Inhibition of circRNA circVPS33B reduces cell malignant behaviors and Warburg effect through regulation of the miR-873-5p/HNRNPK axis in inltrative gastric cancer

Circular RNA VPS33B (circVPS33B) is upregulated in gastric cancer (GC) tissues. However, the role of circVPS33B in inltrative GC is indistinct. Expression of circVPS33B, miR-873-5p, and heterogeneous nuclear ribonucleoprotein K (HNRNPK) mRNA was detected using quantitative real-time polymerase chain reaction (qRT-PCR). The proliferation, colony formation, migration, and invasion of inltrative GC cells (XGC-1) were determined by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbromide (MTT), plate clone, wound healing, or transwell assays. Several protein levels were examined by western blotting. The extracellular acidication rate (ECAR) and oxygen consumption rate (OCR) of XGC-1 cells were evaluated by XF96 extracellular ux analyzer. Glucose uptake and lactate production were analyzed by glycolysis assay. The relationship between circVPS33B or HNRNPK and miR-873-5p was veried by dual-luciferase reporter and/or RNA pull-down assays. In vivo tumorigenesis assay was executed for verifying the in vitro results.

4. HNRNPK served as a target for miR-873-5p in XGC-1 cells. 5. The study provided a new mechanism to understand the progression of in ltrative GC.

Background
Gastric cancer (GC), a malignant tumor that occurs in the stomach, ranks third among cancer-related death causes [1]. GC is divided into in ltrative and expanding types according to the growth and invasive patterns based on Ming's classi cation [2]. Moreover, the prognosis of in ltrative GC is worse than that of expanding GC [3]. Studies have reported that the recurrence of the peritoneal or liver after radical resection of GC is closely related to the pathological in ltrative pattern (INF) [4,5]. Therefore, it is of great signi cance to explore the latent molecular mechanism of in ltrative GC progression. Circular RNAs (circRNAs), endogenous non-coding RNAs, have closed-loop structures formed by reverse splicing of mRNA exons or introns [6]. They are more stable than other conventional linear RNAs due to their resistance to exonucleases [7]. Studies have revealed that circRNAs exert diverse vital roles in cellular physiology through serving as protein translation templates, transcription regulators, RBP-binding molecules, or microRNA (miR) sponges [8]. Also, circRNAs dysregulation are closely related to the advancement of diverse diseases, including cancers [9]. For example, circRNA hsa_circ_0003141 contributed to the tumorigenesis of hepatocellular cancer [10]. CircRNA VPS33B, also termed as hsa_circ_0005529, is derived from the vacuolar protein sorting 33 homologue B (VPS33B) gene. It was reported that circVPS33B was elevated in GC tissues (GSE78092) [11]. However, the role of circVPS33B in in ltrative GC progression is unclear.
MiRs can regulate the expression of genes via binding to the 3' Untranslated Regions (UTR) of target mRNAs [12]. They act as powerful regulatory factors in a series of cellular activities, including cell development, differentiation, growth, and apoptosis [13]. It has been revealed that miRs exert important roles in numerous cases of cancers [14]. For instance, the miR-34 family served as potential therapeutic candidates in cancers [15]. It was reported that miR-873-5p acted as an epigenetic modulator in the early stages of liver brosis and cirrhosis [16]. Studies have shown that miR-873-5p played an inhibitory role in numerous cancers development, such as glioma [17], papillary thyroid cancer [18], and colorectal cancer [19]. Moreover, miR-873-5p was uncovered to curb the malignancy of GC cells [20,21]. However, the mechanism of miR-873-5p in in ltrative GC progression has not been fully clari ed.

Wound healing and transwell assays
The migration of XGC-1 cells was analyzed by wound healing and transwell assays. For wound healing assay, a wound on the cell layer was created using a pipette tip (100 μL) when the transfected XGC-1 cells reached 90% con uence. The pictures of the cells at 0 h and 24 h were photographed with an inverted microscope (Nikon). The migration rate was calculated according to the following equation: cell migration rate (%) = (1-the distance following healing/the distance prior to healing) ×100%.
For transwell migration assay, the serum-free containing transfected XGC-1 cells (1×10 5 cells) was added to the top chamber of the transwell chamber (8 μm, Costar, Cambridge, MA, USA). The cell medium containing 10% FBS was supplemented into the bottom of the transwell chamber. After removing the cells on the upper surface of the membrane, the remaining cells were xed and stained with paraformaldehyde (4%, Sigma) and crystal violet (0.5%, Sigma), respectively. The migrating cells were calculated with an inverted microscope (Nikon) at 100 × magni cation.
For transwell invasion assay, its method was the same as the cell migration assay. It was worth noting that the transwell chamber of the invasion assay was pre-coated with Matrigel (Sigma).
Analysis of extracellular acidi cation rate (ECAR) and oxygen consumption rate (OCR) The XF96 Extracellular Flux analyzer (Seahorse Bioscience, Chicopee, MA, USA) was utilized for the assessment of the ECAR and OCR of transfected XGC-1 cells. In short, the cells (1×10 4 ) were seeded into a Seahorse XF 96 cell culture microplate. The Seahorse XF Glycolysis Stress Test Kit (Seahorse Bioscience) was applied for ECAR analysis. In short, glucose, oligomycin, and 2-deoxy-D-glucose (2-DG) were sequentially injected into each well at the speci ed time point after baseline measurement. The Seahorse XF Cell Mito Stress Test Kit (Seahorse Bioscience) was utilized for OCR evaluation. In brief, oligomycin, p-tri uoromethoxy carbonyl cyanide phenylhydrazone (FCCP), and rotenone (Rote) plus antimycin A (AA) were sequentially injected into each well after baseline measurement. At last, the data were analyzed with the Seahorse XF-96 Wave software, and ECAR and OCR were presented in mpH/min and pmoles/min, respectively.

Glycolysis assay
The supernatant of transfected XGC-1 cells was collected. The amount of lactate and glucose in the supernatant was assessed with a Glucose Assay kit (Sigma) or Lactate Assay kit (BioVision, Mountain View, CA, USA) in light of the manufacturer's instructions. Glucose uptake was analyzed by the difference between the glucose concentration and the control group.

Subcellular fractionation and localization
The nuclear RNA and cytoplasm RNA of XGC-1 cells were isolated by the PARIS kit (Life Technologies).

RNA pull-down assay
The biotinylated (Bio)-miR-873-5p, Bio-miR-NC, and Bio-miR-873-5p-MUT probes were bought from Sigma. The probe-coated bead was established by incubating a probe with the M-280 Streptavidin magnetic bead (Invitrogen). The lysates of XGC-1 cells were incubated with probe-coated beads. The RNA complexes were isolated with TRIzol reagent (Solarbio). QRT-PCR was executed to analyze the enrichment level of circVPS33B in RNA complexes.

In vivo tumorigenesis assay
The protocols of xenograft assay were authorized by the Animal Ethics Committee of the Institute of Gastrointestinal Oncology, School of Medcine, Xiamen University. 10 BALB/c nude mice (4-week-old) were bought from Vital River Laboratory Animal Co., Ltd. (Beijing, China). For xenograft assay, XGC-1 cells (1×10 6 ) with sh-circVPS33B or sh-NC were subcutaneously injected into the right ank of BALB/c nude mice. Tumor volume was measured with calipers every 7 days from the second week after injection. Until day 35, the mice were killed by cervical dislocation under iso urane (5%) to take tumor tissues. Tumors volume was counted in light of the following equation: Volume = (length ×width 2 )/2. The mice were randomly divided into 2 groups with ve mice per group.

Statistical analysis
Correlation among circVPS33B, HNRNPK, and miR-873-5p in in ltrative GC tissues were evaluated with the Pearson's correlation analysis. Differences were deemed signi cant if P < 0.05. One-way variance analysis (ANOVA) with Turkey's post hoc test was utilized to assess the differences among 3 or more groups. Paired or unpaired Student's t test was applied to compare the difference between 2 groups. Statistical analysis was implemented with GraphPad Prism 6 software (GraphPad, San Diego, CA, USA). The experiments in vitro were repeated at least 3 times, and data were exhibited as the mean ± standard deviation.

Results
Expression and characteristics of circVPS33B in in ltrative GC In the research, we rst searched the public database (GES78092) and noted that there were 199 differentially expression circRNAs in GC tissues (Fig. 1A). Among the differentially expression circRNAs, we selected the top 10 upregulated and 10 downregulated differentially expression circRNAs to draw a heat map, as exhibited in Fig. 1B. Moreover, we selected circVPS33B (hsa_circ_0005529) as the target for further investigation. CircVPS33B is a 193 bp transcript derived from the VPS33B gene (Fig. 1C). Next, we detected the expression of circVPS33B in 37 paired in ltrative GC tissues and neighboring non-tumor tissues. QRT-PCR results also displayed that circVPS33B expression was elevated in in ltrative GC tissues in contrast to the neighboring non-tumor tissues (Fig. 1D). Also, circVPS33B expression was increased in XGC-1 cells compared to the GES-1 cells (Fig. 1E). We also explored the characteristics of circVPS33B in XGC-1 cells. Actinomycin D assay revealed that circVPS33B was more stable than linear VPS33B in XGC-1 cells, implying that circVPS33B can be a suitable prognosis and diagnostic marker (Fig. 1F). RNase R digestion assay presented that circVPS33B was resistant to RNase R, whereas the linear VPS33B was degraded by RNase R treatment (Fig. 1G). These data indicated that high circVPS33B expression might be associated with the development of in ltrative GC.
Inhibition of circVPS33B reduced the malignancy and Warburg effect of in ltrative GC cells Subsequently, we investigated the biological role of circVPS33B in in ltrative GC via loss-of-function assays. We designed the siRNA targeting circVPS33B and transfected into XGC-1 cells. Compared to the control si-NC, circVPS33B expression was apparently reduced in XGC-1 cells after si-circVPS33B transfection ( Fig. 2A). MTT assay exhibited that reduced circVPS33B expression inhibited cell proliferation in XGC-1 cells relative to the control group (Fig. 2B). Plate clone assay displayed that circVPS33B silencing decreased the colony formation number of XGC-1 cells (Fig. 2C). Wound healing assay showed that the migration of XGC-1 cells was repressed by circVPS33B knockdown (Fig. 2D). Transwell assay indicated that circVPS33B knockdown played an inhibitory effect on the migration and invasion of XGC-1 cells (Fig. 2E and 2F). Results of western blotting showed that the levels of ki-67, Ncadherin, and Vimentin were downregulated in circVPS33B-silenced XGC-1 cells, while the level of Ecadherin was upregulated (Fig. 2G). Moreover, we also analyzed the ECAR and OCR of XGC-1 cells with an XF96 extracellular ux analyzer. The results exhibited silenced circVPS33B expression decreased the ECAR of XGC-1 cells and increased the OCR of XGC-1 cells (Fig. 2H and 2I). Furthermore, glycolysis assay presented that downregulation of circVPS33B could reduce glucose uptake and lactate production in XGC-1 cells (Fig. 2J and 2K). Taken together, these data indicated that circVPS33B knockdown could decrease cell malignant behaviors and Warburg effect in in ltrative GC cells.
CircVPS33B was identi ed as a sponge for miR-873-5p in in ltrative GC cells In consideration of the above ndings, we further explored the latent regulatory mechanism of circVPS33B in in ltrative GC. We rst explored the subcellular localization of circVPS33B in XGC-1 cells. QRT-PCR presented that circVPS33B was mainly distributed in the cytoplasmic fraction of XGC-1 cells, indicating that circVPS33B had potential post-transcriptional regulation (Fig. 3A). Moreover, we discovered that miR-873-5p possessed complementary binding sites for circVPS33B through circInteractome database (Fig. 3B). In addition, to analyze the e ciency of miR-873-5p overexpression in XGC-1 cells, qRT-PCR was utilized to measure the level of miR-873-5p. And result showed that the expression level of miR-873-5p was markedly up-regulated after transfection with miR-873-5p (Fig. 3C). Dual-luciferase reporter assay indicated that miR-873-5p overexpression could reduce the luciferase intensity of luciferase reporters with circVPS33B-WT in XGC-1 cells, whereas the luciferase intensity of luciferase reporters with circVPS33B-MUT did not affect (Fig. 3D). RNA pull-down assay presented that circVPS33B could be pulled down by Bio-miR-873-5p probe relative to the Bio-miR-NC and Bio-miR-873-5p-MUT probes (Fig. 3E). Moreover, the public database (GES78091) showed that miR-873-5p was downregulated in GC tissues (Fig. 3F). Also, we observed that miR-873-5p expression was decreased in in ltrative GC tissues compared to the neighboring non-tumor tissues (Fig. 3G). Likewise, miR-873-5p was downregulated in XGC-1 cells than that in GES-1 cells (Fig. 3H). Pearson's correlation analysis indicated that miR-873-5p and circVPS33B had a negative correlation in in ltrative GC tissues (Fig. 3I). Furthermore, decreased circVPS33B expression could elevate miR-873-5p expression in XGC-1 cells (Fig.  3J). However, miR-873-5p mimic had no effect on the level of circVPS33B in XGC-1 cells (Fig. 3K).
Collectively, these ndings manifested that circVPS33B acted as a sponge for miR-873-5p in in ltrative GC cells.
CircVPS33B could regulate HNRNPK expression through competitively binding to miR-873-5p Next, we further investigated whether circVPS33B could regulate the target of miR-873-5p via competitively binding to miR-873-5p through acting as a competing endogenous RNA (ceRNA). We predicted the target of miR-873-5p through cross-analyzing four prediction databases: TarBase, TargetScan, miRDB, and miRWalk. As shown in Fig. 5A, there were 13 genes with complementary binding sites for miR-873-5p. Based on the pre-experiment, we chose HNRNPK for further research. The complementary binding sites between miR-873-5p and HNRNPK were displayed in Fig. 5B. Moreover, dual-luciferase reporter assay revealed that elevated miR-873-5p expression reduced the luciferase activity of the luciferase reporters containing HNRNPK 3'UTR-WT in XGC-1 cells, but there was no overt difference in the luciferase reporters containing HNRNPK 3'UTR-MUT (Fig. 5C). We also observed that HNRNPK mRNA and protein levels were elevated in in ltrative GC tissues compared to the neighboring non-tumor tissues (Fig. 5D and 5E). Consistently, HNRNPK mRNA and protein levels were also increased in XGC-1 cells (Fig. 5F and 5G). Moreover, the expression of HNRNPK mRNA in in ltrative GC tissues had a negative correlation with miR-873-5p (Fig. 5H). Furthermore, miR-873-5p mimic reduced the levels of HNRNPK mRNA and protein in XGC-1 cells, while this decrease was reversed by transfection of pcDNA-HNRNPK ( Fig. 5I and 5J). Also, HNRNPK mRNA expression was positively correlated with circVPS33B in in ltrative GC tissues (Fig. 5K). Downregulation of circVPS33B could reduce the levels of HNRNPK mRNA and protein in XGC-1 cells, but this trend was overturned by miR-873-5p inhibition (Fig. 5L and 5M). Collectively, these results manifested that circVPS33B regulated HNRNPK expression via sponging miR-873-5p in in ltrative GC cells.

HNRNPK overexpression abolished miR-873-5p mimic-mediated effects on the malignancy and Warburg effect of in ltrative GC cells
We further investigated whether miR-873-5p regulated the malignancy and Warburg effect of XGC-1 cells via targeting HNRNPK. The results exhibited that the repressive impact of miR-873-5p mimic on cell proliferation, colony formation, migration, and invasion in XGC-1 cells was reversed by HNRNPK elevation (Fig. 6A-6E). Moreover, miR-873-5p overexpression reduced ki-67, N-cadherin, and Vimentin levels and increased E-cadherin level in XGC-1 cells, but these impacts were overturned by forcing HNRNPK expression (Fig. 6F). Additionally, forced miR-873-5p expression decreased ECAR and elevated OCR of XGC-1 cells, but these tendencies were abolished after HNRNPK overexpression ( Fig. 6G and 6H). Furthermore, elevated HNRNPK expression restored the decrease of glucose uptake and lactate production in XGC-1 cells mediated by miR-873-5p upregulation ( Fig. 6I and 6J). Overall, these data indicated that miR-873-5p regulated the malignancy and Warburg effect of in ltrative GC cells via targeting HNRNPK.

Silence of circVPS33B reduced tumor growth in vivo
To verify the role of circVPS33B on tumor growth in vivo, we injected XGC-1 cells stably transfected with sh-circVPS33B or sh-NC into BALB/c nude mice. The results presented that tumor volume and weight were signally reduced in the circVPS33B silencing group in contrast to the control group ( Fig. 7A and 7B). QRT-PCR revealed that circVPS33B and HNRNPK protein were overtly downregulated in mice tumor tissues of the circVPS33B silencing group, but miR-873-5p expression was apparently increased (Fig. 7C-7E). Together, these data manifested that circVPS33B inhibition could decrease tumor growth in vivo.

Discussion
Mounting studies have indicated that circRNAs are promising biomarkers and targets for disease diagnosis and treatment [28,29]. In the research, we explored the role and regulatory mechanism of circVPS33B in in ltrative GC development.
Increasing evidence proved that circRNAs were connected with the advancement of GC. For instance, circRNA circ_MAT2B contributed to cell growth and glycolysis through sponging miR-515-5p and upregulating HIF-1α in GC [30]. Similarly, circRNA circ_CCDC9 inhibited GC progression via modulating CAV1 expression through sponging miR-6792-3p [31]. In the current work, circVPS33B was upregulated in in ltrative GC tissues and cells. Silence of circVPS33B reduced tumor growth in vivo and repressed the malignancy and Warburg effect of in ltrative GC cells in vitro. Warburg effect, also known as aerobic glycolysis, accelerates cancer cell growth through elevating glucose uptake and lactate production [32]. Wang et al. uncovered that circRNA circ_FOXP1 accelerated the Warburg effect and tumor growth in gallbladder cancer [33]. Therefore, these data manifested that circVPS33B knockdown could reduce tumor growth and Warburg effect in in ltrative GC.
Studies have disclosed that some circRNAs can be used as ceRNAs to modulate the expression of some genes and thus participate in the advancement of certain tumors [10,30,31]. In this study, we discovered circVPS33B served as a sponge for miR-873-5p. MiR-873-5p was revealed to be implicated in certain tumors progression. Zhu et al. uncovered that miR-873-5p could curb colon cancer progression via repressing the TUSC3/AKT pathway [34]. Another research disclosed that miR-873-5p reduced cell invasion and migration through targeting CXCL16 in papillary thyroid cancer [18]. In GC, miR-873-5p expression could be inhibited by lncRNA TDRG1 or lncRNA DDX11-AS1, thereby accelerating the aggressiveness of GC [20,21]. Herein, miR-873-5p expression had a negative correlation with circVPS33B in in ltrative GC tissues. Downregulation of miR-873-5p reversed the repressive impact of circVPS33B downregulation on the malignancy and Warburg effect of in ltrative GC cells. Therefore, we inferred that circVPS33B regulated in ltrative GC progression and Warburg effect via miR-873-5p.
HNRNPK, an oncogene, is involved in the development of tumors. One report revealed that lncRNA pancEts-1 could facilitate neuroblastoma progression by HNRNPK-mediated the stabilization of β-Catenin [35]. Xu et al. proved that HNRNPK promoted cell colony formation and proliferation in lung adenocarcinoma cells [36]. Peng et al. reported that HNRNPK contributed to cell invasion, migration, and proliferation via regulation of the alternative splicing of CD44E in GC cells [26]. In our work, we discovered that HNRNPK was a target of miR-873-5p. Moreover, miR-873-5p regulated in ltrative GC cell malignant behaviors and Warburg effect via HNRNPK. Also, circVPS33B could modulate HNRNPK expression via competitively binding to miR-873-5p. Overall, we concluded that circVPS33B regulated in ltrative GC malignant behaviors and Warburg effect via sponging miR-873-5p and regulating HNRNPK expression.

Conclusion
In sum, circVPS33B played an accelerative role in in ltrative GC progression. Moreover, circVPS33B downregulation reduced in ltrative GC advancement and Warburg effect through modulating HNRNPK expression through sponging miR-873-5p, manifesting that circVPS33B might be a hopeful target for in ltrative GC treatment.

Availability of data and materials
The analyzed data sets generated during the present study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
The present study was approved by the ethical review committee of The Second Hospital of Hebei Medical University. Written informed consent was obtained from all enrolled patients.

Figure 7
CircVPS33B silencing decreased the growth of in ltrative GC cells in vivo. (A and B) Tumor volume and weight were decreased in the sh-circVPS33B group compared to the control group. (C-E) QRT-PCR and western blotting were employed to analyze the levels of circVPS33B, miR-873-5p, or HNRNPK protein in mice tumor tissues of the sh-circVPS33B and sh-NC groups. *P < 0.05.