3.1. CircSOD2 is highly expressed in HCC tumor tissues and liver cancer cell lines
Genome-wide RNA-seq studies on HCC tumor tissues and their adjacent nontumorous liver tissues revealed that hsa_circ_0004662, derived from SOD2 gene was significantly upregulated in HCC tumor tissues [26]. However, its role in HCC is still unknown. To validate RNA-seq results, qRT-PCR with divergent primers complementary to circSOD2 exon 1 and 4 (Figure 1A) was used to detect circSOD2 expression from patients and cell lines. In line with RNA-seq data, circSOD2 expression was significantly upregulated in 18/19 HCC patient’s tumor tissues and liver cancer cell lines compared with normal liver tissues and normal liver cells (Figure 1B-C). High circSOD2 expression was also associated with poor survival (Figure 1F) and linked to higher grade tumors (Supplemental table 1). To further characterize circSOD2, RNA was extracted from HEPG2 and HUH7 cells and treated with or without RNase A before reverse transcription. The effect of RNase A on circSOD2 expression was then examined. Similar to other circRNA, circSOD2 was highly resistant to RNase A digestion. However, the RNA level of GAPDH was greatly impaired after RNase A treatment (Figure 1D). Data from circular RNA interactome showed that 13 circSOD2 binding sites exist in an RNA binding protein-Ago2 (data not shown), indicating circSOD2 may interact with Ago2 within the cell. To confirm this, RIP assay with Ago2 antibody or IgG control was performed. Indeed, Ago2 significantly precipitated circSOD2 from HEPG2 and HUH7 cells compared to IgG control (Figure 1E).
3.2. CircSOD2 promoter is intensively modified by H3K27ac and H3K4me3
H3K27ac and H3K4me3 modification indicate active gene transcription [27, 28]. To understand if these modifications contribute to circSOD2 upregulation in HCC. H3K27ac and H3K4me3 ChIP-seq data from HEPG2 cells were then examined. As shown in WashU browser, SOD2 promoter was extensively occupied by H3K27ac and H3K4me3 (Figure 2A), correlating with high circSOD2 expression. ChIP-qPCR with H3K27ac and H3K4me3 antibodies confirmed higher H3K27ac and H3K4me3 modification on circSOD2 promoter from HCC tumor tissues and liver cancer cell lines compared to normal liver tissues (Figure 2B-C) and normal liver cells (Supplementary figure 1A-B). EP300 and WDR5 mediate H3K27ac and H3K4me3 modification respectively. We next ask if EP300 and WDR5 bind to circSOD2 promoter. ChIP-qPCR was then performed. Similar to H3K27ac and H3K4me3 signal, EP300 and WDR5 were also significantly enriched in circSOD2 promoter from HCC tumor tissues compared to normal liver tissues (Figure 2D-E). In addition, partial depletion of EP300 or WDR5 (Figure 2F,I) greatly impaired H3K27ac or H3K4me3 signal on circSOD2 promoter (Figure 2G,J), circSOD2 expression was also downregulated (Figure 2H,K). Thus, these results indicate that the enrichment of EP300 and WDR5 on circSOD2 promoter increased its H3K27ac and H3K4me3 modification and circSOD2 expression in HCC.
3.3. CircSOD2 promotes in vitro liver cancer cell proliferation and tumorigenesis in vivo
To further characterize the role of circSOD2 in HCC, siRNA targeting the back spliced site of circSOD2 was used to silence circSOD2 expression in liver cancer cells. SiRNA efficiently silenced circSOD2 expression ~6 fold lower than its original level (Figure 3A). CircSOD2 downregulation impaired HEPG2 and HUH7 cell growth, and cell migration (Figure 3B-E). Moreover, cell cycle was also arrested in G0/G1 phase and cell apoptosis was increased following circSOD2 depletion (Figure 3F-H). The role of circSOD2 in in vivo tumorigenesis was also examined. In accordance with impaired in vitro cell proliferation, silencing circSOD2 also decreased HCC tumor formation in nude mice compared with scramble control (Figure 3I-J). Taken together, these results suggest that high circSOD2 expression may associate with HCC development.
3.4. CircSOD2 suppresses miR-502-5p expression by acting as a sponge
More and more evidences suggest that, by acting as a sponge, circular RNAs regulate miRNA expression [29, 30]. By searching the circular RNA interactome database, we found that miR-502-5p is a target of circSOD2. The potential interaction site between circSOD2 exon 3 and 5’ miR-502-5p was shown (Figure 4A). Our previous results showed that Ago2 interacts with circSOD2 in cells. In addition to that, here, we found Ago2 co-precipitated circSOD2 and miR-502-5p from HEPG2 and HUH7 cells (Figure 4B). To further determine if circSOD2 could directly interact with miR-502-5p. Wild type miR-502-5p or circSOD2-miR-502-5p interaction sites mutated miR-502-5p was labeled with biotin and used to pull down circSOD2 from cell lysates (Figure 4C-D). As indicated, wild type miR-502-5p efficiently precipitate circSOD2. However, the interaction between mutant miR-502-5p and circSOD2 was almost completely abolished (Figure 4E). Thus, these results confirmed the direct interaction between circSOD2 and miR-502-5p within liver cancer cells.
To further understand if circSOD2 plays a role in regulating miR-502-5p expression. The expression of miR-502-5p in HCC tumor tissues and liver cancer cells was first investigated. In contrast to circSOD2, miR-502-5p expression was downregulated in HCC patient tumor tissues and liver cancer cell lines compared with normal liver tissues and normal liver cells (Figure 4F-G). In addition, the expression level of miR-502-5p was negatively correlated with circSOD2 in HCC tumor tissues (Figure 4H). Furthermore, siRNA mediated circSOD2 knockdown greatly upregulated miR-502-5p expression (Figure 4I). Above all, these results clearly demonstrate that, by acting as a sponge, circSOD2 suppressed miR-502-5p expression in liver cancer cells.
3.5. Overexpression of miR-502-5p downregulates DNMT3a expression
miRNAs interact with the 3′ untranslated region (3′ UTR) of target mRNAs to induce mRNA degradation and translational repression [31]. Database from miRSystem, and TargetScan suggest that DNMT3a, a DNA methyltransferase is a target of miR-502-5p. The interaction site between DNMT3a 3’UTR and miR-502-5p was shown (Figure 5A). To test if miR-502-5p regulates DNMT3a expression in liver cancer cells, miR-502-5p mimic was introduced into HEPG2 and HUH7 cells (Figure 5B). Indeed, introducing miR-502-5p mimic into HEPG2 and HUH7 cells significantly downregulated DNMT3a transcription (Figure 5C), DNMT3a translation was also repressed (Figure 5D). To confirm the direct interaction between DNMT3a 3’UTR and miR-502-5p. Wild type or mutant DNMT3a 3’UTR was labeled with biotin and used to pull down miR-502-5p from cell lysates. Wild type DNMT3a 3’UTR significantly enriched miR-502-5p compared to mutant DNMT3a 3’UTR (Figure 5E-F). The higher DNMT3a expression in liver cancer cells compared to normal liver cells (Figure 5G) suggests that downregulation of miR-502-5p in HCC may otherwise facilitate DNMT3a expression.
3.6. DNMT3a activates JAK2/STAT3 signaling pathway by suppressing SOCS3 expression
Dysregulation of IL-6/STAT3 signaling pathway has been implicated in the pathogenesis of HCC [32-34]. To gain insights into the molecular mechanism underlying the role of DNMT3a in HCC, CRISPR-CAS9 was used to delete DNMT3a from HEPG2 and HUH7 cells, the effect of DNMT3a depletion on the expression of genes involved in JAK/STAT3 signaling pathway was examined. DNMT3a was efficiently deleted by CAS9 in HEPG2 and HUH7 cells. Interestingly, JAK2 inhibitor SOCS3 was greatly upregulated, while phosphorylated JAK2, and phosphorylated STAT3 were all downregulated (Figure 6A), indicating DNMT3a may be involved in SOCS3/pJAK2/pSTAT3 signaling pathway regulation in HCC. Since SOCS3 locates in the upstream of JAK/STAT signaling pathway. We next asked if DNMT3a could directly regulate SOCS3. Indeed, co-transfecting DNMT3a and SOCS3 promoter driven luciferase vector suppressed luciferase activity in a dose dependent manner (Figure 6B). However, no effect was observed in SHP1 or SOCS1 promoter activity in the presence of DNMT3a (Supplemental figure 1C-D). DNMT3a is a DNA methyltransferase that modifies CpG methylation and suppresses gene expression. By analyzing SOCS3 promoter, we found that there is an CpG island in the SOCS3 promoter (Figure 6C), suggesting that DNMT3a may downregulate SOCS3 expression by modifying its promoter methylation status. To test this hypothesis, DNMT3a ChIP-qPCR and DNA methylation pull down assay were performed. DNMT3a was highly enriched in SOCS3 promoter compared to its nearby non-CpG region (Figure 6D), 5mc antibody also significantly precipitated SOCS3 promoter compared to control (Figure 6E). Moreover, knocking down DNMT3a decreased SOCS3 promoter methylation in both HEPG2 and HUH7 cells (Figure 6F). In summary, these results suggest that, in liver cancer cells, DNMT3a upregulation promotes SOCS3 promoter methylation and suppresses SOCS3 expression, which further activates JAK2/STAT3 signaling pathway.
3.7. DNMT3a rescues liver cancer cell proliferation impaired by circSOD2 depletion
To further confirm that DNMT3a is the downstream gene among circSOD2 regulated signaling pathway. CircSOD2 was depleted in the presence or absence of exogenously expressed DNMT3a. In the absence of DNMT3a, depletion of circSOD2 downregulated DNMT3a and upregulated SOCS3 expression. However, SOCS3 expression was restored when DNMT3a was expressed (Figure 7A), suggesting that the effect of circSOD2 depletion on upregulating SOCS3 expression was mediated through suppressing DNMT3a. In line with these results, we showed that the expression of circSOD2 was negatively associated with SOCS3 while positively associated with DNMT3a expression in HCC tissues (Supplemental figure 1). In addition to that, we found that hampered liver cancer cell growth and cell migration induced by circSOD2 depletion were also rescued by DNMT3a (Figure 7B-D). Collectively, these results demonstrate that DNMT3a is the downstream gene of circSOD2 regulated signaling pathway.
3.8. STAT3 upregulates circSOD2 expression in a feedback way
Motif analysis with JASPAR found that eight STAT3 binding sites occupy circSOD2 promoter (Figure 8A), suggesting that STAT3 may regulate circSOD2 expression. To confirm this, ChIP-qPCR with STAT3 antibody was performed. DNA region that is close to circSOD2 promoter without STAT3 binding sites was used as a control. STAT3 antibody significantly enriched circSOD2 promoter compared to control IgG and control region (Figure 8B). Depletion of STAT3 also suppressed circSOD2 expression (Figure 8C-D). These results suggest that STAT3 regulates circSOD2 expression in a feedback way.