#1. Together, these data showed that circ_0001789 was required to maintain the malignancy of GC cells.

To verify the intracellular localization of circ_0001789 in AGS and HGC27 cells, FISH was performed using circ_0001789 as a probe. As shown in Fig. 4A, circ_0001789 was predominantly located in the cytoplasm of AGS and HGC27 cells. As circRNAs can act as molecular sponges of miRNAs [21], the ‘circBank’ and ‘circInteractome’ databases were employed to search for potential miRNA candidates of circ_0001789. A total of 5 miRNAs [Homo sapiens (hsa)-miR-1246, hsa-miR-1286, hsa-miR-140-3p, hsa-miR-658 and hsa-miR-890] were identified as potential targets of circ_0001789 (Fig. 4B). To validate the interactions, RNA pull-down assay was performed using a biotin-labeled circ_0001789 probe or a control oligo. The circ_0001789 probe significantly enriched miR-140-3p in comparison with the findings in the control oligo probe; however, this result was not observed for other miRNAs (Fig. 4C). To further confirm their interaction, dual luciferase reporter assay was performed using a reporter containing a WT or MUT binding site (Fig. 4D). The results showed that the presence miR-140-3p mimic significantly suppressed the luciferase activity of the circ_0001789 WT reporter, while there was no suppression in the MUT reporter activity (Fig. 4E). Furthermore, AGO2 RIP assay also demonstrated that both miR-140-3p and circ_0001789 were enriched in AGO2-containing complex (Fig. 4F). In comparison with its expression in normal gastric mucosa tissues, miR-140-3p was downregulated in GC tumor tissues (Fig. 4G), and patients with high expression of miR-140-3p showed a better overall survival compared with that of patients in the low-expression group (Fig. 4H). Pearson’s correlation coefficient analysis revealed a significant negative correlation between the expression of circ_0001789 and miR-140-3p in patients with GC (Fig. 4I). Furthermore, the RT-qPCR results revealed that miR-140-3p expression was reduced in GC cell lines (Fig. 4J). Together, these findings indicated that circ_0001789 could act as a sponge to negatively regulate miR-140-3p in GC cells.

To further investigate whether miR-140-3p mediates the downstream effects of circ_0001789, a miR-140-3p inhibitor was applied to suppress the expression of miR-140-3p (Fig. 5A). In the presence of a miR-140-3p inhibitor, the inhibitory effect of circ_0001789 silencing on the proliferation of AGS and HGC27 cells was attenuated (Fig. 5B). Inhibiting miR-140-3p also rescued the migratory and invasive abilities of GC cells upon circ_0001789 silencing (Fig. 5C, D), and the mir-140-3p inhibitor partially increased the tubulogenic ability of HUVECs following circ_0001789 silencing (Fig. 5E). Consistently, the protein levels of N-cadherin, vimentin and VEGF-A were largely rescued by the mir-140-3p inhibitor following circ_0001789 silencing, while the expression level of E-cadherin protein was suppressed by the mir-140-3p inhibitor (Fig. 5F). Therefore, these data suggested that miR-140-3p mediated the downstream effects of circ_0001789.

To further explore the mRNA targets of miR-140-3p, the ‘TargetScan’, ‘miRDB’, ‘TarBase’ and ‘UP-GEPIA’ databases were employed to predict the downstream mRNA with a potential binding site for miR-140-3p. It was observed that the mRNAs of a disintegrin and metalloproteinase domain-containing protein 10 (ADAM10), PAK2, nuclear transcription factor Y subunit α (NFYA), cullin associated and neddylation dissociated 1 (CAND1), mediator complex subunit 13 and ubiquitin specific peptidase 31 (USP31) contained a potential miR-140-3p site (Fig. 6A). To validate this finding, GC cells were transfected with miR-NC and miR-140-3p mimics, and RT-qPCR was performed to detect the mRNA expression levels of these candidates in AGS and HGC27 cells. As shown in Fig. 6B, overexpression of miR-140-3p could downregulate the mRNA levels of ADAM10 and PAK2, with PAK2 mRNA showing the strongest downregulation. Furthermore, The Cancer Genome Atlas analysis between normal gastric and GC tissues showed that PAK2 was highly expressed in GC (Fig. 6C), which was further confirmed in clinical samples of normal gastric mucosa tissues and GC tumor samples by RT-qPCR (Fig. 6D).

High PAK2 expression was associated with poor prognosis in patients with GC (Fig. 6E). Pearson’s correlation analysis indicated that the expression levels of PAK2 and circ_0001789 were positively correlated (Fig. 6F), while the expression levels of PAK2 and miR-140-3p showed a negative correlation in the samples of patients with GC (Fig. 6G). Consistently, PAK2 was upregulated in GC cells in comparison with its expression in GES-1 cells (Fig. 6H). A luciferase reporter containing a WT or MUT binding site between PAK2 mRNA and miR-140-3p was also constructed, and dual luciferase reporter assay showed that overexpression of mir-140-3p could significantly inhibit the luciferase activity of the WT reporter, which was not observed in the MUT reporter (Fig. 6J). Knockdown of circ_0001789 also produced a negative effect on the expression level of PAK2 protein, which could be rescued by a miR-140-3p inhibitor (Fig. 6K).

To demonstrate that PAK2 mediates the effect of miR-140-3p, cells were transfected with miR-140-3p mimics that could downregulate PAK2 in GC cells (Fig. 7A), and co-transfection with the pcDNA-PAK2 vector increased PAK2 expression. Although overexpression of miR-140-3p inhibited cell proliferation, co-transfection with the pcDNA-PAK2 expression vector partially rescued cell proliferation in the presence of miR-140-3p mimics (Fig. 7B). Similarly, PAK2 overexpression also rescued the migratory and invasive abilities of GC cells following miR-140-3p overexpression (Fig. 7C, D), and PAK2 overexpression partially increased the tubulogenic ability of HUVECs upon miR-140-3p overexpression (Fig. 7E). Consistently, the protein levels of N-cadherin, vimentin and VEGF-A were reduced by miR-140-3p mimics, but largely rescued by PAK2 overexpression, while the expression level of E-cadherin protein was suppressed by PAK2 overexpression (Fig. 7F). Together, these data indicated that PAK2 mediated the functional role of miR-140-3p in controlling the malignancy of GC cells.

To further validate the oncogenic role of circ_0001789 in vivo, AGS cells transfected with sh-NC and sh-circ-0001789 were injected subcutaneously into nude mice. The tumor growth curve showed that, compared with the effects of cells transfected with sh-NC, circ_0001789 silencing in the sh-circ_0001789 group attenuated the increase in tumor volume and reduced the tumor weight in vivo (Fig. 8A, B). Consistent with the in vitro data, silencing circ_0001789 led to the upregulation of mir-140-3p and the downregulation of PAK2 in the isolated xenograft samples (Fig. 8C).

IHC staining of different markers was performed, which showed that the expression levels of Ki-67, PAK2, CD34 and N-cadherin were significantly reduced in the silencing group compared with those in the sh-NC group (Fig. 8D). In addition, H&E staining was performed in lung tissues to examine metastasis. In the group of GC tumor with circ_0001789 silencing, the number of observed metastatic nodules in lung tissues was lower than that in the sh-NC group (Fig. 8E). These data supported that circ_0001789 was an oncogenic factor supporting the tumorigenesis and metastasis of GC in vivo.