Functional roles of circular RNAs during epithelial-to-mesenchymal transition

Previous studies have demonstrated that TGF-β can enhance the EMT process. The TGF-β/Smad pathway has been demonstrated to be a positive regulator in the growth and metastasis of various cancer types [58, 59]. Wang L et al. reported that circPTK2 inhibits non-small-cell lung carcinoma (NSCLC) metastasis by disrupting oncogenic miR-429 and miR-200b-3p and promoting transcriptional intermediary factor 1γ (TIF1γ) expression [60]. TIF1γ, which is considered as a negative regulator of the TGF-β/Smad pathway [61, 62], represses the TGF-β-related EMT. Zeng K et al. demonstrated that the high expression level of circANKS1B can be indicative of a poor prognosis in triple-negative breast cancer (TNBC). The upregulated circANKS1B has been implicated in promoting breast cancer invasion and metastasis. The pro-metastatic effect of circANKS1B is mediated through the interaction with miR-148a-3p and miR-152-3p, thereby increasing upstream transcription factor 1 (USF1) expression [63]. USF1 binds to TGF-β1 in murine tissues [64]. Zeng K et al. confirmed that upregulated USF1 can interact with TGF-β1 in TNBC tissues to substantially upregulate the expression of vimentin, p-Smad2, and p-Smad3 but significantly decrease the expression of E-cad. Moreover, circANKS1B expression is regulated by USF1. USF1 overexpression upregulates circANKS1B expression via the splicing factor Epithelial splicing regulatory protein 1 (ESRP1), showing that a positive feedback regulatory loop is formed between circANKS1B, USF1, and ESRP1 [63]. Splicing factors (SFs) can also regulate circRNA formation and EMT progression by inserting SF-binding motifs into flanking introns, which has been demonstrated in several research reports [18, 63, 65, 66].

In general, the Wnt pathway can be classified as a canonical pathway (Wnt/β-catenin pathway) and a noncanonical pathway (Wnt/PCP and Wnt/Ca pathways). Here, we mainly discuss the canonical Wnt/β-catenin pathway. In brief, in the absence of Wnt signaling, β-catenin, Axin, glycogen synthase kinase-3, adenomatous polyposis coli (APC), and casein kinase 1 form a destruction complex in the cytoplasm in which the initially phosphorylated and subsequently ubiquitinated β-catenin is degraded by the proteasome. The presence of Wnt ligands blocks the formation of this complex. As a result, β-catenin accumulates to a certain level in the cytoplasm and then translocated into the nucleus to activate the Wnt-targeted genes [67‐70]. The interaction between circRNAs and the Wnt/β-catenin pathway to promote cancer progression has been demonstrated (Table 2).

Frizzled gene family proteins (FZDs) are important cell surface receptors in the Wnt/β-catenin signaling pathway and consists of FZD1/2/3/6/7, FZD5/8, FZD4, and FZD9/10 subfamilies. FZDs are positive membrane receptors in the Wnt/β-catenin pathway and are significantly associated with cancer progression [71‐73]. Guan F et al. demonstrated the obvious overexpression of circRNA_100,290 and FZD4 in colorectal cancer (CRC) and found their target relationship with miR-516b. CircRNA_100,290 not only increases proliferation and metastasis by regulating FZD4 but also upregulates the expression of MYC, CCD1, CCD2, TCF7, and SOX4, which are target genes of the Wnt/β-catenin signaling pathway [74]. In thyroid cancer, circNEK6 is a competing endogenous RNA of FZD8 that absorbs miR-370-3p, resulting in the activation of the Wnt/β-catenin signaling pathway. The upregulation of circNEK6 enhances the expression of c-myc and CCD1, whereas silencing FZD8 abrogates their expression [75]. Chen et al. demonstrated that hsa_circ_0000177 is a miR-638 sponge that targets FZD7. Furthermore, hsa_circ_0000177 activates the Wnt/β-catenin pathway and promotes the growth of glioma cells by enhancing FZD7 expression [76]. Xia et al. showed that circ-CBFB ultimately leads to CLL progression via the FZD3-induced activation of Wnt/β-catenin signaling by sponging miR-607 [77].

Dickkopf-1 (DKK1) can specifically bind to LRP5/6, thereby interfering with the formation of the Wnt-LRP5/6-FZD complex and inhibiting the downstream pathway [78, 79]. Yao et al. illustrated that the significantly downregulated circ_0006427 is involved in the tumorigenesis of lung adenocarcinoma (LUAD) by releasing miR-6783-3p and by activating the Wnt signaling pathway through DKK1 knockdown. Circ_0006427 downregulation or DKK1 knockdown promotes cell migration and invasion [80]. Y. Jin et al. demonstrated a significant decrease in hsa_circ_0000523 and DKK1 in CRC and identified their target correlation with miR-31. A decreased DKK1 level not only enhances β-catenin in the nucleus but also promotes CRC proliferation [81].

APC is considered a negative regulator in the Wnt/β-catenin signaling pathway and is associated with the β-catenin destruction complex [82‐84]. Hsa_circ_0002052 and APC2 are obviously decreased in osteosarcoma (OS) and are strongly correlated. As a result, the downregulation of hsa_circ_0002052 promotes proliferation and metastasis through the Wnt/β-catenin pathway. MiR-1205 has been identified as a common target miRNA of hsa_circ_0002052 and APC2 and is upregulated in OS. Hsa_circ_0002052 also provides a novel therapeutic target for OS therapy [85].

ITCH is a E3 ubiquitin ligase enzyme and can inactivate the Wnt/β-catenin pathway by degrading phosphorylated dishevelled Dvl [86‐88]. Dvl can bind to Axin, which is an important component of the β-catenin destruction complex, to positively modulate the β-catenin level. Several studies have demonstrated the positive relationship between circ-ITCH and ITCH. In TNBC, circ-ITCH exhibits breast tumor inhibition and can promote the expression of ITCH tumor suppressor by absorbing miR-214 and miR-17 [89]. Feng Li et al. illustrated the anti-oncogenic effect of circ-ITCH on glioma. Decreased circ-ITCH and ITCH can promote migration and invasion by sponging miR-214. In this mechanism, miR-214 and miR-17 can bind to the 3′UTR of ITCH to enhance ITCH expression [89, 90].

Sirtuin 1 (SIRT1) exerts diverse effects on tumor cells as an oncoprotein or tumor suppressor [91‐93]. Yao Y et al. demonstrated that circ_0001946 expression is increased in lung adenocarcinoma. Circ_0001946 upregulates SIRT1 and activates the Wnt/β-catenin pathway in LAC by targeting miR-135a-5p, thus promoting LAC progression [94]. However, the mechanism involving circRNA and SIRT1 remains largely unknown.

Previous studies have found that catenin beta-1 (CTNNBIP1) is a negative regulator in the Wnt/β-catenin pathway and can interact with β-catenin. Mechanistically, increasing evidence has shown that the CTNNBIP1/β-catenin interaction can prevent the formation of the β-catenin/TCF/LEF complex. As a result, it prevents the activation of the Wnt/β-catenin signaling pathway [95‐98]. The upregulated circRNA_102171 facilitates the malignant behavior of papillary thyroid cancer by regulating the CTNNBIP1-mediated activation of the Wnt/β-catenin pathway. CircRNA_102171 overexpression enhances the β-catenin/TCF/LEF interaction while blocking the association between CTNNBIP1 and β-catenin. Nevertheless, how circRNA_102171 modulates CTNNBIP1 still needs further studies [99].

In addition to the regulation exerted by circRNAs at the level of transcription factors and EMT-related signaling pathways, certain circRNAs function by directly or indirectly regulating the expression of cell adhesion molecules and cytoskeletal proteins. Such deregulated circRNAs can act either as tumor suppressors or oncogenes to control cell proliferation, migration and metastasis. For example, circ_0058063 promotes bladder cancer progression by sponging miR-145-5p and regulating CDK6 expression [100]. In HCC, the circRNA circMTO1 (mitochondrial translation optimization 1 homologue) significantly downregulated and circMTO1 overexpression suppressed HCC cell proliferation and invasion by sponging oncogenic miR-9. Moreover, circMTO1 is an important circRNA in bladder cancer tissue. A decreased circMTO1 level was positively correlated with bladder cancer cell metastasis. Further research found that circMTO1 was able to sponge miR-221, and ectopic expression of circMTO1 negatively regulated the E-cad/ N-cad pathway to inhibit bladder cancer cell EMT by competing for miR-221 [101].

Several circRNAs have been demonstrated to regulate vimentin transcription. For instance, Tao Wang et al. reported on the role of circP4HB in NSCLC. In NSCLC, circP4HB levels are significantly higher than in healthy lung. Of note, vimentin has a documented role in promoting EMT and metastasis in NSCLC as a regulator of cell to cell adhesion and cell motility. Interestingly, circP4HB can positively regulate vimentin and enhance EMT and metastatic disease through miR-133a-5p sequestration [102]. A recent circRNA profiling study showed that circRBM23 expression was upregulated, whereas miR-138 expression was decreased, in HCC tissues. Downregulation of circRBM23 decreased cell viability, proliferation, and migration and promoted the expression of miR-138 and its related target genes vimentin and CCND3 [103]. Thus, the circRNA-miRNA-vimentin regulatory axis in EMT is important for tumor invasion and metastasis.

Protocadherins (PCDHs) are classified as the largest subgroup within the cadherin superfamily of calcium-dependent cell-cell adhesion molecules that are mainly expressed in the nervous system and have been implicated in neural cell-cell interactions. Little is known about the functions of PCDHs, but some members, such as PCDH8 and PCDH10, have been shown to suppress tumor activity [104‐107], suggesting that PCDHs may act as tumor suppressors by influencing tumor growth and metastasis. In prostate cancer (PCa), Zhan Yang et al. have shown that an Amotl1-derived circRNA termed circAMOTL1L, is downregulated in PCa and that low expression of circAMOTL1L facilitates PCa cell migration and invasion by downregulating E-cad and upregulating vimentin, which leads to EMT and PCa progression. Mechanistically, the study demonstrated that circAMOTL1L serves as a sponge for binding miR-193a-5p in PCa cells, relieving the miR-193a-5p-mediated repression of the Pcdha gene cluster, which is a subset of PCDHs [108].

Carcinoma cells that undergo EMT reorganize the epithelial actin cytoskeleton into one with actin stress fibers and form migratory organelles, such as lamellipodia and filopodia. In this way, cells acquire directional motility and metastatic potential [109]. Rho GTPases belong to the group of small G proteins that cycle between an inactive GDP-bound state and active GTP-bound state. Among the members of the Rho GTPase family, RhoA, cdc42 and Rac have been studied in most detail. RhoA facilitates the formation of actin stress fibers, while cdc42 and Rac1 generally assist in filopodia and lamellipodia formation. Rho-kinase (ROCK), as the downstream effector of RhoA, induces myosin light chain (MLC) phosphorylation by inhibiting MLC phosphatase to enhance myosin contractility. ROCK also activates LIM kinase (LIMK) and then represses cofilin. The Rho/ROCK/MLC/myosin and Rho/ROCK/LIMK/cofilin pathways all facilitate stress fiber formation. Moreover, actin polymerization dependent on mDia is an essential factor for the assembly of stress fibers [110, 111]. CircHIAT1 expression was suppressed by androgen receptor (AR) in clear cell renal cell carcinoma (ccRCC). Interestingly, circHIAT1 could stabilize the expression of miR-195-5p/29a-3p/29c-3p as a reservoir. It has been demonstrated that the downregulating miR-195-5p/29a-3p/29c-3p targets the cdc42 3′-UTR to enhance the migration and invasion of ccRCC cells. Notably, upregulation of AR-targeting cdc42 significantly promoted filopodia formation, while the increase could be partially reversed by CircHIAT1 overexpression [112]. Jie Li et al. investigated the role of circ-IARS in pancreatic cancer tissues. Circ-IARS was secreted by pancreatic cancer cells and upregulated in plasma exosomes. Then, circ-IARS enters HUVECs through exosomes. In this study, upregulation of circ-IARS in HUVECs changed their biological functions. It has been demonstrated that circ-IARS sponges miR-122 to increase RhoA activity and F-actin expression and reduce ZO-1 expression. As a result, circ-IARS/RhoA modulates the permeability of the endothelial monolayer to promote metastasis [113].

High-mobility group box 1 (HMGB1) is considered a chromatin-binding factor that regulates HCC progression [114‐116]. HMGB1 knockdown can significantly attenuate the migration and invasion behavior [117, 118]. Li S et al. demonstrated the correlation between a circRNA and the HMGB1/RAGE pathway. Upregulating circRNA_101,368 suppresses miR-200a expression, thus enhancing the metastasis of HCC cells by activating the miR-200a-mediated HMGB1/RAGE pathway [119].

MiR-7 is considered a tumor cell suppressor whose dysregulation facilitates malignant behavior [120, 121]. The experimental basis for the targeted treatment from the perspective of the circRNA CDR1as/miR-7 pathway has been validated in several cancer cells. Su C et al. demonstrated that the remarkably increase in ciRS-7 (circRNA CDR1as) induces the metastasis of NSCLC cells through the miR-7 downstream NF-κB pathway [122]. Furthermore, the circRNA CDR1as/miR-7 pathway regulates the EMT phenotype. Furthermore, a miR-7 inhibitor can enhance the metastatic ability of HCC and OS cells [123, 124].

Yu J et al. linked the circRNA cSMARCA5/miR-17-3p/miR-181b-5p/TIMP3 regulatory pathway to EMT in HCC [125]. TIMP3 has been reported to be a tumor cell suppressor in several studies and inactivates the EMT program [126, 127]. It has been found that the decrease in circRNA cSMARCA5 facilitates HCC cell migration and metastasis and downregulates TIMP3 transcription by absorbing miR-17-3p and miR-181b-5p [125].

Human mesenchymal stem cells (MSCs) are multipotent cells that possess the ability to self-renew and differentiate into mesodermal lineage cells [128, 129]. Due to this differentiation potential and other properties to regenerate injured tissues indirectly via growth factor secretion and immunomodulation, MSCs hold promise for regenerative medicine. A full understanding of the molecular mechanisms that regulate the maintenance of human MSC identity and their uncommitted state is helpful for improving therapeutic efficacy. Stem cell plasticity and identity are controlled by master regulatory genes and complex circuits involving noncoding RNAs [130‐132]. Alessandro Cherubini et al. showed that compared to that in differentiated mesodermal cells, circFOXP1 levels were enriched in MSCs. The authors demonstrated a direct interaction between circFOXP1 and miR-17-3p/miR-127-5p, which resulted in the modulation of the EGFR and noncanonical Wnt pathways. They found a regulatory role for circFOXP1 as a gatekeeper of pivotal stem cell molecular networks [133]. Yanjie Wang et al. demonstrated that circRNA_014511 could affect the expression of P53, regulate cell apoptosis and cell cycle arrest, and influence the radiosensitivity of bone marrow mesenchymal stem cells (BMMCs) by adsorbing miR-29b-2-5p. Furthermore, Yan-Jing Zhu et al. found that circZKSCAN1 suppressed cell stemness in HCC by regulating the function of the RBP fragile X mental retardation protein (FMRP), whose downstream target gene is cell cycle and apoptosis regulator 1 (CCAR1) and showed that CCAR1 acts as a coactivator of the Wnt/β-catenin signaling pathway and upregulates cell stemness [134].