Circ_0053943 is upregulated in UM tissues
In the present study, we focused on upregulated circRNAs based on the published circRNA microarray data between UM and normal uveal tissues (Fig. S1A) . Therefore, the qRT-PCR was performed to detect these upregulated circRNAs in 5 UM and normal tissues for further validation. Hsa_circ_0053943 was prominently overexpressed in UM tissues compared to normal group (Fig. 1A), similar to hsa_circ_0119872 reported previously (Fig. S1B) [10, 11]. Then, we detected the expression level of circ_0053943 in UM cell lines; as shown in Fig. 1B, its expression in ARPE-19 cells was significantly lower than that in UM cell lines. Thus, the uncharacterized hsa_circ_0053943 was chosen for our subsequent in-depth study.
Hsa_circ_0053942 (cric_0053943) is generated from exons 9, 10, and 11 of the RasGRP3 gene and is 567 nucleotides in length according to the circBase annotation (Fig. 1C) . To validate the back-splice junction site of circ_0053943, we performed a series of experiments, including PCR amplification, agarose gel electrophoresis, and Sanger sequencing. Subsequently, convergent and divergent primers were designed to amplify the linear and back-splicing products based on cDNA and genomic DNA (gDNA) from UM cells. The subsequent agarose gel electrophoresis demonstrated that the convergent primers could amplify products of expected size from cDNA and gDNA, both for circ_0053943 and GAPDH. However, divergent primers for circ_0053943 could only amplify a PCR product from cDNA but not from gDNA. Meanwhile, no product was observed after being amplified by divergent primers for GAPDH in cDNA or gDNA (Fig. 1D). We next conducted an RNase R treatment assay to confirm the circular characteristics of circ_0053943 further. The results demonstrated that circ_0053943 was more resistant to RNase R and actinomycin D treatment than linear RasGRP3 mRNA (Fig. 1E and Fig. S1C). In addition, subcellular fractionation assays, including RNA-FISH and RNA nucleus/cytoplasm separation, revealed that circ_0053943 was predominantly localized in the cytoplasm (Fig. 1F, G). These data suggested that circ_0053943 was a circular structure and upregulated in a large portion of UM, which might serve as a prognostic biomarker.
Circ_0053943 promoted the proliferation and metastasis of UM cells
Three shRNAs targeting circ_0053943 were transfected into MUM2B and C918 cells to elucidate the biological role of circ_053943 in UM proliferation and metastasis, according to the comparatively high expression in these two cell lines. Meanwhile, OCM-1 and OCM-1A cells were stably developed with a circ_0053943 overexpression lentivirus. At the same time, the expression of RasGRP mRNA remained unchangeable in either circ_0053943 knockdown or stable overexpressing cells. Further, shRNA#1 and #2, and a scrambled non-target shRNA control, were chosen for the subsequent cell phenotype assays, considering their greater knockdown efficiency (Fig.S1D, E).
The growth curves performed by CCK8 assays were suppressed significantly by down-regulated circ_0053943 (Fig. 2A), whereas upregulated circ_0053943 prominently promoted OCM-1 and OCM-1A cells (Fig. 2B). Similarly, the EDU staining assay confirmed that the number of EDU-positive MUM2B and C918 cells (proliferative cells), following transfected with shRNAs, was decreased compared with the control group (Fig. 2C). Conversely, the OCM-1 and OCM-1A cells transfected with overexpressed lentivirus revealed a significant increase compared with the control group (Fig. 2D). Transwell assays and scratch wound healing assays revealed that circ_0053943 knockdown led to a significant decrease in cell migration and invasion ability in MUM2B and C918 cells. In contrast, circ_0053943 overexpression increased the ability of OCM-1 and OCM-1A cells (Fig. 2E-I).
The flow cytometric assays of cell cycle distribution indicated that downregulating circ_0053943 increased the percentage of G0/G1 phase and decreased S phase populations in MUM2B and C918 cells, compared with the control group (Fig. 3A). Conversely, the overexpression of circ_0053943 induced the progression of G1-to-S phase transformation and suppressed apoptosis remarkably in OCM-1 and OCM-1A cells (Fig. 3B). Apoptosis assay showed that cells transfected with shRNAs had higher apoptotic rates than the control group, while transfection of overexpressed lentivirus decreased the apoptotic cells (Fig. 3C, D). In addition, Cyclin D1/CDK4 was reported to induce G1/S transition, while Bcl-2 (anti-apoptotic protein) /Bax (pro-apoptotic protein) was described to regulate cell apoptosis antagonistically. Thus, western blot was used to determine critical molecules involved in the cell cycle and apoptosis. Results demonstrated that when knocking down circ_0053943, the protein expression levels of Cyclin D1, CDK4, Bcl-2, and Bax were reduced or increased in a manner consistent with the results of the flow cytometric assays while overexpressing circ_0053943 had the opposite effects (Fig. 3E).
The abovementioned findings indicated that circ_0053943 might behave as an oncogene by promoting proliferation, metastasis, and invasion in UM cells.
Circ_0053943 and IGF2BP3 cooperate to play oncogenic roles
Given that circRNAs can influence the bio functions via sponging miRNAs, a RIP assay was conducted to test if circ_0053943 regulates targets as a miRNA sponge in UM. Interestingly, the results showed that the AGO2 antibody significantly enriched ciRS-7 (a circRNA binding with AGO2), but not circ_0053943 (Fig. S2A), suggesting circ_0053943 may not act as a miRNA sponge in UM progression.
Then we conducted an RNA pulldown assay coupled with mass spectrometry (MS) to explore whether circ_0053943 exerted function via interacting with proteins (Fig. S2B). The MS assay revealed 167 differential proteins between the sense and antisense circ_0053943 transcript pulldown groups in MUM2B cells (Table S6). Afterward, overlapped differential proteins with RBPs (Table S7) and 50 proteins were found to be potential partners. Since circ_0053943 was mainly localized in the cytoplasm, cytoplasmic protein could interact with circ_0053943. Biotin-labeled RNA pulldown assay and separated by SDS-PAGE were performed with cytoplasmic protein in MUM2B cells. After silver staining, the sense-specific band at about 55-70 kDa was excised (Fig. 4A). Among the 50 differential proteins, 3 proteins might be potential protein partners of circ_0053943. Notably, only IGF2BP3 could be detected in the input group and circ_0053943 pulldown products but not in the antisense pulldown products (Fig. 4B and Fig. S2C). Furthermore, Kaplan-Meier survival analysis in TCGA cohorts showed that the patients with higher IGF2BP3 expression were associated with shorter overall survival (OS) and disease-free survival (DFS) rates (Fig. 4C).
As shown in Fig. 4D and Fig. S2D, the anti-IGF2BP3 antibody could specifically enrich circ_0053943 compared with the anti-IgG antibody following the RIP assay results. In addition, the Dual RNA-FISH and immunofluorescence assay showed the co-localization of circ_0053943 and IGF2BP3 in MUM2B and C918 cells, thereby further supporting their interaction (Fig. 4E). Then, to determine which domain is interacted with circ_0053943, we designed 6 truncated IGF2BP3 plasmids aiming at 6 functional domains. Protein domain mapping and RIP assay showed that the KH1 and KH2 domains were essential for the interaction between IGF2BP3 and circ_0053943 (Fig. 4F-H). Overexpression or knockdown of circ_0053943 could not affect the mRNA and protein levels of IGF2BP3 (Fig. 4I and Fig. S2D, E), while no significant changes in the expression of circ_0053943 were also observed after IGF2BP3 was knocked down (Fig. 4J and Fig. S2F). Of note, IGF2BP3 depletion could abolish the induction of cell proliferation, migration, and invasion elicited by the circ_0053943 overexpression. In contrast, IGF2BP3 overexpression could not affect the proliferation and metastatic ability of UM cells while circ_0053943 was simultaneously knocked down (Fig. S3A-F). Collectively, these results suggest that circ_0053943 and IGF2BP3 may cooperate to play oncogenic roles in UM cells.
EGFR was identified as a downstream target of circ_0053943 and IGF2BP3
Given that circ_0053943 specifically interacts with IGF2BP3 but does not affect the protein levels, we next investigated how circ_0053943 promoted UM progression through IGF2BP3. Previous studies reported that IGF2BP3, which stabilized an extensive repertoire of target mRNA transcripts, was dysregulated and played an oncogenic role in diverse types of cancers [21-22]. To confirm this hypothesis, RNA-sequence was used to profile the global effects of circ knockdown compared with the negative control in MUM2B cells. Gene expression profiling after circ_0053943 knockdown showed a total of 4,293 genes with expression changes, including 2,308 upregulated genes and 1,985 downregulated genes (Fig. 5A). Furthermore, pathway enrichment analysis based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) database showed the top 12 signaling pathways where the MAPK pathway was most enriched (Fig. 5B). Considering the MAPK signaling pathway was identiﬁed by KEGG pathway enrichment analysis, the critical components of the MAPK pathway, p38, ERK, JNK, and their corresponding phosphorylation forms were evaluated using western blot. As displayed in Fig. 5C, the levels of P38, ERK1/2, and JNK did not show significant changes after silencing the circ_0053943. Meanwhile, the phosphorylated forms of these proteins were reduced, and the reduction in phosphorylated ERK showed the most significance. These findings indicated that circ_0053943 might promote UM proliferation and metastasis by regulating the MAPK/ERK signaling pathway.
To this end, we performed an integrated analysis among our RNA-sequence data, two published IGF2BP3 RIP-sequence data , and critical components in the MAPK signaling pathway to define the target genes potentially regulated by circ_0053943 and IGF2BP3. Among 1,985 downregulated genes via depletion of circ_0053943, 1,159 genes had IGF2BP3’s enrichment on their transcripts through two independent IGF2BP3 RIP-sequence data, and 25 genes act as possible key targets in the MAPK signaling pathway (Fig. 5D). Strikingly, survival analysis was performed based on the expression status of these 25 genes by GEPIA 2 , and 10 genes showed prognostic impacts in UM (Fig. 5E and Fig. S4A, B). We then selected these 10 genes as possible targets for further validation in circ_0053943 and IGF2BP3 knockdown cells. Similar to the RNA-seq results, the transcript levels of ARRB2, EGFR, FGFR4, PDGFA, and PDGFB were decreased via depletion of circ_0053943, while only EGFR was showed a significant decrease of more than 50% both in circ_0053943 depletion and IGF2BP3 depletion cells (Fig. 5F).
Thus, EGFR was chosen as the IGF2BP3-bound target altered by circ_0053943 for further research. Western blot analysis confirmed that knockdown of either circ_0053943 or IGF2BP3 reduced the protein levels of EGFR (Fig. S5A). In addition, q-PCR and western blot showed that knocking down circ_0053943 abrogated the long-lasting effect of IGF2BP3 overexpression on the EGFR transcript. Conversely, reducing IGF2BP3 abolished the half-life and mRNA level of increased by circ_0053943 overexpression (Fig. 5G, H and Fig. S5B, C). Collectively, these data suggest a functional interdependency between circ_0053943 and IGF2BP3 when stabilizing the EGFR transcript.
Circ_0053943 cooperates with IGF2BP3 to stabilize EGFR mRNA in an m6A-dependent manner
Considering that IGF2BPs (including IGF2BP1/2/3) was characterized as a new m6A reader that regulates mRNA of m6A-modified genes (such as MYC), we hypothesize that circ_0053943 and IGF2BP3 might regulate the stability of EGFR mRNA in an m6A-dependent manner. To confirm this hypothesis, we first identified the potential m6A-modified regions of EGFR based on the m6A RIP-sequence data. Through analyzing the m6A-RIP sequencing data, we detected several significant m6A peaks that were markedly reduced by knockdown of m6A “writers” silencing (i.e., methyltransferase-like 3 and 14 (METTL3 and METTL14)), in 5’- and 3'-untranslated regions (3’UTR and 5’UTR) of EGFR mRNA (Fig. 6A). Moreover, enrichments of the canonical “GGAC” m6A motif in these m6A peaks confirmed our hypothesis that EGFR mRNA is modified by m6A methylation.
Next, we performed m6A-specific RIP, followed qRT-PCR assay to determine the m6A methylation level of EGFR in METTL3 and METTL14 knockdown cells (Fig. S6A). The result showed that the m6A methylation of the 3’UTR, but not the 5’UTR of the EGFR transcript, was substantially decreased. In line with a previous study, the coding region instability determinant (CRD) of MYC (positive control) was also reduced. However, the negative control of HPRT1 did not show a similar reduction, which confirmed that EGFR was modified by m6A (Fig. 6B). Due to IGF2BP3 being early proven to preferentially bind to the “GGAC” m6A core motif of its targets , we further checked whether circ_0053943 affects IGF2BP3 binding the m6A-modified EGFR. The IGF2BP3 RIP qRT-PCR assays showed that knockdown of circ_0053943 significantly reduced IGF2BP3 binding to the m6A site in the 3’UTR region of EGFR, and the overexpression of circ_0053943 facilitated this binding, but not in the EGFR 5’UTR or MYC CRD (Fig. 6C). These results further indicated that circ_0053943 specifically enhances the occupancy of IGF2BP3 at 3’UTR of EGFR. Moreover, using RNA pulldown assay with in vitro-transcribed circ_0053943, we found that the binding abilities of circ_0053943 to the 3’UTR of EGFR were nearly abolished following IGF2BP3 knockdown and increased after IGF2BP3 overexpressed (Fig. S6B). Thus, it confirmed that circ_0053943 might bind the 3’UTR of EGFR in an IGF2BP3-dependent manner.
In the 3’UTR region of EGFR, we preliminarily found 2 “GGAC” m6A motifs predicted by m6A modification site predictors SRAMPA (http://www.cuilab.cn/sramp) and RMBase V2.0 (http://rna.sysu.edu.cn/rmbase/index.php) (Fig. 6D). Subsequently, we constructed luciferase reporters that contain the wild-type EGFR 3′UTR (WT-3’UTR), mutant #1 and #2 3′UTR (Mut#1- and Mut#2-3’UTR; changing GGAC to GGCC), to elucidate the potential roles of circ_0053943 in m6A modification of EGFR (Fig. 6E). Compared to the Mut#2-3’UTR reporter, both IGF2BP3 and m6A RIP qRT-PCR assays showed higher enrichment with the WT and Mut#1-3’UTR reporters. Additionally, overexpression of circ_0054943 or IGF2BP3 remarkably increased the binding of IGF2BP3 and m6A in the WT and Mut#1-3’UTR reporters, but not in the Mut#2-3’UTR report. Conversely, knocking down circ_0053943 and IGF2BP3 independently reduced this binding except in the Mut#2-3’UTR report (Fig. 6F, G). Furthermore, luciferase reporter assays demonstrated that the relative luciferase activities were decreased by knockdown circ_0053943 or IGF2BP3 in WT and Mut#1-3’UTR reporters, but not in Mut#2-3’UTR reporter. Similarly, the relative luciferase activity of MUT#2-3’UTR reporter was also not altered by overexpressed circ_0054943 or IGF2BP3 (Fig. 6H, I). Thus, methylation of m6A-modified site c.4510A in the 3’UTR of EGFR mRNA contributed to the EGFR expression.
Collectively, these data revealed that circ_0053943 cooperates with IGF2BP3 to increase EGFR mRNA stability in an m6A-dependent manner and further indicated that circ_0053943 might strengthen the m6A recognition by IGF2BP3 and recruitment of RNA stabilizers to m6A-methylated EGFR mRNA.
Circ_0053943 promoted UM proliferation, metastasis, and angiogenesis through upregulating EGFR in vitro
Given that circ_0053943 could promote the proliferation and metastasis of UM and the expression of EGFR, we hypothesized that circ_0053943 might promote UM progression by inducing EGFR expression. Hence, we developed an EGFR-specific shRNA(sh-EGFR) with negative control (sh-Ctrl) to suppress the expression of EGFR and an overexpression plasmid to upregulate the expression in UM cells. Transfection efficiency was examined at the mRNA levels and displayed in Fig. S7A.
As shown in cell proliferation assays, EGFR silencing could remarkably inhibit the proliferation ability of circ_0053943 overexpression cells. In contrast, EGFR overexpression could promote UM cell proliferation ability of circ_0053943 knockdown cells (Fig. 7A, B and Fig. S7B, C). Transwell assays indicated that metastasis abilities in sh-EGFR cells were significantly increased compared to those in sh-Ctrl cells when circ_0053943 was silenced (Fig. 7C and Fig. S7D). Similar results were also observed in the wound healing assay (Fig. 7D and Fig. S7E).
In addition to cell division, increased motility, and decreased apoptosis, EGFR expression was described to play an essential role in tumor angiogenesis. Given these findings, we analyzed the potential implication of the circ_0053943/IGF2BP3/EGFR mRNA-protein ternary complex on angiogenesis using HUVEC tube formation assay. As depicted in Fig. S7F, conditioned medium (CM) from sh-circ_0053943 MUM2B cells reduced the formation of capillary-like structures by HUVECs dramatically compared to control conditioned medium. In contrast, EGFR overexpression reversed the capillary-like structure formation decreased by circ_0053943 depletion. In contrast, CM from circ_0053943 OCM-1A cells showed the promoting effects on the formation of capillary-like structures, which were inhibited by silencing EGFR as shown in Fig. S7G. Together, these results demonstrated that overexpression of circ_0053943 increased capillary-like structure formation in HUVECs by regulating EGFR expression.
To this end, the expression of cycle-related proteins (Cyclin D1, CDK4) and apoptosis-related proteins (Bcl-2, Bax) were measured by performing western blotting. The results showed that the knockdown of circ_0053943 impaired the G1 to S translation by downregulating CDK4 and Cyclin D1, and the overexpression of EGFR could rescue this effect. The overexpression of circ_0053943 could promote the G1 to S transition, while silencing EGFR could impair this translation. On the other hand, the alterations of Bcl2 and Bax demonstrated that the anti-apoptotic effect of circ_0053943 could be weakened by the knockdown of EGFR (Fig. 7E, F).
Taken together, all the above demonstrated that increased circ_0053943 might upregulate the expression of EGFR, which further promotes UM proliferation and metastasis.
Circ_0053943 promoted UM proliferation and metastasis through upregulating EGFR in vivo
To investigate the effects of circ_0053943 on UM proliferation and metastasis in vivo, we next established subcutaneous xenograft nude mice models. MUM2B and OCM-1A cells transfected with sh-circ_0053943/sh-EFGR/overexpression plasmids, along with relative control group and co-transfection group, were separately injected into nude mice. As shown, tumor growth was repressed by circ_0053943 knockdown, with lesser tumor volume and weight than control groups, while overexpression of EGFR could reverse the growth inhibition caused by circ_0053943 depletion (Fig. 8A, B). In OCM-1A cells, circ_0053943 overexpression had an opposite effect, which was repressed by EGFR depletion (Fig. 8C, D). Additionally, the results of IHC showed that Ki-67 (the marker for tumor proliferation) and EGFR expression levels were decreased in the circ_0053943 depletion group and increased in the overexpression group (Fig. 8E).
Next, we investigated the effect of circ_0053943 on tumor metastasis in vivo. The anatomical results confirmed that the metastatic nodules were lessened in the circ_0053943 knockdown group compared with the control group. In contrast, the metastatic lesions at the liver surface were more abundant in the EGFR overexpression group. Further analysis showed that the reduction of metastatic nodules caused by circ_0053943 depletions was increased by EGFR overexpression. In contrast, the opposite effect was obtained in nude mice injected with circ_0053943/sh-EGFR cells (Fig. 8F-H).
Taken together, the results of the in vivo experiments strongly suggested that circ_0053943 upregulation enhanced the proliferation and metastatic capacity of UM by upregulating EGFR in vivo.