The Expression, Functions and Mechanisms of Circular RNAs in Gynecological Cancers

Circular RNAs (circRNAs) are covalently closed, endogenous non-coding RNAs and certain circRNAs are linked to human tumors. Owing to their circular form, circRNAs are protected from degradation by exonucleases, and therefore, they are more stable than linear RNAs. Many circRNAs have been shown to sponge microRNAs, interact with RNA-binding proteins, regulate gene transcription, and be translated into proteins. Mounting evidence suggests that circRNAs are dysregulated in cancer tissues and can mediate various signaling pathways, thus affecting tumorigenesis, metastasis, and remodeling of the tumor microenvironment. First, we review the characteristics, biogenesis, and biological functions of circRNAs, and describe various mechanistic models of circRNAs. Then, we provide a systematic overview of the functional roles of circRNAs in gynecological cancers. Finally, we describe the potential future applications of circRNAs as biomarkers for prognostic stratification and as therapeutic targets in gynecological cancers. Although the function of most circRNAs remains elusive, some individual circRNAs have biologically relevant functions in cervical cancer, ovarian cancer, and endometrial cancer. Certain circRNAs have the potential to serve as biomarkers and therapeutic targets in gynecological cancers.


Introduction
The Encyclopedia of DNA Elements (ENCODE) project consortium reported that only 2% of the transcribed RNAs are translated into proteins, whereas the majority of other transcripts have no apparent coding potential [1]. Increasing evidence suggests that a significant part of such non-protein coding RNAs (ncRNAs) participate in complex regulatory networks composed of other nucleic acids or proteins [2,3]. The complexity of these networks allows ncRNAs to regulate the expression of a broad spectrum of genes [2,3].
Generally, ncRNAs can be separated into small non-coding RNAs (small ncRNAs), which are shorter than 50 nt, medium non-coding RNAs, which are between 50-200 nt, and long non-coding RNAs (lncRNAs), which are longer than 200 nt [4]. MicroRNAs (miRNAs) are the most studied small ncRNAs and they typically bind to the 3 -untranslated regions (3 -UTR) of protein-coding mRNA by Figure 1. Functional mechanisms of circular RNAs (circRNAs). CircRNAs, like mRNAs, are transcribed from protein-coding genes by RNA polymerase II (Pol II). However, they are spliced in a non-canonical splicing process known as back-splicing. Following biogenesis, the majority of circRNAs are exported to the cytoplasm in a size-dependent manner. (a) CircRNAs may act as microRNA (miRNA) sponges by binding to miRNAs and inhibiting their functions, thereby releasing downstream target genes from miRNA-mediated repression. (b) CircRNAs that harbor binding sites for RNA-binding proteins (RBPs) may function as protein sponges and thus regulate gene expression. In addition, circRNAs may work as scaffolds in the assembly of protein complexes. (c) CircRNAs may associate with Pol II to enhance the transcription of their parental genes. (d) CircRNAs may compete with their linear counterparts against canonical pre-mRNA splicing, thus suppressing the expression of their parent genes. (e) CircRNAs that contain internal ribosomal entry site (IRES) elements and open reading frame may be translated into protein or polypeptide.

MiRNA Sponge
One of the most frequently studied functions of circRNA is its ability to act as a miRNA sponge. Recent observations have found that some lncRNAs act as competing endogenous RNAs (ceRNAs), which regulate other RNA transcripts by competing for shared miRNAs [25] (Figure 1a). CircRNAs that harbor multiple binding sites for miRNAs can function as efficient miRNA sponges to indirectly regulate the expression of protein-coding genes [34]. For example, circHIPK3  (Pol II). However, they are spliced in a non-canonical splicing process known as back-splicing. Following biogenesis, the majority of circRNAs are exported to the cytoplasm in a size-dependent manner. (a) CircRNAs may act as microRNA (miRNA) sponges by binding to miRNAs and inhibiting their functions, thereby releasing downstream target genes from miRNA-mediated repression. (b) CircRNAs that harbor binding sites for RNA-binding proteins (RBPs) may function as protein sponges and thus regulate gene expression. In addition, circRNAs may work as scaffolds in the assembly of protein complexes. (c) CircRNAs may associate with Pol II to enhance the transcription of their parental genes. (d) CircRNAs may compete with their linear counterparts against canonical pre-mRNA splicing, thus suppressing the expression of their parent genes. (e) CircRNAs that contain internal ribosomal entry site (IRES) elements and open reading frame may be translated into protein or polypeptide.

MiRNA Sponge
One of the most frequently studied functions of circRNA is its ability to act as a miRNA sponge. Recent observations have found that some lncRNAs act as competing endogenous RNAs (ceRNAs), which regulate other RNA transcripts by competing for shared miRNAs [25] (Figure 1a). CircRNAs that harbor multiple binding sites for miRNAs can function as efficient miRNA sponges to indirectly Cancers 2020, 12, 1472 4 of 22 regulate the expression of protein-coding genes [34]. For example, circHIPK3 modulates tumor cell growth by silencing nine miRNAs [23]. Despite the consistency of these findings, there are a few examples that challenge this ceRNA hypothesis for circRNAs [35,36]. For example, bioinformatics analysis has indicated that miRNA-binding sites are generally not enriched in circRNAs any more than could be expected by chance, with the exception of the reported circRNA, ciRS-7 [35]. In another study, the circRNAs downloaded from the circBase database were screened for Ago2-bound regions [36]. As a result, 58,063 circRNAs were found to own Ago2-bound regions. Six circRNAs were selected, and the potential ability of these circRNAs to interact with miRNAs was explored using RNA-binding protein immunoprecipitation-PCR experiments. The results revealed that these six circRNAs did not exhibit strong potential to serve as miRNA sponges [36].

Protein Sponge and Scaffold
RNA-binding proteins are a class of proteins that bind RNAs to regulate the metabolic processing of RNAs by mediating their maturation, transport, localization, translation and turnover [37] (Figure 1b). CircRNAs that contain multiple complementary binding sites to one or multiple proteins might function as protein sponges. Several circRNAs were suggested to sequester HuR, an extensively studied RNA-binding protein, which consequently affects its action on target gene regulation [38][39][40]. The interactions between circRNAs and RNA-binding proteins have been shown to influence cancer progression [41]. In addition, circRNAs were observed to act as dynamic scaffolds that bring different proteins together to form a complex [41]. In mouse fibroblast cells, circ-Foxo3 can interact with both p21 and CDK2 to form a ternary complex that results in the inhibition of cell cycle progression [42].

Transcription and Splicing
Unlike ecircRNAs, which are located in the cytoplasm, ciRNAs and EIciRNAs are primarily enriched in the nucleus. Previous studies have shown that a class of EIciRNAs interact with RNA polymerase II (Pol II) and accumulate at transcription sites, thereby promoting the transcription of their parental genes [43,44] (Figure 1c). Moreover, certain circRNA can compete with their linear counterparts against canonical pre-mRNA splicing, thus suppressing the expression of their parent genes [45,46] (Figure 1d).

Translation of circRNAs
CircRNAs were initially regarded as ncRNAs that lack protein-coding abilities, owing to the absence of a 5 cap structure and poly-(A) tail. However, subsequent studies have revealed that some circRNAs carry open reading frames, and can be translated into peptides or proteins [47,48] ( Figure 1e). The presence of internal ribosomal entry sites (IRES) was shown to drive circRNA translation through a cap-independent mechanism [49]. N6-methyladenosine (m6A) refers to the methylation that occurs in the mRNAs, and the presence of m6A on circRNA has the potential to initiate cap-independent translation [50]. The peptides or proteins encoded by circRNAs have critical functions in mediating cancer development [51,52]. Therefore, it seems necessary to determine whether the proteins originating from circRNA translation are functional in gynecological cancers.

Methods for Detecting and Characterizing circRNAs
The detailed description of bioinformatic tools, laboratory techniques and general workflow used for circRNA research has been reviewed elsewhere [53][54][55][56][57].
Previous studies have demonstrated that tumor-adjacent tissues have similar molecular changes to those found in the tumor tissues or present a unique intermediate state between healthy and tumor tissues [58][59][60]. One example is miR-29c, which is significantly upregulated in tumor-adjacent tissues when compared with gastric cancer tissues as well as non-cancerous tissues [60]. Most of the previous studies that analyze circRNA global expression in human cancer tissues have only used matched tumor-adjacent samples as normal control [61,62]. The results indicate that the expression of circRNAs in cancer tissues is frequently downregulated when compared with the matched tumor-adjacent tissues [61,62]. However, some circRNAs are actually overexpressed in tumor-adjacent tissues in comparison to cancer samples and normal samples without cancer [63], suggesting that the use of tumor-adjacent tissues for comparison purposes may lead to experimental biases regarding the true circRNA expression profile associated with tumors [63].
The functional characterization of circRNAs is commonly conducted by following two types of strategies: RNA interference (RNAi)-induced circRNA silencing and gain-of-function studies using circRNA expression plasmids. One of the challenges for circRNA overexpression is that the vectors used for overexpressing circRNA might generate many undesired transcripts, such as linear RNAs or concatenated and/or intertwined circRNAs [64,65]. Therefore, it might be difficult to determine if a specific circRNA has an impact on a particular phenotype in tumor cells. Although recently developed expression vectors were believed to efficiently drive circRNA formation, but not linear RNAs [66], the combination of RNAi-based gene silencing with vector-mediated overexpression approaches may present more opportunities to fully understand the exact functions of circRNAs in tumor cells.
Recently, the novel CRISPR/Cas13-based approaches using guide RNAs, which target the circRNA-specific back-splicing junction sites, have been used to achieve precise and robust circRNA silencing without disturbing their linear cognate mRNAs [67,68]. These studies suggest that the application of CRISPR/Cas-based circRNA engineering systems may be a promising method for performing functional studies of circRNA in human cancers.

Evidence Acquisition
This review included literature from the PubMed and Google Scholar databases published up to 20 February 2020, using the following keywords: circular RNA, circRNA, cervical cancer, endometrial cancer, ovarian cancer, vaginal cancer and vulvar cancer. All studies recognized were assessed for relevance by two authors by checking the title and abstract. After excluding the irrelevant articles, studies without access to the full-text of the publication, case reports, letters, expert opinions, meeting proceedings, review articles and non-English articles, our screening identified 83 articles that addressed the expression, roles and mechanisms of circRNAs in cervical cancer (CC), ovarian cancer (OC) and endometrial cancer (EC). These selected articles were evaluated independently by two authors. A flow diagram of the study selection process was shown in Figure 2. have only used matched tumor-adjacent samples as normal control [61,62]. The results indicate that the expression of circRNAs in cancer tissues is frequently downregulated when compared with the matched tumor-adjacent tissues [61,62]. However, some circRNAs are actually overexpressed in tumor-adjacent tissues in comparison to cancer samples and normal samples without cancer [63], suggesting that the use of tumor-adjacent tissues for comparison purposes may lead to experimental biases regarding the true circRNA expression profile associated with tumors [63]. The functional characterization of circRNAs is commonly conducted by following two types of strategies: RNA interference (RNAi)-induced circRNA silencing and gain-of-function studies using circRNA expression plasmids. One of the challenges for circRNA overexpression is that the vectors used for overexpressing circRNA might generate many undesired transcripts, such as linear RNAs or concatenated and/or intertwined circRNAs [64,65]. Therefore, it might be difficult to determine if a specific circRNA has an impact on a particular phenotype in tumor cells. Although recently developed expression vectors were believed to efficiently drive circRNA formation, but not linear RNAs [66], the combination of RNAi-based gene silencing with vector-mediated overexpression approaches may present more opportunities to fully understand the exact functions of circRNAs in tumor cells.
Recently, the novel CRISPR/Cas13-based approaches using guide RNAs, which target the circRNA-specific back-splicing junction sites, have been used to achieve precise and robust circRNA silencing without disturbing their linear cognate mRNAs [67,68]. These studies suggest that the application of CRISPR/Cas-based circRNA engineering systems may be a promising method for performing functional studies of circRNA in human cancers.

Evidence Acquisition
This review included literature from the PubMed and Google Scholar databases published up to 20 February 2020, using the following keywords: circular RNA, circRNA, cervical cancer, endometrial cancer, ovarian cancer, vaginal cancer and vulvar cancer. All studies recognized were assessed for relevance by two authors by checking the title and abstract. After excluding the irrelevant articles, studies without access to the full-text of the publication, case reports, letters, expert opinions, meeting proceedings, review articles and non-English articles, our screening identified 83 articles that addressed the expression, roles and mechanisms of circRNAs in cervical cancer (CC), ovarian cancer (OC) and endometrial cancer (EC). These selected articles were evaluated independently by two authors. A flow diagram of the study selection process was shown in Figure 2.

CircRNAs in CC
CC is the fourth most common female malignancy in the world and kills more than 300,000 deaths worldwide each year [69]. High-risk subtypes (16 and 18) of human papillomavirus (HPV) are well established as the causative agents responsible for CC [69]. Massive studies have shown that a large number of circRNAs were differentially expressed in CC tissues compared with adjacent normal cervical tissues, and play oncogenic or tumor-suppressive roles in the initiation and progression of CC . All of the circRNAs that are abnormally expressed in CC are summarized in Table 1, and their functions in CC are shown in Figure 3. As demonstrated by many studies, circRNAs function as miRNA sponges to regulate different aspects of CC progression through regulating multiple signaling pathways ( Figure 3). These studies demonstrate that the circRNA-miRNA-mRNA regulatory networks and an HPV E7 oncoprotein-encoding circRNA play crucial roles in controlling CC tumorigenesis and progression.  Diverse circRNAs (such as circEIF4G2, circ0007534, circ0005576, circ0075341, circ0000515, circAMOTL1, and circSLC26A4) were previously described to promote the tumorigenesis of CC by increasing the respective downstream targets mRNAs, including HOXA1, BMI-1, KIF20A, AURKA, ELK1, AMOTL1 and HOXA7 via sponging miRNAs [84,90,91,94,95,97,101]. Additionally, circRNAs (such as circ0018289, circ0000285 and circMYBL2,) were found to be overexpressed in CC tissues, and their upregulation enhanced CC cell invasion and metastasis [76,92,98].
The PI3K/AKT signaling pathway controls multiple cellular processes including proliferation, growth, invasion and chemoresistance, and it is one of the most frequently dysregulated pathways in human cancers, including CC [106,107]. Integrin β1 can promote cancer initiation and progression through the activation of multiple downstream pathways, including the PI3K/AKT signaling [108]. By sponging miR-361-5p, circCSPP1 stimulates the expression of integrin β1, allowing downstream activation of the PI3K/AKT pathway, which contributes to CC cell proliferation and migration [105]. The RAF/MEK/ERK cascade plays a critical role in the development and progression of cancer [109]. A recent study suggested that circAGFG1 was upregulated in CC cell lines compared with normal cervical cells, and it could induce CC cell proliferation and migration by elevating RAF1 expression and activating the MEK/ERK pathway, via sponging miR-370-3p (a suppressor of RAF1) [93]. Epithelial-mesenchymal transition (EMT) is a biological process in which epithelial cells acquire mesenchymal phenotypes, thus conferring the malignant properties to cancer cells including invasive behaviors, cancer stem cell properties, and greater resistance to cancer treatment [110]. Experimental studies have revealed that circRNAs affect the EMT program in human cancers (including CC) through different mechanisms [77,80,89,99,100,104,111]. For instance, circ000284 indirectly increases Snail-2 expression to promote EMT and invasion of CC cells through sponging miR-506 [77]. CircCLK3 induces the levels of FOXM1 via sponging miR-320a, and overexpression of FOXM1 could trigger EMT-like changes, including the acquisition of Vimentin and N-cadherin expression and reduction of E-cadherin expression [99]. Also, the upregulation of circ0067934, circ0000745, ciRS-7 and circHIPK3 in CC cells was shown to increase the expression of mesenchymal markers and decrease the expression of epithelial markers [80,89,100,104]. Inactivation of p53 in human tumors occurs through missense mutations in the DNA-binding domain of the p53 protein, and can also result from the amplification/overexpression of its specific inhibitors (such as MDM2 and MDM4) [112]. Circ0000263 can increase the expression of MDM4 by sponging miR-150-5p, thus causing degradation of p53 in CC cells [81].

CircRNA Encodes HPV E7 Oncoprotein
The oncoproteins E6 and E7 encoded by HPV16 and 18 contribute to the neoplastic transformation of squamous cervical epithelium [113]. The E6 protein from high-risk HPVs binds the p53 tumor suppressor protein for degradation [114], whereas the E7 protein interacts with the retinoblastoma protein and inactivates this cellular protein [115]. A recent report showed that oncogenic HPV16 generates a 472nt circular RNA called circE7, which contains the entire E7 open reading frame [87]. In this study, circE7 was shown to be translated to the E7 oncoprotein, and knockdown of circE7 in CC-derived CaSki cells reduced E7 protein levels and suppresses tumor growth and malignancy both in vitro and in vivo [87]. These results suggest that circE7 appears to be essential for E7 protein expression and the transformation of CaSki cells.

CircRNAs as Tumor Suppressors in CC
Recent data suggest that circSMARCA5 was downregulated in CC tissues compared with adjacent normal tissues, and overexpression of circSMARCA5 suppresses the proliferation and invasion of HeLa cells by sponging miR-620 [82]. However, controversial findings have been reported in another study showing that circSMARCA5 was expressed at higher levels in CC tissues, and knockdown of circSMARCA5 suppressed cellular growth and invasion [102]. The oncogenic functions of circSMARCA5 were mediated through a miR-432/EGFR axis [102]. Besides, the expression of circITCH was significantly decreased in CC tissues compared with the adjacent normal tissues [103]. CircITCH acts as an endogenous miR-93-5p sponge to induce the levels of FOXK2 in CC cells, leading to the suppression of the malignant behaviors [103].

CircRNAs in OC
Expression profiling studies and bioinformatics analysis have revealed significant changes in circRNA expression between OC tissues and normal ovarian tissues [116,117], between miliary-type OC and non-miliary OC [118], between serum samples from OC patients and those samples from healthy controls [119], between primary and metastatic sites of OC [120], and between OC cells and non-malignant human ovarian surface epithelial cells [62]. The expression and mechanisms of circRNAs in OC are presented in Table 2 and their biological roles are shown in Figure 4.    Several circRNAs (such as circEPSTI1, circGFRA1, circ0051240, circPIP5K1A, circPVT1 and circABCB10) have been found to increase the proliferative and invasive capacities of OC cells by acting as miRNA sponges [121][122][123][124][125][126]. Knockdown of circSMAD7 by short-hairpin RNA (shRNA) was associated with decreased cell proliferation, migration and invasion and with increased expression of KLF6 in CC cells [127]. Silencing of circANKRD12 reduced OC cell proliferation, probably through downregulating the expression of cyclin D1 [128].
CircCSPP1 was shown to be highly expressed in OC tissues and borderline tumor tissues when compared with normal ovarian tissues [129]. Further investigations suggest that circCSPP1 sponged miR-1236-p to increase the expression of ZEB1, leading to increased cell invasion and EMT phenotypes [129]. Circ0061140, which was upregulated in OC cell lines relative to normal ovarian epithelial cells, was reported to stimulate OC cell proliferation, invasion and EMT properties, probably via binding to miR-370 [130]. Overexpression of FOXM1, a target of miR-370, rescued the EMT properties that were reduced by either circ0061140 silencing or miR-370 overexpression [130]. By sponging miR-144, circUBAP2 promoted the expression of N-cadherin in OC cells [131]. CircWHSC1 was increasingly expressed in OC tissues compared with normal ovarian tissues, and there was a tendency towards increased levels of circWHSC1 expression in poorly-and moderately-differentiated OCs compared with well-differentiated OCs [132]. By interacting with miR-145/miR-1182, circWHSC1 was shown to elevate the levels of two EMT-inducer genes (MUC1/hTERT) in OC cells [132][133][134]. Inhibition of a circRNA VPS13C-has-circ-001567 by shRNA attenuated cellular invasion, resulting in the downregulation of N-cadherin and upregulation of E-cadherin [135].
CircCELSR1 acts as a sponge for the tumor suppressor miR-1252, and suppression of miR-1252, in turn, contributes to paclitaxel resistance in OC cells, possibly by inducing the expression of its target gene FOXR2 [136].
CircPUM1 was found to be highly expressed in OC tissues compared with normal ovarian tissues, and higher circPUM1 expression was positively correlated with advanced FIGO stage [137]. Studies in both OC cell lines and mouse OC models suggest that circPUM1 promoted cell invasiveness and growth by sponging miR-615-5p and miR-6753-5p to upregulate NF-κB and MMP2 expression [137]. Interestingly, circPUM1 was present in OC cell-derived extracellular vesicles, and exosomal circPUM1 facilitated tumor peritoneal dissemination in vivo [137].

CircRNAs Promotes Angiogenesis in OC
Angiogenesis is influenced by the microenvironment and tightly regulated by a balance between pro-and anti-angiogenic factors [138]. Vascular endothelial growth factor (VEGF-A) has a prominent role in tumor angiogenesis [139]. CircRhoC, a circRNA derived from the RhoC mRNA, was overexpressed in OC tissues relative to normal ovarian tissues, and higher circRhoC expression was detected in advanced-stage tumors compared with early-stage tumors [140]. Suppression of circRhoC with shRNA significantly attenuated cell migration and invasion in vitro, and inhibited the intraperitoneal dissemination of OC cells in vivo [140]. By performing mechanistic analysis, the same study indicated that circRhoC can serve as a sponge for miR-302e to upregulate the expression of VEGF-A, which potentially promotes OC angiogenesis and metastasis [140].

CircRNAs as Tumor Suppressors in OC
Some circRNAs exert their tumor-suppressive roles in OC cells [141][142][143][144][145][146][147][148][149][150] (Table 2 and Figure 4). According to these reports, the "miRNA" sponge seems to be the most common mechanism by which circRNAs regulate the malignant properties of OC cells. Interestingly, one study proposed a different potential mechanism whereby circITCH might inhibit the proliferation of OC cells by downregulating the levels of lncRNA HULC, although the interaction between circITCH and HULC has not yet been defined [143]. Aberrant activation of the Wnt/β-catenin signaling plays a key role in promoting the cancer stem cell self-renewal, metastasis and chemoresistance in OC [151]. By sponging miR-9, circPLEKHM3 functions as a tumor suppressor to enhance the expression of BRCA1, DNAJB6 and KLF4, which consequently inactivate the PI3K/AKT and Wnt/β-catenin signaling pathways in OC cells [149].

CircRNAs in EC
To date, very few reports on circRNAs in EC have been published [152][153][154][155][156]. The expression profile of circRNAs was found to be altered in EC tissues compared with the adjacent normal tissues [152]. Similarly, another study identified 62,167 circRNAs that were differentially expressed in grade 3 EC tissues when compared with the adjacent normal tissues [153]. A recent study indicated a positive correlation between RNA-binding protein QKI and circRNAs, and a negative correlation of QKI with the activity of specific miRNAs, suggesting a potential model where QKI, circRNAs, and miRNAs form a regulatory feedback loop in EC [154]. Another study identified differentially expressed circRNAs (including 209 upregulated circRNAs and 66 downregulated circRNAs) in extracellular vesicles isolated from the serum of EC patients compared with healthy controls [155]. CircPUM1 was significantly overexpressed in EC tissues and EC cell lines, and it promoted the development of EC by regulating the miR-136/NOTCH3 axis [156].

CircRNAs as Potential Biomarkers in Gynecological Cancers
Early studies have provided evidence that circRNAs could be used as biomarkers in distinguishing cancer tissues from normal tissues and predicting the progression and prognosis, and assessment of response to therapy in many cancer types [15]. Dysregulation of circRNAs has been reported to be closely related to the clinicopathological characteristics of patients with CC, OC and EC [79,80,94,117,149,153]. In addition, levels of some circRNAs have been associated with survival times in patients with CC and OC [78,80,141,142,[157][158][159][160]. In an in silico analysis of an OC dataset obtained from the GEO database, the expression of six-circRNAs was found to be an independent predictor of overall survival in patients with OC [160].
Moreover, circRNAs can be released into the extracellular space, and subsequently detected in body fluids (such as blood, plasma, serum and exosomes) from patients with gynecological cancers [117,119,137,[161][162][163]. Higher serum levels of circSETDB1 have been found in patients with OC compared to healthy controls [161]. Serum levels of circSETDB1 were positively correlated with chemoresistance and shorter survival in patients with OC [161]. CircBNC2 was downregulated in the plasma of patients with OC compared to healthy individuals, and plasma levels of circBNC2 provided a potential tool to distinguish patients with early-stage OC from healthy individuals [164].
The circRNAs that might serve as potential prognostic biomarkers in gynecological cancers are summarized in Table 3.

CircRNAs as Potential Therapeutic Targets for Gynecological Cancers
Recent data have proved that several circRNAs can influence the growth, survival, invasion, EMT, angiogenesis, and drug resistance in gynecological cancers. Manipulation of circRNA abundance has been shown to possess anti-tumor activities in experimental xenograft models of CC [88,91,95], OC [121,123,124] and EC [156].
The introduction of circRNAs with binding sites for oncogenic miRNAs (or proteins) might sequester miRNAs (or proteins) from native targets, and consequently inhibit the biological activities of tumor-promoting miRNAs (or oncoproteins) in tumor cells [15]. Generally, numerous oncogenic miRNAs are overexpressed in a given tumor, thus artificial circRNAs that contain multiple binding sites for multiple oncogenic miRNAs would be expected to simultaneously bind and inhibit the functions of different tumor-causing miRNAs. Alternatively, circRNAs might be used as stable carriers that deliver tumor-suppressive miRNAs (or proteins) to tumor cells.
In addition, exosomal circRNAs could modulate the progression of cancer, the remodeling of tumor microenvironment, anti-tumor immune response, and the occurrence of metastasis [165]. The discovery that mRNAs from tumor cells can be transfected into dendritic cells and further translated into proteins to function as antigens are intriguing [166,167]. Similar to mRNA, circRNA can be detected in exosomes.
Thus, the circRNAs with protein-coding potential might function as potential antigens and be used as potent targets for cancer immunotherapy [168]. Collectively, the research indicates the potential use of circRNAs as therapeutic targets for the treatment of gynecological cancers.

Conclusions
Despite many exciting advances in the circRNA research area, there are still arguments regarding the functionality of circRNAs [169]. At least, some individual circRNAs have biologically relevant functions in ovarian, cervical and endometrial cancer cells [87,137,156]. Their biogenesis, functions, and action mechanisms in gynecological cancer cells need to be fully investigated. Recent studies suggest that circRNAs may fulfill a regulatory role in gynecological cancers by sponging miRNAs, although the possibility of other mechanisms (such as scaffolds for protein complexes) also exists. CircRNAs have recently been recognized as emerging prognostic biomarkers for gynecological cancers. The unique ability of circRNAs to sponge RNAs (or proteins) and code for proteins make them valuable as potential therapeutic tools to treat gynecological cancers.
Author Contributions: Writing-review and editing, P.D., D.X., Y.X., J.Y., K.I., Y.K., and H.W. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by a grant from JSPS Grant-in-Aid for Scientific Research (C) (18K09278 and 19K09769) and an NIH/NCI grant 1R21CA216585-01A1 to J. Yue.