ebv-circRPMS1 Promotes the Progression of EBV-Associated Gastric Carcinoma via Sam68 Complex Dependent Activation of METTL3

Background: Emerging studies have showed that circular RNAs (circRNAs) are important regulators for tumorigenesis by modulating malignant behaviors of tumor cells. However, the functions of EBV-encoded circRNAs in EBV-associated gastric carcinoma (EBVaGC) remain poorly understood. Methods: The expression of ebv-circRPMS1 in EBVaGC tissues, xenografts and cell lines were analyzed by BaseScope, qRT-PCR and in situ hybridization (ISH). The effects of ebv-circRPMS1 on gastric carcinoma (GC) cell proliferation, apoptosis, migration and invasion were measured by CCK8, EdU, immunouorescence (IF), FACS and Transwell assays. qRT-PCR, Western blotting, ChIP, RNA uorescence in situ hybridization (RNA-FISH), luciferase reporter assays, mass spectrum, RNA immunoprecipitation (RIP), and pulldown assays were used to investigate the molecular mechanisms. Xenograft mouse model was also used to analyze the effect of ebv-circRPMS1 on GC growth and metastasis in vivo. Results: We demonstrated that ebv-circRPMS1 promoted the proliferation, migration, invasion and anti-apoptosis of EBVaGC cells. Mechanistically, ebv-circRPMS1 recruited the Sam68 complex to the promoter of METTL3 and enhanced its transcription. Moreover, overexpression of METTL3 induced transcriptional activation of downstream genes (such as SNAI1, ZMYM1 and SOCS2) via m 6 A modications on their mRNAs, which were associated with tumor progression. Besides, RNA binding proteins (RBPs) such as QKI, DHX9 and ILF3, might involve in ebv-circRPMS1 biogenesis. In clinical EBVaGC samples, ebv-circRPMS1 was associated with distant metastasis and poor prognosis. Conclusion: These ndings indicated that ebv-circRPMS1 contributed to EBVaGC progression through recruiting the Sam68 ebv-circRPMS1 expression cell proliferation, and an ebv-circRPMS1-dependent manner. m 6 A modications latent and lytic of 6 A)-binding protein YTHDF1 EBV EBV a circRNA, EBV-mediated oncogenesis.

EBV-encoded small non-coding RNAs (EBERs, including EBER1 and EBER2) could be detected by in situ hybridization (ISH) in about 10% of GC worldwide, which are de ned as EBV-associated GC (EBVaGC) 2,3 .
In addition to EBERs, EBV expresses microRNAs (ebv-miR-BARTs), the BamHI A region long non-coding RNAs, EBV nuclear antigen 1 (EBNA1) and/or latent membrane 2A (LMP2A) in EBVaGC. Recently, it has been shown that EBV also produces circular RNAs (ebv-circRNAs) 4,5 . For example, ebv-circRPMS1, which is derived from the RPMS1 gene, was found to be expressed in B-cell lymphoma, NPC and EBVaGC 6,7 . A lower abundance of the circEBNA_U, which derived from the EBNA locus, was detected in B cells that displayed type I or type III latency or underwent reactivation 8 . While recent studies have revealed that EBV generates a diverse repertoire of viral circRNAs, the functions of these viral circRNAs remain largely unknown. Our group demonstrated that ebv-circLMP2A was enriched in cancer stem cells (CSCs) of EBVaGC and involved in inducing and maintaining stemness phenotypes through targeting the miR-3908/TRIM59/p53 axis 9 . Liu et al. showed that ebv-circRPMS1 promoted epithelial-mesenchymal transition (EMT) in NPC through sponging multiple miRNAs, such as miR-203, miR-31 and miR-451 7 .
N 6 -methylation of adenosine (m 6 A) is the most abundant modi cation ubiquitously occurred in eukaryotic mRNAs, which plays an important role in the regulation of mRNA stability, splicing, transport, localization and translation 10,11 . The m 6 A modi cation is dynamic and reversible in mammalian cells. It can be installed by m 6 A methyltransferases, such as METTL3, METTL14 and WTAP, and removed by m 6 A demethylases including FTO and ALKBH5. Additionally, speci c RNA-binding proteins (RBPs), such as YTHDF1/2/3, YTHDC1/2, eIF3, IGF2BP1/2/3 and HNRNPA2B1, can bind to the m 6 A motifs directly or indirectly to affect RNA function 12,13 . It has been reported that m 6 A modi cation is involved in various biological processes, including stem cell differentiation, tissue development, and tumor progression [14][15][16][17] .
To date, little is known about the roles that ebv-circRNAs play in the development of EBVaGC. In this study, we found that ebv-circRPMS1, an EBV-encoded circRNA derived from back-splicing of RPMS1 (from exon 4 to exon 2), was highly expressed in EBVaGC tissues and positively correlated with distant metastasis and poor outcome. In addition, ebv-circRPMS1 promoted EBVaGC cell proliferation, migration and invasion, and inhibited apoptosis in vitro and in vivo. Mechanistically, ebv-circRPMS1 recruited the Sam68 complex to the promoter of METTL3, resulting in METTL3 transcription enhancement and subsequent activation of downstream genes (such as SNAI1, ZMYM1 and SOCS2) via m 6 A modi cation on their mRNAs. These ndings indicated that ebv-circRPMS1 contributed to EBVaGC progression through Sam68 complex dependent activation of METTL3, a key enzyme of m 6 A modi cation. The relative expression of mRNAs was calculated using the 2 -ΔΔCt method. GAPDH or U6 was used as the endogenous control to normalize the data. All primer sequences are shown in Supplementary Table S2  and Table S3.

Materials And Methods
Actinomycin D and RNase R treatment Cells were planted into six-well plates, up to 60% con uency after 24 h, cells were treated with Actinomycin D (5 μg/ml, Millipore) or DMSO and collected at 0h, 4h, 8h, 12h and 24h. Total RNA (2 μg) was incubated with 3 U/μg of RNase R (Epicentre Technologies) for 20 min at 37 °C. After treatment with Actinomycin D or RNase R, the levels of ebv-circRPMS1 and RPMS1 mRNAs were analyzed by qRT-PCR.

Nuclear and cytoplasmic extraction
Cytoplasmic and nuclear fractions were isolated using the reagents supplied in PARIS™Kit (Thermo Fisher Scienti c). Brie y, SNU719 cells were lysed in Cell Fraction Buffer on ice for 10 min. After centrifugation at 500 × g for 3 min at 4 °C, the supernatant was collected as cytoplasmic fraction. Then, by washing the pellet with Cell Fraction Buffer, the nuclei fraction was collected. Flow cytometry analysis of cell apoptosis Cells were planted into six-well plates, up to 80% con uency, cells were treated with 5-Fluorouracil at the concentration of 2.5μM or 5μM. At 48 hours after culture, cells were stained with Annexin V-PE and 7-AAD (BD) according to the manufacturer's instructions. The cells were analyzed by using a BD FacScanto II ow cytometer and analyzed with FlowJo software.
RNA uorescence in situ hybridization (RNA-FISH) RNA-FISH was performed following the manufacturer's instructions. Firstly, the probes targeting the backsplicing site of ebv-circRPMS1 were designed by RiboBio (Guangzhou, China). Cells were xed with 4% paraformaldehyde for 10 min and then permeabilized in PBS with 0.5% Triton X-100 for 5 min. Next, the cells were hybridized with the labeled ebv-circRPMS1 probes at 37 °C overnight. Afterwards, the cells were washed with 4× sodium citrate buffer containing 0.1% Tween-20 for 5 min and then 1× SSC for 5 min.
Finally, cells were stained with DAPI for 10 min. The images were acquired using a laser confocal microscope (Leica).

Fluorescence immunocytochemical staining
Cells were seeded on glass coverslips in 24-well plates. Cells were xed with 4% paraformaldehyde at room temperate for 30 min and incubated with antibodies speci c for SAM68 or P53. Then, coverslips were treated with Alexa Fluor 594 or 488 goat anti-rabbit IgG, whereas nuclei were counterstained using DAPI. Images were acquired using a laser confocal microscope (Leica).

RNA tag-labeled assay
The RNA-protein complex was pulled down by incubating the cell lysates with RNA tag-labeled-coupled MS2 (GENESEED) according to a previous study 19 . The level of ebv-circRPMS1 in the complex was evaluated by qRT-PCR, whereas the proteins in the complex were detected by western blotting or mass spectrometry.

RNA-binding protein immunoprecipitation (RIP)
RIP assay was performed using a Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore) according to the manufacturer's instructions. Brie y, cells were lysed in RIP lysis buffer on ice for 30 min. After centrifugation, the supernatant was incubated with 30 μl of Protein-A/G magnetic beads (MedChemExpress) and antibodies. After incubation overnight, the immune complexes were centrifuged then washed six times with 1× washing buffer. The beads bound proteins were further analyzed using western blotting. The immunoprecipitated RNA was subjected to qRT-PCR analysis.

Co-immunoprecipitation (Co-IP)
Cells were incubated in lysis buffer containing protease inhibitor Cocktail (MedChemExpress) for subsequent Co-IP. The cell lysates were incubated with speci c antibodies targeting Flag, METTL3, SAM68, MS2 and P53 at 4 °C for 2 h, then incubated with 30 μl Protein-A/G magnetic beads at 4 °C overnight. The beads were separated and washed by using cold phosphate-buffered saline, and then subjected to western blotting analysis.

ChIP assays
Brie y, cells were crosslinked with formaldehyde for 15 min and terminated with 0.125 M glycine. Cells were then sonicated to generate chromatin with average size of 500 bp. Samples were rst pre-cleared with Protein-G magnetic beads (Cell Signaling) for 1 h at 4 °C. Cleared chromatin was then incubated with the antibodies overnight at 4 °C, and then with Protein-A/G magnetic beads for 6 h at 4 °C. Beads were then recovered, washed multiple times, and de-cross-linked by incubating them overnight at 65 °C in 1% SDS, 0.1 M NaHCO 3 . DNA was then puri ed by using the E.Z.N.A. Gel Extraction Kit DNA (Omega). The promoters of the speci c genes were then analyzed by qRT-PCR. Primer sequences used for qRT-PCR are included in Supplementary Table S2.

Luciferase reporter assay
The sequence of METTL3 promoter was cloned downstream of p-GLO Dual-Luciferase vector (VectorBuilder). 293T cells in 96 well plates were transfected with luciferase reporter and indicated expression constructs. All cells were harvested 48 h after transfection and analyzed using the dualluciferase reporter gene assay system (Promega, Madison, WI). The relative ratio of re y luciferase activity to Renilla luciferase activity was determined.
m 6 A RNA immunoprecipitation (MeRIP) assay m 6 A RNA immunoprecipitation (MeRIP) assay was conducted according to the previously described protocol with a slight modi cation 20 . Brie y, 100 ng of the total RNA was collected as input; the remaining RNA was used for m 6 A-immunoprecipitation with m 6 A antibody and Protein A/G to obtain m 6 A pull down portion. Finally, immunoprecipitated RNA was analyzed through qRT-PCR.

Total m 6 A RNA quanti cation
The EpiQuik m 6 A RNA Methylation Quanti cation Kit (Colorimetric) was used to measure the m 6 A level of total RNA in GC cells. First, 200ng RNA was added into the assay wells. Then, the detection antibody solution was added into assay wells. The m 6 A level was quanti ed by colorimetry, and the absorbance of each well was measured at 450 nm, and the calculation of m 6 A level according to the standard curve. m 6 A dot blot assay Firstly, a serial dilution of RNA was spotted onto a HybondN+ membrane (GE Healthcare). The membranes were then UV crosslinked, blocked, incubated with m 6 A antibody and horseradish peroxidase conjugate anti-rabbit immunoglobulin G. Finally, the signals from the dot blot were visualized by an ECL Western Blotting Detection Kit (Thermo Fisher Scienti c). Methylene blue staining served as a loading control.
Animal studies NOD/SCID mice were obtained from GemPharmatech Laboratory (Nanjing, China). For the xenograft model, 3 × 10 6 cells were injected into the subcutaneous of mouse (n = 5 per group). For the in vivo lung metastases model, 5 × 10 6 cells suspended in 100 μl PBS were injected intravenously into the tail vein of each 4-week-old NOD/SCID mice (n = 5 per group). After eight weeks, mice were anaesthetized with iso urane and images were acquired with the Xenogen IVIS Lumina series II for 5 min and analyzed using the Living Image 2.11 software package (Xenogen Corp). All the mice were sacri ced after eight weeks, and the xenografts and lungs were xed with phosphate-buffered formalin and sectioned for H&E staining and immunohistochemical analysis. All animal studies were performed in accordance with the institutional ethics guidelines for the animal experiments which were approved by Immunohistochemistry and in situ hybridization Immunohistochemistry (IHC) analysis was performed using a GT Vision III Kit (Leica) according to the manufacturer's instructions. The staining results were scored as follows: staining intensity score, 0 (no staining), 1 (weak), 2 (moderate), or 3 (strong); staining area score, 0 (≤10% positive staining), 1 (11-50% positive staining), 2 (51-75% positive staining), and 3 (≥75% positive staining). Staining intensity score and staining area score were multiplied to yield a nal score. Finally, the expression was divided into low and high expression group according to ROC curve. In situ hybridization (ISH) assay was performed with a commercially available EBV oligonucleotide probe complementary to EBER-1 (PanPath, Amsterdam, Netherlands), as previously described by Chen et al 21 .
BaseScope assay BaseScope assays were performed in accordance with the guidelines provided by the manufacturer (Advanced Cell Diagnostics). 4 μm thick sections were placed onto Superfrost plus slides (Fisher Scienti c) and baked at 60 °C for 1 h before depara nizing in xylene (2 × 5 min) and ethanol (2 × 2 min). After drying by baking at 60 °C for 2 min, pretreat 1 (hydrogen peroxide) was applied for 10 min at RT,

Statistical analysis
All experiments were carried out at least three times. Data are presented as mean ± standard deviation (S.D.). The statistical signi cance of differences was evaluated by Chi-square test, two-tailed Student's ttest or one-way ANOVA. OS was assessed with the Kaplan-Meier method and compared by the Log-rank test. The correlations among ebv-circRPMS1, METTL3 and SAM68 expression in xenografts and EBVaGC tissues were calculated by Pearson correlation analysis. Statistical signi cance was set at a value of P < 0.05. All statistical analyses were carried out using SPSS 22.0.

ebv-circRPMS1 promotes EBVaGC cell proliferation, migration, invasion, and inhibits apoptosis
Ebv-circRPMS1 is a circRNA that encoded by EBV 6,8, 22 . To further characterize ebv-circRPMS1, we designed divergent primers to detect the circRNA through RT-PCR and Sanger sequencing (Fig.S1B). The results revealed that ebv-circRPMS1 was derived from back splicing between exon 4 and exon 2 of RPMS1 (Fig. S1A). RNase R treatment, which degrades linear transcripts but not circular ones, showed that the linear transcript (RPMS1 mRNA) was dramatically reduced, while ebv-circRPMS1 remained untouched (Fig. S1C). In addition, the half-life of ebv-circRPMS1 exceeded 24 h, whereas that of RPMS1 mRNA was about 8 h (Fig. S1D), indicating that ebv-circRPMS1 is more stable than its cognate mRNA. Moreover, ebv-circRPMS1 localized both in the cytosol and the nucleus (Fig. S1E-F). Taken together, these data indicated that ebv-circRPMS1 is a bona de circRNA.
In order to investigate the potential roles of ebv-circRPMS1 in EBVaGC tumorigenesis, ebv-circRPMS1 expressing vector was constructed by cloning the circularizing sequence into the vector, which favored the formation of ebv-circRPMS1 (Fig. 1A). The nuclear localization of ebv-circRPMS1 was further analyzed by RNA uorescent in-situ hybridization assay in ebv-circRPMS1-transfected AGS cells. RNA FISH clearly demonstrated ebv-circRPMS1 exclusively in the nucleus (Fig. S2). Ectopic expression of ebv-circRPMS1 promoted cell proliferation, migration and invasion, and inhibited apoptosis in AGS and BGC823 cells (Fig. 1B-H, Fig. S3 & Fig. S4). In contrast, effectively knocked down the expression of endogenous ebv-circRPMS1 by small interfering RNAs (siRNAs) targeting the ebv-circRPMS1 backsplice junction (Fig. S5A), signi cantly inhibited the proliferation, migration and invasion of SNU719 and YCCEL1 cells, and induced more apoptosis (Fig. S5B-F). These results demonstrated that ebv-circRPMS1 might participate in EBVaGC tumorigenesis through promoting cell migration and invasion.

ebv-circRPMS1 regulates EBVaGC cell malignant behaviors through activating METTL3
Given that m 6 A methylation is potentially involved in tumor progression 17 , we examined the level of total m 6 A methylation in AGS and BGC823 cells with ebv-circRPMS1 overexpression. Compared with mock controls, the RNA m 6 A levels were signi cantly higher in ebv-circRPMS1 overexpression cells ( Fig. 2A-B).
Next, we examined the expressions of m 6 A related enzymes by qRT-PCR and western blot. METTL3 was the only enzyme that was up-regulated in GC cells with ebv-circRPMS1 overexpression (Fig. 2C-D, Fig.  S6).
Overexpression of METTL3 signi cantly upregulate the m 6 A levels on RNA in AGS and BGC823 cells (Fig.  3A-D). In addition, it enhanced the proliferation, migration and invasion of AGS and BGC823 cells, and inhibited apoptosis (Fig. 3E-I, Fig. S7 & Fig. S8). On the contrary, stable METTL3 knockdown in SNU719 and YCCEL1 cells dramatically reduced the m 6 A levels on RNA (Fig. S9 & Fig. S10A-D). Besides, knockdown of METTL3 obviously suppressed GC cell proliferation, migration and invasion and induces apoptosis in SNU719 and YCCEL1 cells (Fig.S10E-I).
Taken together, these ndings suggested that ebv-circRPMS1 regulates EBVaGC cell malignant behaviors through METTL3.
3. ebv-circRPMS1 directly interacts with Sam68 complex to activate METTL3 transcription The most acknowledged function of circRNAs is a microRNA sponge function, while we failed to identify miRNA interactions with ebv-circRPMS1 (Fig. S11), it suggests that the viral circRNA probably do not function as miRNA sponges. To further investigate the molecular mechanisms, RNA tag-labeled pulldown was performed using MS2-labeled protein (Fig. 4A, Fig. S12). Mass spectrometry revealed that 48 proteins consistently pulled down by exogenous ebv-circRPMS1 (Fig. 4B), and 8 of them were RBPs de ned by RBPDB. Further comprehensive analysis indicated Sam68 was the potential ebv-circRPMS1interacting partners (Fig. 4C-E, Fig. S13). The RNA recognition motif domain [174-202 amino acids (aa)] but not amino-or carboxyl-terminus of Sam68 protein was necessary for its interaction with ebv-circRPMS1 (Fig. 4F). Mutation of RNA recognition motif domain, potential interacting regions analyzed by catRAPID, abolished the interaction of Sam68 with ebv-circRPMS1 (Fig. 4G). In addition, Sam68 was signi cantly up-regulated with ebv-circRPMS1 overexpression (Fig. 4H). These results suggested that ebv-circRPMS1 directly interacted with Sam68 protein in EBV positive cells.
Furthermore, the present study investigated how ebv-circRPMS1 targets and inhibits the expression of METTL3 gene. By treating with cycloheximide (CHX) and MG132, it was revealed that ebv-circRPMS1 did not affect the protein stability of METTL3 (Fig. S14), indicating that ebv-circRPMS1 may affect the protein level of METTL3 at post-transcriptional level. Sam68 is a transcriptional regulator that activates downstream gene expression 23 . Therefore, we wondered whether Sam68 participates in the regulation of METTL3 transcription. ChIP assay showed that Sam68 enriched on METTL3 promoter region (-800 -600 bp from the transcription start site) (Fig. 5A, Fig. S15). Notably, ebv-circRPMS1 silencing impaired the enrichment of Sam68 on METTL3 promoter (Fig. 5A). Additionally, ebv-circRPMS1 silencing suppressed the enrichment of transcriptional active marker H3K27ac on METTL3 promoter (Fig. 5B). Moreover, P53, a cofactor reported to interact with Sam68, which can synergistically activate the transcription and regulate gene expression 23 , enriched on METTL3 promoter region (Fig. 5C-D). Also, ebv-circRPMS1 colocalized on METTL3 promoter with Sam68 (Fig. 5E). Overexpression of ebv-circRPMS1, Sam68 and P53, respectively, promoted the luciferase activity of METTL3 promoter (Fig. 5F), demonstrating that ebv-circRPMS1, Sam68 and P53 may form a complex to promote METTL3 transcription.
In ebv-circRPMS1 overexpression AGS cells, silencing of Sam68 decreased the expression of METTL3 and the RNA m 6 A levels, and reversed the effects of ebv-circRPMS1 overexpression on cell proliferation, apoptosis, migration and invasion. In ebv-circRPMS1-depleted YCCEL1 cells, upregulation of Sam68 signi cantly increased the expression of METTL3 and the RNA m 6 A levels, and reversed the effects of ebv-circRPMS1 silencing on cell proliferation, apoptosis, migration and invasion (Fig. 6, Fig. S16 & Fig.   S17). These data suggested that ebv-circRPMS1 directly interacted with Sam68 complex to activate METTL3 transcription, thus promoting tumor progression.

SNAI1, ZMYM1 and SOCS2 are potential genes involved in EBVaGC progression
Since METTL3 is an m 6 A writer, we sought to nd out its potential targets involved in m 6 A-regulated tumor progression. According to the previous studies 15,24,25 , eight genes (SNAI1, ZMYM1, SOCS2, HDGF, MALAT1, MET, ADAT2, TGFB1) were selected as candidate genes. The m 6 A levels of Snail (encoded by SNAI1), ZMYM1 and SOCS2 mRNAs were signi cantly increased in ebv-circRPMS1 expression cells than in negative controls (Fig. S18). In addition, their mRNA levels were also higher in ebv-circRPMS1 overexpression cells than those in ebv-circRPMS1 knockdown cells (Fig. S18).

Up-regulated ebv-circRPMS1 in EBVaGC tissues positively correlated with poor prognosis
In EBVaGC, the expression of ebv-circRPMS1 was positively correlated with the expression of METTL3 ( Fig. 8A-B) and distant metastasis (p=0.001, Table 1). Besides, the expressions of METTL3 and Sam68 were also positively correlated (Fig. 8B). In human gastric carcinoma samples, RNA-FISH showed that ebv-circRPMS1 and Sam68 co-localize in the nucleus (Fig. S22). The expression of METTL3 was signi cantly correlated with lymph node metastasis (p=0.023) and distant metastasis (p=0.006), while the expression of Sam68 was signi cantly correlated with Lauren classi cation (p=0.030), lymph node metastasis (p=0.013) and distant metastasis (p=0.017) ( Table 1). Furthermore, higher expression of ebv-circRPMS1 in EBVaGC patients predicted shorter overall survival (Fig. 8C). The patients with METTL3 up-regulation also had a worse overall survival (Fig. 8C). These results indicated that ebv-circRPMS1 exerts an oncogenic role in EBVaGC.
7. Identi cation of potential splicing regulators in ebv-circRPMS1 biogenesis QKI, DHX9 and ILF3 (NF90/NF110) are RBPs that have been identi ed as splicing-associated factors involved in circRNA biogenesis 26-28 . Knockdown of these three RBPs in YCCEL1 cell line showed a reduction of ebv-circRPMS1; however, the expression of RPMS1 mRNA was not affected (Fig. 9A).

Discussion
In this study, we characterized the function and molecular mechanisms of ebv-circRPMS1 in EBVaGC. We found that ebv-circRPMS1 promoted EBVaGC cell proliferation, migration and invasion and inhibited apoptosis both in vitro and in vivo. Mechanistically, ebv-circRPMS1 recruited the Sam68 complex to METTL3 promoter, which subsequently activated the transcription of METTL3 and regulated downstream genes that associated with tumor progression, such as SNAIL, ZMYM1 and SOCS2, via m 6 A modi cations on their mRNAs. In addition, we demonstrated that QKI, DHX9 and ILF3 might involve in ebv-circRPMS1 biogenesis. In clinical EBVaGC samples, ebv-circRPMS1 was highly expressed in EBVaGC tissues and predicted poor prognosis. These ndings indicated that ebv-circRPMS1 contributed to EBVaGC progression through recruiting the Sam68 complex to activate METTL3 expression and its downstream targets.
Recent studies have found that EBV, a human DNA tumor virus, produced circRNAs 6,8, 9,22 . ebv-circRPMS1 was found to be expressed in a panel of EBV-associated diseases representing latency types I, II, and III, including EBV-positive PTLD, BL, EBVaGC, NPC, and AIDS-associated lymphoma 6,8, 22 . In addition, ebv-circRPMS1 was also expressed during reactivation 29 . ebv-circEBNA_U, which derived from the EBNA locus, was detected in B cells that displayed type I or type III latency or underwent reactivation 29 . ebv-circBHLF1 was detected in most latency cell models but displayed extraordinarily high expression under reactivation conditions 29 . In contrast, ebv-circLMP2 was seldom expressed in latency, but expressed speci cally during reactivation 22 .
Despite extensive identi cation of EBV circRNAs, the function of EBV circRNAs has not been extensively explored. Our group has demonstrated that ebv-circLMP2A was enriched in cancer stem cells (CSCs) of EBVaGC and could induce and maintain stemness through targeting the miR-3908/TRIM59/p53 axis 9 .
Here, we showed that ebv-circRPMS1 was stably expressed in EBVaGC and played an essential role in EBVaGC progression. Ectopic expression of ebv-circRPMS1 enhanced GC cell proliferation, migration and invasion and inhibited apoptosis. Similarly, Liu et al. found that ebv-circRPMS1 promoted tumor metastasis in NPC, probably through sponging multiple miRNAs, such as miR-203, miR-31 and miR-451 7 .
In EBVaGC, we also tried to see whether ebv-circRPMS1 could also act as a miRNA sponge. Although some miRNA binding sites of ebv-circRPMS1 were predicted by bioinformative software, we were unable to identify interactions between ebv-circRPMS1 and these cellular miRNAs (Fig. S9).
Considering the nding that ebv-circRPMS1 was mainly located in the nucleus of EBVaGC cells, we wondered whether ebv-circRPMS1 exerted its function through RNA-protein interaction, because most nuclear circRNAs regulate gene expression at post-transcriptional or transcriptional levels through RNAprotein interaction. For example, cis-acting circ-CTNNB1 acted as a mediator of β-catenin signaling through DDX3-mediated transactivation of YY1 and promoted cancer progression 30 . Circ-DONSON was localized in the nucleus, recruited the NURF complex to SOX4 promoter and initiated its transcription in gastric cancer cells 31 . CircSamd4 bound with PUR proteins and thereby prevented their interaction with the Mhc promoter, which repressed myogenic transcriptional activity 32 . In the present study, we found that ebv-circRPMS1 can bind to RNA-binding protein Sam68, and regulate METTL3 transcription.
Sam68 is a transcription factor involved in a variety of cellular processes, including alternative splicing, cell cycle regulation, RNA 3'-end formation, tumorigenesis, and regulation of human immunode ciency virus (HIV) gene expression 33,34 . Sam68 usually acts as a potent transcriptional co-factor in tumor progression via interacting with transcription regulatory proteins, such as P53 23 . As a transcriptional regulator, Sam68 has been reported to regulate the promoter activity of NF-κB/P65 35 . In this study, we demonstrate that Sam68 functions as a transcription co-factor with P53 to facilitate the transactivation of METTL3. In addition, we found that RNA recognition motif domain of Sam68 was necessary for its interaction with ebv-circRPMS1. Mutation of RNA recognition motif domain abolished the interaction of Sam68 with ebv-circRPMS1.
METTL3, a methyltransferase of mRNA m 6 A, exhibits functions for the self-renewal of cancer stem cells, promotion of cancer cell proliferation, and resistance to radiotherapy or chemotherapy 36,37 . We found that METTL3 expression was dramatically up-regulated in ebv-circRPMS1 expression cells and promoted GC cell proliferation, migration and invasion in an ebv-circRPMS1-dependent manner. Studies have found that m 6 A modi cations were associated with latent and lytic infection of virus 38 . Xia et al. showed that N 6 -methyladenosine (m 6 A)-binding protein YTHDF1 suppresses EBV replication and promotes EBV RNA decay 39 . Blocking m 6 A inhibited splicing of the pre-mRNA encoding replication transcription activator (RTA) and halted lytic replication of KSHV 40 . Depletion of the m 6 A machinery had different anti-viral impacts on gene expression of KSHV depending on the cell-type analyzed 41 . In this study, we showed for the rst time that the host RNA m 6 A modi cation machinery can be targeted and manipulated speci cally by a virus-encoded circRNA, ebv-circRPMS1. It suggested a novel link between the host epitranscriptome and EBV-mediated oncogenesis.
So far, there are three well-studied mechanisms about circRNA biogenesis 42 . The rst one is that RBPs binding to intronic regions and facilitating their pairing through homo-or hetero-dimerization promote circRNA biogenesis. The second one is that the presence of anking inverted Alu repeat elements that stabilize proximity of the splice donor and acceptor sequences ease circRNA formation. The last one, the orientation-opposite complementary sequences residing in the anking introns of the circularized exon(s) facilitated the formation of circRNA. In the present study, knockdown of QKI, ILF3, and DHX9 repressed circRNA biogenesis. QKI is a dimer, promotes circRNA biogenesis by bringing the circle-forming exons into close proximity. It is essential for enhanced production of many circRNAs and acts by binding to recognition elements within introns, in the vicinity of the circRNA-forming splice sites 27 . ILF3 is a dsRNA binding protein, which strongly bind to and stabilize transient RNA pairs formed between intronic complementary sequences, leading to the enhanced circRNA biogenesis in the nucleus 26 . DHX9 is a helicase which stabilizes internal repeat Alu pairing and promotes circRNA biogenesis 28 . Our data suggested that RNA binding/processing factors such as these RBPs play an important role in facilitating EBV backsplicing, however, the exact mechanisms still need to be further investigated.
In conclusion, we found that ebv-circRPMS1 promoted EBVaGC cell proliferation, migration and invasion and inhibited apoptosis. Availability of data and materials All data are included in the manuscript.

Con icts of interest
The authors declare no potential con icts of interest   used to track the transfection. B and C, RT-PCR and qRT-PCR assay revealing the relative levels of ebv-circRPMS1 (normalized to GAPDH) in GC cell lines. D and E, CCK8 and EdU assays were used to analyze the proliferation of AGS and BGC823 cells after ebv-circRPMS1 overexpression. Scale bar, 200 µm. F and G, Transwell assay illustrated that ebv-circRPMS1 overexpression facilitated the migration and invasion of AGS and BGC823 cells. Scale bars, 20 µm. H, Annexin/PI staining followed by FACS analysis indicated that ebv-circRPMS1 overexpression inhibited apoptosis. *p < 0.05, **p < 0.01, ***p < 0.001 METTL3 promotes GC cell proliferation, migration and invasion, and inhibits apoptosis in vitro. A and B qRT-PCR and western blotting assay revealing the levels of METTL3 overexpression. C, RNA m6A levels were detected by colorimetric ELISA-like assay using the RNA m6A methylation quanti cation kit. D, RNAs isolated from mock or METTL3-overexpressing GC cells were used in dot blot analyses with an anti-m6A antibody. Methylene blue staining served as the loading control. E and F, CCK8 and EdU assays were used to analyze the proliferation of AGS and BGC823 cells after METTL3 overexpression. Scale bar, 200 µm. G and H, Transwell assay illustrated that METTL3 overexpression facilitated the migration and invasion of AGS and BGC823 cells. Scale bars, 20 µm. I, Annexin/PI staining followed by FACS analysis indicated that METTL3 overexpression inhibited apoptosis. *p < 0.05, **p < 0.01, ***p < 0.001 Figure 4 ebv-circRPMS1 directly interacts with Sam68 complex. A, Schematic diagram showing the process to generate RNA tag-labeled ebv-circRPMS1 by ligation of in vitro transcribed linear transcripts. B, RNA taglabeled ebv-circRPMS1 was incubated with SNU719 cell lysates, followed by mass spectrum identi cation. 48 proteins were pulled down by exogenous ebv-circRPMS1, including 28 up-regulated proteins and 20 down-regulated ones. C, Heat map showing the differentially expressed peptide. Venn diagram indicating the Sam68 was identi ed as a potential candidate for ebv-circRPMS1-interacting partner. D, RIP assay using anti-Sam68 showed that Sam68 precipitated ebv-circRPMS1 in SNU719 cell lysates. E, Pulldown assay con rmed that RNA tag-labeled ebv-circRPMS1 interacted with Flag-Sam68. F, Domain mapping assay indicated that the region of 174-202 aa in Sam68 was essential for the interaction with ebv-circRPMS1. G, RIP assay con rmed the interaction of Sam68 (174-202 aa) with ebv-circRPMS1. H, qRT-PCR and western blotting assay revealing the up-regulation of Sam68 after ebv-circRPMS1 overexpression. Figure 5 ebv-circRPMS1 recruits the Sam68 complex to activate METTL3 transcription. A, ChIP assay showed that Sam68 was enriched on METTL3 promoter, and ebv-circRPMS1 silencing attenuated the enrichment of Sam68 on METTL3 promoter. B, ChIP assay showed that depletion of ebv-circRPMS1 impaired the enrichment of active markers H3K27ac on METTL3 promoter. C, Co-IP assay showed that Sam68 interacts with P53. D, ChIP assay showed that P53 was enriched on METTL3 promoter. E, RNA-FISH assay veri ed the colocalization of ebv-circRPMS1, Sam68 and P53 in YCCEL1 cells. Scale bar, 20 μm. F, Luciferase reporter assay showed that overexpression of ebv-circRPMS1, Sam68, or P53 increased the luciferase activity. *p < 0.05, **p < 0.01

Figure 6
The ebv-circRPMS1/Sam68-METTL3 axis modulates GC progression. A and B, qRT-PCR and western blotting analysis of Sam68 expression in knockdown of ebv-circRPMS1 or overexpression ebv-circRPMS1 systems. C, RNA m6A levels were detected by colorimetric ELISA-like assay using the RNA m6A methylation quanti cation kit. D, RNAs isolated from GC cells were used in dot blot analyses with an anti-m6A antibody. Methylene blue staining served as the loading control. E and F, CCK8 and EdU assays were Effects of ebv-circRPMS1 on GC growth in vivo. A, Representative images of tumor tissues in each group were presented in the left. Tumor volumes were determined every 5 days. After sacri ced, the tumor weights were measured (n = 5 for each group). B, ISH analysis of EBER, RNAscope analysis of ebv-circRPMS1 and IHC analysis of METTL3, Ki67 and Sam68 expression in tumor tissues of each group.
Scale bar, 20 μm. C, qRT-PCR analysis of ebv-circRPMS1 and METTL3 in tumor tissues. D, western blotting analysis of METTL3 in tumor tissues. E, left and middle, representative hematoxylin and eosin staining images; right, the number of metastatic nodules formed in the lungs of mice for each group (n = 5 for each group). Scale bar, 500 μm. *p < 0.05, **p < 0.01, ***p < 0.001 Figure 8 ebv-circRPMS1 is upregulated in EBVaGC tissues and positively correlated with poor prognosis. A, ISH analysis of EBER, RNAscope analysis of ebv-circRPMS1 and IHC analysis of METTL3 and Sam68 expression in EBVaGC tissues. Scale bars, 100 μm. B, ebv-circRPMS1 was positively correlated with the mechanisms underlying ebv-circRPMS1 promoted GC progression. ebv-circRPMS1 binds Sam68 to facilitate its interaction with p53, resulting in transactivation of METTL3 and transcriptional alteration of downstream target genes, which associated with metastasis and cancer progression. *p < 0.05, **p < 0.01

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.