Circ_SMAD4 promotes gastric carcinogenesis by activating wnt/β‐catenin pathway

Abstract Objectives Circular RNAs (circRNAs) are essential participants in tumour progression. This study focused on investigating the mechanism of a novel functional circRNA in gastric cancer (GC). Methods Gene expression was detected by qRT‐PCR or Western blot. Survival curves were generated via Kaplan‐Meier method. In vitro and in vivo assays were used to investigate the impact of circ_SMAD4 on GC cell growth and tumorigenesis. Agarose gel electrophoresis assay, RNase R treatment and Sanger sequencing were utilized for confirming the circular structure of circ_SMAD4. Relationship between molecules was monitored by a series of mechanical experiments, as needed. Results Circ_SMAD4 expression was potentiated in GC. Circ_SMAD4 depletion impeded GC cell growth in vitro and restrained tumorigenesis in vivo. Mechanically, nuclear circ_SMAD4 recruited TCF4 to facilitate CTNNB1 transcription, while cytoplasmic circ_SMAD4 sequestered miR‐1276 to prevent the silence of CTNNB1 mRNA, leading to activation of Wnt/β‐catenin pathway. Rescue experiments validated that circ_SMAD4 depended on miR‐1276/TCF4‐regulated CTNNB1 to elicit accelerating effects on GC cell growth. Conclusion Circ_SMAD4 facilitated GC tumorigenesis by activating CTNNB1‐dependent Wnt/β‐catenin pathway. Hopefully, the findings could provide new clues for improving GC prognosis and treatment.

via sponging miR-195, resulting in GC aggravation. 13,14 Another common mechanism behind circRNA regulating gene expression involves RNA binding proteins (RBPs). 15,16 As an example, nuclear circRNA_102171 interacts with CTNNBIP1 to activate WNT/βcatenin pathway in papillary thyroid cancer. 17 However, preliminary documents still lack thorough explanation of most circRNAs in GC.
As a classic pathway, Wnt/β-catenin signalling is frequently activated in the tumorigenesis of diverse cancers, including GC. 18 For instance, LINC01133 inactivates Wnt/β-catenin pathway to impair GC progression via miR-106a-3p/APC axis. 19 As is well-known, the hallmark of Wnt/β-catenin activation is the nuclear accumulation of β-catenin. 20 However, whether circRNAs mediate modulation on this pathway in GC still requires deeper understanding.
With the employment of microarray analysis, circ_SMAD4 was unveiled to be overexpressed in GC. Therefore, present study aimed at unravelling the function and mechanism of circ_SMAD4 in GC tumorigenesis. The present study might provide the potential for circ_SMAD4 to be a technically effective biomarker for GC therapies.

| Microarray
CircRNA microarray analysis was undertaken with Human CircRNA Array v2.1 (CapitalBio, Beijing, China). The differentially expressed circRNAs were deemed to be significantly different between groups in line with following conditions: fold change > 2.0 and P < 0.05.

| Actinomycin D and RNase R treatments
For RNase R treatment, total RNA was cultivated without or with RNase R (Epicentre, Madison, WI, USA) for 30 minutes at 37℃. The relative levels of circ_SMAD4, linear SMAD4 and GAPDH were assayed by qRT-PCR, normalizing to those measured in mock group.
Ct values from triplicate samples were normalized to the internal housekeeping transcript and were then normalized to that of samples before adding Actinomycin D. Results were shown as the ratio of mRNA abundance at indicated times relative to 0 hour.

| Colony formation assay
Colony formation assay was implemented to assess the proliferation of indicated AGS or BGC-823 cells. Cells (800 per well) were added in 6-well plates and cultured for 14 days. Afterwards, colonies were fixed in methanol (Sigma-Aldrich), stained by crystal violet (Sigma-Aldrich) and calculated manually.
Tumour volume was recorded every 4 days. Mice were killed and dissected at 4 weeks following injection, and the tumours were weighed. Animal assay was conducted with the approval of the

Animal Ethics Committee of Affiliated Hospital of Guilin Medical
College.

| Immunohistochemistry (IHC)
Tissues gained from xenograft model were put in formalin (Sigma-Aldrich), fixed and paraffin-embedded. Following deparaffinating and rehydrating, sections were incubated with primary antibodies against Ki67 (Abcam, Cambridge, USA) or PCNA (Abcam) and corresponding secondary antibodies. Following staining using DAB (Sigma-Aldrich), sections were observed under a light microscopy (Nikon).

| Subcellular fractionation
According to the previous protocol, 22 RNA in the cytoplasm and nucleus of GC cells was isolated with a PARIS™ Kit (Ambion, Austin, TX, USA). Isolated RNA was assayed utilizing qRT-PCR, with GAPDH or U6 as the cytoplasmic or nuclear control, respectively.

| Fluorescence in situ hybridization (FISH)
FISH assay was implemented in line with the procedures described previously. 23 Alexa Fluor 555-labeled circ_SMAD4 probe was synthesized by RiboBio. A Fluorescent in Situ Hybridization Kit (RiboBio) was attained to study the probe signals. Images were examined applying the fluorescence microscopy.

| RNA Pull-down assay
RNA pull-down assay was implemented with the Pierce™ Magnetic RNA-protein pull-down kit (Thermo Fisher Scientific). The pull-down probe of circ_SMAD4 was the oligonucleotide containing complementary binding site for the spliced junction region of circ_SMAD4.
Biotin-labelled RNA was mixed with cell extracts, with the mixture containing non-biotin RNA as the NC. The mixtures were incubated with Dynabeads Myone Streptavidin T1 beads. Then, proteins in the RNAprotein complex were eluted and separated by SDS-PAGE (Bio-Rad, Hercules, CA, USA). Afterwards, the gel was stained using a Fast Silver Stain Kit (Beyotime, Shanghai, China), followed by mass spectrometry.

| Chromatin immunoprecipitation (ChIP)
As described previously, ChIP experiments were executed. 24 Crosslinked chromatin was initially sonicated to fragments of 500-bp.
Immunoprecipitation was conducted adopting antibodies against TCF4 or IgG (negative control). Following adding magnetic beads, the precipitated chromatin fragments were purified, isolated and examined via qRT-PCR.

| Electrophoretic mobility shift assay (EMSA)
EMSA experiment was conducted as depicted previously. 25 The nuclear extracts were obtained from AGS cells. Probes were produced by annealing single-strand oligonucleotides containing the TCF4 consensus sequence of circ_SMAD4 promoter and labelling the ends with [γ-32P] ATP using T4 polynucleotide kinase (TaKaRa Bio). Anti-TCF4 (Proteintech) and anti-IgG (Santa Cruz) were used as primary antibodies.

| Statistical analysis
Results shown as mean ± SD were derived from at least three independent assays and imported into SPSS 22.0 (IBM, Chicago, IL, USA).
The Kaplan-Meier and log-rank rest method was applied for survival analysis. Correlations between RNAs were analysed via Pearson's correlation analysis. Differences in experimental variables were confirmed via Student's t test or one-way ANOVA, with the significant level of P < 0.05.

| Characterization of circ_SMAD4 level and its stable structure
To search dysregulated circRNAs in GC, microarray analysis was carried out in 3 GC tissues and paired non-tumour samples.
Consequently, 86 circRNAs exhibited elevated levels and 114 exhibited suppressed levels in GC tissues relative to controls. We focused on the most overexpressed circRNA and hsa_circ_0047718 ( Figure 1A). Hsa_circ_0047718 was derived from SMAD4 (called circ_SMAD4 subsequently) and had 6786 nt in length ( Figure 1B), and its splice junction forming the circular structure was determined by Sanger sequencing (Supporting Information Figure S1A). Agarose gel electrophoresis (AGE) assay displayed that circ_SMAD4 was amplified only from cDNA with divergent primers but could not be found in the products from genomic DNA ( Figure 1C and Supporting Information Figure S1B-C). In addition, it was proved that circ_SMAD4 was more stable than linear SMAD4 ( Figure 1D-E). Thereafter, qRT-PCR data depicted a higher circ_SMAD4 expression in 40 GC tissues in contrast to matched non-cancerous samples ( Figure 1F), so was the level of linear SMAD4 (Supporting Information Figure S1D). More importantly, high circ_SMAD4 level predicted dreadful survival in GC patients ( Figure 1G). Consistently, circ_SMAD4 level was also increased in GC cells compared with normal GES-1 cells ( Figure 1H). To summarize, circ_SMAD4 was upregulated in GC.

| Loss of circ_SMAD4 blunted GC tumorigenesis
In vitro and in vivo assays were executed to probe the role of circ_SMAD4 in GC. First, we confirmed that circ_SMAD4 not SMAD4 was interfered in AGS and BGC823 cells by specific shR-NAs targeting the splice junction of circ_SMAD4 (Figure 2A and Supporting Information Figure S2A). Interestingly, silencing circ_

| Circ_SMAD4 activated WNT/βcatenin pathway
CircRNAs have been increasingly documented to modulate functional pathways to affect tumour growth. 13 Figure S4A-C). Next, to probe whether circ_SMAD4 could regulate WNT/βcatenin pathway in GC, we investigated the potential regulation of circ_SMAD4 on CTNNB1. Unsurprisingly, CTNNB1 expression was F I G U R E 2 Loss of circ_SMAD4 blunted GC tumorigenesis. A, Circ_SMAD4 or SMAD4 expression in AGS and BGC823 cells was detected by qRT-PCR and AGE after transfection with sh/circ_SMAD4#1/2. B, Images and quantification of colonies formed under circ_SMAD4 interference. C, Images and quantification of EdU-stained AGS and BGC823 cells under circ_SMAD4 deficiency. D, The influence of circ_SMAD4 depletion on AGS and BGC823 cell apoptosis was assessed via flow cytometry analysis. (E-F) The growth curve and weight of subcutaneous xenograft tumours obtained from nude mice under sh/ctrl and sh/circ#1 sets. G, IHC staining evaluated Ki-67 and PCNA expressions in tumours with or without circ_SMAD4 silencing. *P < 0.05, **P < 0.01. n.s. indicated no significance F I G U R E 3 Circ_SMAD4 affected the activation of WNT/β-catenin pathway. A, Circ_SMAD4 expression was investigated upon the application of several pathway inhibitors in AGS and BGC823 cells. B, CTNNB1 mRNA and protein (β-catenin), and WNT/β-catenin pathwayrelated proteins (CCND1 and c-myc) were determined after circ_SMAD4 blockade by using qRT-PCR or Western blot. (C-D) Colony formation and EdU assays proved that LiCl treatment rescued the hindered AGS and BGC823 cell proliferation caused by circ_SMAD4 inhibition. E, Flow cytometry analysed that LiCl abrogated the effects of circ_SMAD4 blockade on AGS and BGC823 cell apoptosis. **P < 0.01

F I G U R E 4 Circ_SMAD4 bound to TCF-4. (A-B)
The distribution of circ_SMAD4 in AGS and BGC823 cells was assayed through subcellular fractionation plus qRT-PCR and immunofluorescence staining. C, Nuclear and cytoplasmic β-catenin level in two GC cells was assessed by Western blot. D, Under the absence of circ_SMAD4, CTNNB1 promoter activity was assayed by means of luciferase reporter assay. E, RNA pull-down and mass spectrometry interrogated circ_SMAD4-binding proteins. F, RIP and AGE confirmed that circ_SMAD4 bound to TCF4 protein. SNRNP70 and U1 as positive controls. G, TCF4 was effectively knocked down as illustrated via qRT-PCR and Western blot. (H-I) The impacts of TCF4 downregulation on regulating CTNNB1, c-myc, CCND1 and β-catenin levels were appropriately validated through qRT-PCR and Western blot. **P < 0.01 significantly higher in GC tissues (Supporting Information Figure   S5A). Importantly, CTNNB1 level was positively correlated with circ_SMAD4 in these 40 GC samples (Supporting Information Figure   S5B). Moreover, circ_SMAD4 deficiency impaired CTNNB1 level and also decreased the levels of β-catenin, CCND1 and c-myc proteins ( Figure 3B and Supporting Information S5C). Then, we conducted rescue experiments via WNT/β-catenin agonist, LiCl. Results demonstrated that the effects of circ_SMAD4 interference on GC cell proliferation and apoptosis were counteracted after LiCl treatment ( Figure 3C-E). Altogether, circ_SMAD4 activated WNT/β-catenin pathway to expedite GC cell growth.

| circ_SMAD4 bound to TCF-4 in GC cells
Since subcellular location of circRNAs can be informative to their regulatory mechanism, we subsequently examined circ_SMAD4 distribution in GC cells. It was uncovered that circ_SMAD4 existed in both the nuclear and cytoplasmic fractions of GC cells, although with a larger proportion in nucleus ( Figure 4A-B). It was depicted that inhibition of circ_SMAD4 reduced the level of both cytoplasmic and nuclear β-catenin in AGS and BGC823 cells ( Figure 4C).
Such data implied that circ_SMAD4 might regulate total β-catenin expression but not directly affect the transport of β-catenin, which echoed the previous observation that CTNNB1 level was restrained by circ_SMAD4 knock-down. Therefore, we then probed into how circ_SMAD4 influenced CTNNB1 expression in GC. Results suggested that circ_SMAD4 depletion evidently reduced the luciferase activity of CTNNB1 promoter (Figure 4D), indicating a possible transcriptional regulation of circ_SMAD4 on CTNNB1. Hence, the circ_SMAD4-binding proteins were surveyed via RNA pull-down and mass spectrometry. It turned out that transcription factor 4 (TCF4) was abundant in Bio-circ_SMAD4 group ( Figure 4E). Next, the interaction of circ_SMAD4 with TCF4 was further certified by RIP assays, RNA pull-down and EMSA assays ( Figure 4F and Supporting Information Figure S5D-E). Furthermore, we certified TCF4 upregulation and the positive expression correlation between circ_SMAD4 and TCF4 in GC tissues (Supporting Information Figure   S5F-G). Thereafter, we downregulated TCF4 and observed declined CTNNB1 level and lowered protein levels of β-catenin, CCND1 and c-myc ( Figure 4G-I and Supporting Information S5H). Taken together, circ_SMAD4 interacted with TCF4 to affect CTNNB1 expression and WNT/β-catenin activation.

| Circ_SMAD4 recruited TCF4 to potentiate CTNNB1 transcription
Subsequently, we intended to confirm the impact of circ_SMAD4-TCF4 interaction on CTNNB1 expression. By utilizing UCSC (http:// genome.ucsc.edu/) and JASPAR (http://jaspar.gener eg.net/), we obtained TCF4 binding motif and also found two potential TCF4 binding sites in CTNNB1 promoter ( Figure 5A-B). Further, luciferase reporter assay data revealed that upregulating TCF4 enhanced the luciferase activity of wild-type CTNNB1 promoter, while such enhancement was partly mitigated when mutating site 1 or site 2 alone but completely offset when both sites were mutated ( Figure 5C).
These data corroborated that both sites were responsible for the interaction between TCF4 and CTNNB1 promoter. Subsequently, ChIP assay confirmed the binding between TCF4 and CTNNB1 promoter ( Figure 5D). Interestingly, such binding was hindered upon circ_SMAD4 depression, but totally recovered under further augmentation of TCF4 ( Figure 5E). However, the suppression of silenced circ_SMAD4 on CTNNB1 expression and the levels of β-catenin, CCND1 and c-myc were only partially counteracted after overexpressing TCF4 ( Figure 5F and Supporting Information Figure S6A).
Functionally, loss of TCF4 retarded GC cell growth, while such influence was then antagonized byCTNNB1 overexpression or LiCl treatment (Supporting Information Figure S6B-D). All data substantiated that circ_SMAD4 potentiated CTNNB1 transcription via the recruitment of TCF4.

| circ_SMAD4 upregulated CTNNB1 as a miR-1276 sponge
Since circ_SMAD4 modulated CTNNB1 expression not only depending on its nuclear function, we wondered whether cytoplasmic circ_SMAD4 also had some contribution to elevate CTNNB1 in GC cells. Further, considering the high potential for cytoplasmic circR-NAs as a ceRNA, we focused on miRNAs that could bind to both circ_SMAD4 and CTNNB1. Through analysing starBase and miRDB, we discovered 10 miRNAs shared by circ_SMAD4 and CTNNB1 ( Figure 6A), and three (miR-1276, miR-4429 and miR-320b) were further selected since their overexpression decreased the luciferase activity of circ_SMAD4 ( Figure 6B). Nevertheless, only miR-1276 mimics could reduce the luciferase activity of CTNNB1 in both cells ( Figure 6C). In addition, miR-1276 expressed at a low level in GC tissues and cells (Supporting Information Figure S7A-B), and its level was negatively related to circ_SMAD4 in GC samples (Supporting Information Figure S7C). Based on these data, we guessed miR-1276 mediated the regulation of circ_SMAD4 on CTNNB1. Corresponding miR-1276 binding sites in circ_SMAD4 and CTNNB1 were exhibited in Figure 6D. Luciferase reporter assay results displayed that only the luciferase activities of CTNNB1-WT and circ_SMAD4-WT declined upon miR-1276 mimics ( Figure 6E-F). Additionally, miR-1276 failed to affect the activity of SMAD4 3′UTR reporters (Supporting Information Figure S7D). Moreover, co-existence of circ_SMAD4, miR-1276 and CTNNB1 was found exclusively in the anti-Ago2 group

| Circ_SMAD4 aggravated GC cell growth by augmenting CTNNB1 via miR-1276 and TCF4
Previously, we disclosed that circ_SMAD4 boosted CTTNB1 expression in GC via both TCF4-and miR-1276-mediated manners.
Here, we detected whether it functioned in GC also through these two pathways. It turned out that TCF4 upregulation or miR-1276 inhibition alone resulted in partial recovery of circ_SMAD4 deficiency-hampered cell growth, whereas the combined effect of them completely rescued the impeded cell growth induced by circ_SMAD4 interference (Supporting Information Figure S8A-C). At length, we investigated whether circ_SMAD4 relied on CTNNB1 to function in GC cell growth. Before that, we validated that CTNNB1 was effectively upregulated by pcDNA3.1/CTNNB1 ( Figure 7A). Consequently, CTNNB1 overexpression offsets the impacts of circ_SMAD4 depletion on cell proliferation and apoptosis ( Figure 7B-D). In sum, circ_SMAD4 drove GC tumorigenesis by upregulating CTNNB1 via miR-1276 and TCF4.

F I G U R E 5
Circ_SMAD4 recruited TCF4 to potentiate CTNNB1 transcription. A, TCF4 motif obtained from JASPAR. B, JASPAR predicted two binding sites in CTNNB1 promoter for TCF4. C, Luciferase reporter assay determined the effect of TCF4 overexpression on the reporters containing different CTNNB1 promoter sequences. D, ChIP assay tested the interaction between TCF4 and CTNNB1 promoter. E, ChIP analysed the impact of circ_SMAD4/TCF4 on the binding of TCF4 to CTNNB1 promoter. F, qRT-PCR and Western blot depicted the impact of TCF4 overexpression on circ_SMAD4 silence-inhibited CTNNB1 mRNA and c-myc, CCND1 and β-catenin levels. *P < 0.05, **P < 0.01 F I G U R E 6 Circ_SMAD4 upregulated CTNNB1 by serving as a miR-1276 sponge. A, With the predication of miRDB and starBase, potential miRNAs for circ_SMAD4 and CTNNB1 were unveiled. B, Luciferase reporter assay tested the luciferase activity of circ_SMAD4 under the respective overexpression of 10 miRNA mimics. C, Luciferase reporter assay examined the luciferase activity of CTNNB1 in the presence of 3 indicated miRNA mimics. D, MiR-1276 binding sites in circ_SMAD4 and CTNNB1 were delineated. (E-F) Luciferase reporter assay evaluated the luciferase activity of CTNNB1-WT/MUT or circ_SMAD4-WT/MUT upon miR-1276 overexpression. G, Ago2-RIP assay substantiated the existence of circ_SMAD4, miR-1276 and CTNNB1 in Ago2 precipitates. H, qRT-PCR investigated the influence of circ_ SMAD4 inhibition on miR-1276 level in AGS and BGC823 cells. I, qRT-PCR measurement of CTNNB1 expression under miR-1276 mimics. *P < 0.05, **P < 0.01

| D ISCUSS I ON
In this study, circ_SMAD4 was selected as the most upregulated circRNA in GC through microarray assays. Further, we proved the overexpressed circ_SMAD4 in GC tissues and cells. With regard to its clinical features, we demonstrated that circ_SMAD4 contributed to shorter survival duration. Next, in vitro and in vivo assays functionally displayed that circ_SMAD4 exerted oncogenic effect on tumorigenesis in GC. These findings linked circ_SMAD4 to GC pathogenesis.
Mechanically, we unveiled WNT/β-catenin pathway as the downstream of circ_SMAD4 in GC. Previously, the modulation of cir-cRNAs on this pathway has been substantiated by various reports. 28 The specific role of circRNAs varies according to their cellular distribution. 29 Presently, we discovered circ_SMAD4 located in both nucleus and cytoplasm of GC cells. Subsequently, the nuclear circ_ SMAD4 was uncovered to influence CTNNB1 transcription. Inspired by the previous hypothesis that circRNAs could interact with specific nuclear proteins to transcriptional regulate gene expressions. 22 We searched the binding proteins of circ_SMAD4 and discovered TCF4.
TCF4 can bind to β-catenin to trigger WNT/β-catenin pathway. 30,31 Interestingly, here we proved that nuclear circ_SMAD4 could interact with TCF4 and then recruit TCF4 to CTNNB1 promoter to potentiate CTNNB1 transcription, resulting in WNT/β-catenin signalling transmission. Similar function of other circRNAs has also been described by a number of reports. For instance, circ-DONSON initiated SOX4 transcription through enhancing the recruitment of NURF complex to SOX4 promoter. 22 Also, circAnks1a recruited YBX1 to Vegfb promoter to activate Vegfb transcription. 23 Additionally, rescue assay results verified that TCF4 contributed to GC cell growth by CTNNB1-activated WNT/β-catenin pathway. However, we uncovered that TCF4-induced elevation of CTNNB1 could not offset the suppression of silenced circ_SMAD4 on CTNNB1, indicating circ_SMAD4 regulated CTNNB1 through another pathway.
It is proposed that cytoplasmic circRNAs, mRNAs and lnRNAs share and compete for the binding of miRNAs to establish a competing endogenous network which post-transcriptionally modulates protein-coding genes. 32,33 For example, cytoplasmic circAKT3 promoted PIK3R1 by sponging miR-198 in GC cells. 34 Current study manifested a post-transcriptional modulation of circ_SMAD4 on CTNNB1 as a sponge of miR-1276. Further, we proved that circ_ SMAD4 affected WNT/β-catenin pathway by miR-1276/CTNNB1 axis. Moreover, miR-1276 inhibition and TCF4 upregulation completely offset the impacts of interfered circ_SMAD4 on CTNNB1 expression and GC cell growth.
In conclusion, our work first illuminated the pro-growth role of circ_SMAD4 in GC through activating CTNNB1-dependent WNT/β-catenin signalling via interacting with miR-1276 and TCF4.
Consequently, targeting circ_SMAD4 might be a promising method for treating GC. However, how WNT/β-catenin pathway affects circ_SMAD4 expression in turn still remains covered, which is the biggest limitation of our present work.

ACK N OWLED G EM ENTS
Thank you to everyone who participated in this study.

F I G U R E 7
Augmentation of CTNNB1 restored GC cell growth inhibition induced by circ_SMAD4 suppression. A, qRT-PCR confirmation of CTNNB1 overexpression in AGS and BGC823 cells. (B-C) Colony formation and EdU assays tested cell proliferation under diverse conditions. D, Flow cytometry analysed the apoptosis of indicated cells. **P < 0.01