According to the formed sequence, circRNAs can be divided into six categories: exonic circular RNAs (ecircRNAs), circular intronic RNAs (ciRNAs), exon-intron circular RNAs (EIciRNAs), intergenic circRNAs, anti-sense circRNAs, and tRNA intronic circRNAs (tricRNAs) (10).

CircRNAs have no ends so it is highly stable and have specific spatiotemporal expression patterns. Numerous circRNAs usually express in specific tissues and specific developmental stages (11). CircRNAs were found to be evolutionarily conserved in diversity and the conservation is different in different tissues, among which the most conservative in the brain (12). Jeck et al. identified over 25,000 different circRNAs in human fibroblasts (13). In addition, circRNAs are widely distributed and have been reported in thousands of animal cells, such as humans, mice, and nematodes, and expressed in high abundance (11, 14). Rybak-wolf et al. found that circRNAs are abnormally enriched in the mammalian brain and are specifically and dynamically expressed in neuronal differentiation (15). New classification methods such as differences in length, stability, function, and characteristics of circRNAs still need to be continuously explored.

CircRNAs are derived from the reverse splicing mechanism of pre-mRNAs to form a single-stranded closed loop (16). However, the biogenesis mechanism has not been fully elucidated, circRNAs can be derived from exons, introns, 3’UTR, 5’UTR, intergenic regions, or antisense sequences ( Figure 1 ). As early as 2013, Jeck et al. proposed two circRNAs cyclization models: Lariat-driven circularization and intron-pairing-driven circularization (13). Lariat-driven circularization, also known as exon skipping, is connected by non-adjacent exons as donors and acceptors to form a lasso structure. Then the introns in the lasso are removed, resulting in ecircRNAs. The biogenesis of circRNAs is mainly studied in ecircRNAs, other types of circRNAs are rarely studied. EcircRNAs account for 80% of identified circRNAs (17). The biogenesis of eicircRNAs is similar to ecircRNAs. In some cases, the intron portion is completely sheared to form eicircRNAs. Another model is intron-pairing-driven circularization, also known as direct reverse splicing. This model is based on the direct base pairing of the flanking introns, and then the introns are removed to form a ring structure (17). Zhang et al. reported a class of intron-derived circRNAs, namely CiRNAs (5). The formation of ciRNAs depends on the 7 nucleotides GU enrichment element near the 5’ splicing site and the 11 nucleotides C-rich element near the branch site. Zhang et al. believed that the exon cycle depends on the complementary sequences of the flanking introns. Intergenic circRNAs are formed from genes outside known genetic loci (11). TricRNA is formed by intron excision of the pre-tRNA by the tRNA splicing mechanism, followed by intron release and binding into tRNA and TricRNA (18).

In addition, multiple factors are involved in the biogenesis of circRNAs. Zhang et al. reported that the exon cycle depends on the complementary sequences of the flanking introns (19). Some RBPs play an important role in reverse splicings, such as MBL(splicing factor muscleblind), QKI(Quaking), and FUS(fused in sarcoma), binding to both sides of the flanking intron sequence enhances exon cycling by tightly linking the 3 ‘and 5’ ends of circRNAs. Thereby promoting exon circulation. Muscleblind is a splicing factor for MBL-derived genes and MBL in Drosophila promotes the production of circRNAs from the second exon of its own pre-mRNA by binding to flanking introns (20). The QKI of the STAR family is a tumor suppressor protein with three isoforms, all of which have the same KH domain but have different 3’UTRs. Among them, QKI-5, the most abundant nuclear isoform, acts on circRNAs during splicing. QKI dimerizes through its N-terminal Qua1 domain and binds to two-part sequence motifs that can be located on the same or separate RNA molecules (21). The investigation of PAR-CLIP cross-linking in human embryonic kidney cells (HEK293T) indicated that the majority of QKI binding occurs within introns and is responsible for circRNA synthesis, which limits proliferation and EMT during human cancer (22). In addition, Conn et al. also introduced consensus binding sequences for QKI in flanking introns to enable circRNAs to be generated from exons that normally only undergo canonical linear splicing (21). A recent study showed that overexpression of circ-SHPRH in cadmium-transformed BEAS-2B cells promoted the expression of QKI and significantly inhibited cell proliferation, EMT, invasion, migration, and non-anchored growth. Conclusions of Conn et al. (23). FUS was first reported to be involved in circRNA generation in the nervous system in 2017 (24). Cao et al. recently discovered that the nematode homologous gene FUST-1 promotes the creation of numerous circRNAs while having no effect on the analogous linear mRNA, regulating exon skipping and reverse splicing, surprisingly, CLIP-seq results suggest that FUS attaches to stem-loop secondary structure rather than particular sequences (25).

Besides, negative regulators destroy the stability of intron interactions, thereby reducing the cyclization efficiency, such as adenosine deaminases acting on RNA 1 (ADAR1) reduces the efficiency of cyclization by disrupting the base pairing between flanking introns through the A to I RNA editing mechanism (15). ADAR is an adenosine deaminase that is widely expressed in humans and can be applied to RNA modification. ADAR systems are used for programmable RNA editing in vitro and in vivo by recruiting ADARs to target RNA sequences using ADAR recruitment guide RNAs (adRNAs). Two recent studies have designed circRNAs that can recruit ADARs to improve RNA editing efficiency. Katrekar et al. engineered a highly stable circular ADAR-recruiting guide RNA (cadRNA) to recruit endogenous ADARs, improving the efficiency and durability of RNA editing (26). The engineered circ-arRNAs designed by Yi et al.’s LEAPER2.0 system have much higher editing efficiency than the corresponding linear arRNAs, which greatly improves the efficiency and robustness of RNA editing (27).

Interestingly, UAP56 and URH49 proteins can assist the transport of circRNAs from the nucleus to the cytoplasm (28). This discovery is novel, however, many of the regulatory factors involved in circRNAs biogenesis remain unclear and require more research. A better understanding of the biogenesis mechanism of circRNAs will lead to a better understanding of their specific roles in cancer development. Future studies can explore the levels of specific key factors that regulate the biogenesis of circRNAs, which will also provide innovative strategies for cancer treatment and prevention.

In addition to its unique way of formation, how circRNAs participate in the process of biogenesis has also deepened our understanding of circRNAs. However, the functions of most circRNAs are still unknown. Recent studies show that the functions of circRNAs are mainly ceRNA or miRNA sponging, binding with proteins, regulation of pre-RNA splicing, regulation of parental gene expression, and potential translation templates for proteins or peptides ( Figure 1 ).

Early RNA sequencing did not identify circular RNAs without a ployA tail, and non-linear fragments were often considered errors and were ignored. This section introduces some traditional circRNAs detection techniques and emerging technical methods ( Table 1 ). Northern blot, qRT-PCR, RNA-seq, and Microarrays are examples of traditional circRNA detection techniques (63). However, previous RNA analysis methods are also difficult to study circRNAs. RNA-seq detection of RNA detection efficiency is low, so many low abundance circRNAs may be missed. And microarray technology has been used to detect linear RNA for a long time. The detection efficiency of circRNAs microarray is much higher than RNA-seq because it contains probes that target the head-to-tail connection (51). But it produces data that is difficult to compare between studies. Therefore, there is an urgent need to develop simple, effective, and sensitive new methods to study circRNAs.

Recently, a newly reported exome capture RNA sequencing technology can detect and characterize circRNAs in more than 2000 cancer samples (64). In addition, Zhang et al. developed a new method for quantitative detection of circular RNA with high sensitivity and specificity (56). The two cleverly designed DNA probes can be precisely connected by using ligase at the connection site of circular RNA. Distinguish circular RNA from corresponding linear RNA. Liu et al. proposed a reverse transcription-rolling cycle amplification (RT-RCA) process that can selectively amplify target circular RNA (57). Zhang et al. designed a pair of Stem-loop primers (SLPs) based on loop-mediated isothermal amplification (LAMP), an excellent nucleic acid amplification method, that could accurately recognize circRNA junction sequences, thereby establishing an SLP-induced dual amplification system (58).The exponential amplification method makes circRNA detection simple and accurate. Additionally, Li et al. designed an electrochemical method for the recognition and capture of circRNAs with hairpin probes, avoiding additional RNase R treatment, and the method exhibited good sensitivity and selectivity (62).

According to the characteristics of circRNAs, some detection methods for circRNAs have been developed, including traditional detection methods and newly developed methods, which have their own advantages and disadvantages. However, the way forward is clear, and the detection method of circRNAs still needs to be developed or improved, so that its sensitivity, specificity, rapidity, and convenience can be applied to biomedical research and clinical detection. All these make it possible for circRNAs to become diagnostic tools and therapeutic targets.

In recent years, with the intensive study of circRNAs, researchers have developed many high-quality online databases. This section introduces some databases that can be used for circRNAs research in Table 2 . These artificially established databases are of great significance to the study of the biological functions of circRNAs. In addition to the online database listed in the table, in 2022, the University of Padova developed a circRNA function prediction software CRAFT, which can predict circRNA sequences and molecular interactions with miRNAs and RBPs, as well as their coding potential (81). Nevertheless, the existing circRNAs collection is largely limited to certain well-studied species, such as humans and mice. In addition, the current annotations and naming are rather incomplete. Most databases only use one or two resources for annotations. Searching for the same circRNAs has different naming methods in different databases, which increases the difficulty of studying circRNAs.

Lung cancer is the malignant tumor with the highest mortality rate in the world (199). Circ0003222 sponges miR-527 to down-regulate the expression of PHF21B and its downstream β-catenin. Thereby promoting the proliferation, migration, and invasion of tumor cells. Yu et al. (91) discovered a new type of circHMGA2 (hsa_circ_0027446) molecule through microarray, which is highly expressed in lung adenocarcinoma (LUAD). Mechanically, circHMGA2 promotes LUAD cell metastasis through the miR-1236-3p/ZEB1 axis. Yao et al. (99) found that circ_0018414 was down-regulated in LUAD tissues and cells, and inhibited the progression of LUAD through the Wnt/β-catenin pathway of miR-6807-3p/DKK1 axis inactivation. Some circRNAs have been described as binding to proteins in lung cancer. For example, circNDUFB2 (101), which is down-regulated in non-small cell lung cancer, acts as a scaffold to enhance the interaction between TRIM25 and IGF2BPs. It inhibits the growth and metastasis of NSCLC cells by regulating protein ubiquitination and degradation and cellular immune responses. In addition, Huang et al. (103)found that circXPO1 is highly expressed in LUAD through RNA sequencing. In terms of mechanism, circXPO1 can bind to IGF2BP1 to enhance the stability of CTNNB1 mRNA, thereby promoting the progress of LUAD.

The development of new NSCLC-specific biomarkers to aid in diagnosis and clinical decision-making has always been a pressing concern. Li et al. found that circ0003222 is related to the staging, metastasis, and survival rate of patients with non-small cell lung cancer (NSCLC) (6). Additionally, high expression of circ 0070354 was demonstrated to be substantially linked to advanced TNM staging and poor differentiation in NSCLC and was an independent predictor of poor prognosis. CEA, SCC, and Cyfra21-1 are the acronyms for CEA, SCC, and Cyfra21-1, respectively. The AUC of circ0070354, when combined with the other three mature tumor markers, was 0.730, which was much higher than the solitary diagnosis (200). According to the findings, some circRNAs potentially outperform traditional tumor markers in terms of diagnosis, and the combined diagnosis has higher sensitivity and specificity for lung cancer diagnosis and treatment.

Colorectal cancer (CRC) is the second leading cause of death from cancer worldwide (199). Jian et al. (110) tested the gene expression in 42 pairs of colorectal cancer tissues and normal tissues adjacent to cancer. The results showed that circ001680 was overexpressed in 71.4% of colorectal cancer tissues. In terms of mechanism, circ001680 promotes the proliferation and migration of colorectal cancer cells by targeting miR-340. Yang et al. discovered a new circRNA, circPTK2, and found that circPTK2 binds to the Ser38, Ser55, and Ser82 sites of vimentin protein to promote EMT of CRC cells in vivo and in vitro ( 112). In addition to interacting with proteins to regulate the expression of target genes, circRNAs encoding proteins or peptides have also been found in colorectal cancer, such as circ0006401 (116), circPLCE1 (117), and circFNDC3B (118). Among them, circ000641 encoding peptide fragment promotes the proliferation and migration of CRC and promotes the stability of the host gene col6a3 mRNA, and thus promotes the proliferation and metastasis of CRC. The circPLCE1-411 protein encoded by circPLCE1 combined with the HSP90α/RPS3 complex plays a key role in the NF-κB activation of CRC and ultimately inhibits tumor proliferation and metastasis in CRC cells (117). The tumor suppressor circFNDC3B is mainly located in the cytoplasm and encodes a new protein circFNDC3B-218aa, thereby inhibiting the proliferation, invasion, and migration of colon cancer cells (118).

Wang et al. (108) found that circSPARC is highly expressed in the tissues and plasma of CRC patients, is associated with advanced TNM staging, lymph node metastasis, and a low survival rate of CRC. Mechanistically, circSPARC can upregulate the expression of JAK2 by sponge miR-485-3p, and ultimately promote the accumulation of phosphorylated p-STAT3, thereby promoting the proliferation and migration of cancer cells. The most commonly used colorectal tumor marker CEA has limited sensitivity in early CRC (201, 202). While circRNAs can be employed as reliable biomarker complements for CEA in CRC early diagnosis and treatment monitoring. According to the ROC curve analysis of Hu et al., the AUC (0.831 vs 0.657), sensitivity (0.677 vs 0.532), and specificity (0.915 vs 0.675) values of circ 001659 in the early diagnosis of CRC were higher than those of CEA as a novel biomarker of successful treatment and response for cancer tracking thing (203). These findings indicate that circRNAs can become potential diagnostic and prognostic biomarkers and therapeutic targets for the treatment of CRC.

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related deaths. There have been multiple reports that a variety of circRNAs inhibit or promote tumor progression in liver cancer. Hu et al. (120) found that circASAP1 promotes the proliferation and invasion of liver cancer cells by regulating the miR-326/miR-532-5p-MAPK1 signaling pathway, and then mediates tumor-associated macrophages by regulating the miR-326/miR-532-5p-CSF-1 pathway Cell infiltration. The circRNAs array analyzes the expression of circRNAs in tumor tissues and normal tissues. In a study by Dong et al. (125), it was found that 28 up-regulated and 18 down-regulated circRNAs were found in liver cancer tissues. circMEMO1 is significantly down-regulated in HCC samples and can act as a sponge of miR-106b-5p to regulate TCF21 promoter methylation and gene expression, thereby regulating HCC progression. Li et al. (128)found that circMRPS35 was highly expressed in 35 pairs of HCC patients compared with normal tissues. It is worth noting that circMRPS35 can not only adsorb miR-148a-3p, regulate the expression of Syntaxin 3 (STX3), thereby regulating the ubiquitination and degradation of phosphatase and tensin homolog (PTEN) but can also encode a peptide (circMRPS35-168aa), this peptide promotes cisplatin resistance in HCC cells. CircLRIG3 is significantly up-regulated in HCC, forming a ternary complex with EZH2 and STAT3, promoting EZH2-induced STAT3 methylation and subsequent phosphorylation, leading to the activation of STAT3 signal, thereby promoting the proliferation, migration, and invasion of liver cancer cells. Reduce cell apoptosis (129).

Wei et al. reported that the expression of circCDYL or the combined expression of HDGF and HIF1AN are independent markers for distinguishing early HCC, providing the possibility for the detection and early treatment of liver cancer (204). Yang et al. found that circFN1 promotes sorafenib resistance by regulating the miR-1205/E2F1 signaling pathway, which is a potentially valuable target for HCC resistance (205).

Gastric cancer (GC) is the fourth leading cause of death in the world (199), especially in Asian countries, the incidence of gastric cancer is increasing year by year. Cao et al. found that circ0008259 (circLMO7) is highly expressed in GC tissues, circLMO7 sponge miR-30a-3p regulates the WNT2/β-Catenin pathway and affects the glutamine metabolism of GC cells, and ultimately promotes the growth and migration of GC (131). Peng et al. found that the level of circCUL2 in GC tissues and cells was significantly reduced, and the sponge engulfed miR-142-3p to regulate ROCK2, thereby inhibiting malignant transformation and inhibiting tumorigenicity in vivo ( 140). In addition, Yan et al. (206) found that circEVI5 was significantly down-regulated in GC tissues and cells. circEVI5 sponges swallowed miR-4793-3p and increased the expression level of FOXO1 to inhibit the proliferation of GC and delay the cell cycle. Wang et al. (148)analyzed the circRNAs of five pairs of human stomachs and corresponding non-tumor adjacent specimens and found that circURI1 was significantly highly expressed in GC and metastasized in GC. It regulates a small part of genes involved in cell movement by isolating hnRNPM protein to inhibit GC metastasis. Zhang et al. (149) found that circDIDO1 is down-regulated in gastric cancer tissues, and its low level is associated with larger tumors, distant metastasis, and poor prognosis. In mechanism, circDIDO1 encodes a new 529aa protein, which interacts with poly ADP-ribose polymerase 1 (PARP1). Effect and inhibit its activity. Interestingly, circDIDO1 also binds to peroxide reduction protein 2 (PRDX2), which promotes the ubiquitination and degradation of PRDX2 mediated by rbx1, leading to inactivation of its downstream signaling pathways.

Further, numerous research has explored the clinical utility of circRNAs as biomarkers for the early detection and prognosis of gastric cancer. For instance, Song et al. (207) detected the expression profile of circRNAs and found that hsa_circ_000780 was significantly downregulated in GC tissue samples, and its level was correlated with the level of tumor clinicopathological features. Interestingly, circ000780 was also found in gastric juice of patients with early GC. In another report, circERBB2 (208) in plasma can be used as a prognostic biomarker for gastric cancer patients. CircERBB2 levels in preoperative plasma (high group) were significantly correlated with lymph node metastasis (P = 0.035), suggesting that it could be used to predict noninvasively the prognosis of GC.

Breast cancer (BC) is the main cause of cancer in women and the main cause of death in women. Wang et al. (151) used the circRNAs microarray data set and found that four circRNAs were abnormally expressed in TNBC. Among them, circBACH2 is most significantly elevated in BC tissues, and its high expression promotes epithelial-mesenchymal transition and cell proliferation and is positively correlated with the malignant progression of BC patients. Mechanistically, circBACH2 sponges miR-186-5p and miR-548c-3p, thereby releasing the expression of C-X-C chemokine receptor type 4 (CXCR4). Li et al. (160) discovered that circ-EIF6 encodes a new peptide called EIF6-224 amino acid (aa). EIF6-224aa directly interacts with the oncogene MYH9 in BC and inhibits the ubiquitin-proteasome pathway and subsequently activates the Wnt/β-catenin pathway to reduce the degradation of MYH9, thereby playing a carcinogenic effect. In addition, Wang et al. reported that circACTN4 can competitively bind to far upstream element-binding protein 1 (FUBP1) to prevent FUBP1 from binding to FIR, thereby activating MYC transcription and promoting tumor progression in breast cancer (158).

Some potential circRNAs biomarkers for early diagnosis of BC and prediction of recurrence and metastasis have emerged from the detection of clinical samples in the tissues and peripheral blood of BC patients and healthy controls, combined with the correlation analysis of clinicopathological factors and the analysis of prognosis and survival. For example, CircSMARCA5 can form an R-loop with its parental locus, causing a transcriptional pause at SMARCA5 exon 15, and SMARCA5 DNA is involved in chromatin remodeling in damaged regions.circSMARCA5 may serve as a therapeutic target for patients with drug-resistant BC (209). The above studies have provided new insights into the role of circRNAs in BC.

According to recent research findings, the involvement of circRNAs in hematological malignancies is becoming more widely recognized (210). Among them, aberrant circRNAs expression might upset the balance between self-renewal and differentiation of hematopoietic stem cells (HSCs), resulting in bone marrow failure or hematological malignancies.

Acute myeloid leukemia (AML), the most common leukemia in adults, is uncommon in children, but has a poor prognosis and is prone to relapse (211). Liu et al. discovered that circRNF220 is abundantly and precisely expressed in children’s peripheral blood and bone marrow using microarray technology. CircRNF220 knockout can reduce the proliferation of AML cell lines and primary cells while also promoting cell death (162). FLT3-ITD+ AML is a significant subtype of AML, and Zhang et al. observed that the effect of circ0000370 on the development of FLT3-ITD-positive AML may be directly connected to miR-1299 and S100A7A (164). Acute lymphoblastic leukemia (ALL) is the most prevalent malignant tumor in children, and circ0000094 has been demonstrated to be a molecular sponge of miR-223-3p, which can upregulate the expression of FBW7 by limiting the expression of miR-223-3p, hence preventing ALL progression (165). Zhu et al. found that circADD2, as a tumor suppressor gene in ALL, inhibited cell proliferation and promoted cell apoptosis both in vitro and in vivo. Mechanistically, circADD2, which can sponge miR-149-5p, may serve as a potential biomarker for ALL diagnosis or treatment (166). Interestingly, circRNAs expression profiles can also clearly distinguish Acute leukemia (AL). For example, Guo et al. reported circ0001857 and circ0012152 ALL and AML (212). The recently discovered Circ0009910 can regulate ULK1-induces autophagy by targeting miR-34a-5p and accelerating the resistance of CML cells to imatinib (167). High expression of circ-RPL15 was positively correlated with IGHV mutation status, which is crucial for evaluating CLL prognosis. MiR-146b-3p-mediated RAS/RAF1/MEK/ERK pathway inhibition could be alleviated by circRPL15 overexpression in CLL. CircRPL15 may represent a promising novel plasma biomarker for the diagnosis of CLL (168). Mei et al. found that the relative expression of circADARB1 was significantly increased in the plasma of Natural killer/T-cell lymphoma (NKTCL), which binds to miR-214-3p and regulates p-Stat3, promotes the proliferation of NKTCL cells, and inhibits apoptosis (169). Furthermore, Zhao et al. reported that CircEAF2 inhibited Epstein-Barr infection positive diffuse large B cell expansion and advanced apoptosis via the miR-BART19-3p/APC/-catenin axis (170).

Some of the circRNAs reported above play a regulatory role in the occurrence and development of hematological tumors through various molecular mechanisms, suggesting some potentials of circRNAs in the research of hematological malignancies in the future. These findings also aid in the diagnosis and prediction of hematological malignancies.

The abnormal expression of many circRNAs has been verified in many cancers. In renal cell carcinoma (RCC), Cen et al. found that circSDHC competitively binds to miR-127-3p, preventing it from inhibiting the downstream genes CDKN3 and E2F1 pathways, leading to RCC Malignant progress (172). NONO-TFE3 TRCC (Xp11.2 translocation/NONO-TFE3 fusion renal cell carcinoma) is a subgroup of renal cell carcinoma. Yang et al. found that highly expressed circMET accelerates the decay of CDKN2A mRNA by recruiting YTHDF2, while competitively binding miR- 1197, Regulates SMAD3 expression (177). In bladder cancer (BC), Yang et al. (179) used high-throughput sequencing and RT-qPCR to verify the abnormally high expression of circUBE2K BC tissue. As a ceRNA, the expression of ARHGAP5 was regulated by sponge miR-516b-5p to promote tumor development. The down-regulated circZKSCAN1 in BC tissues and cell lines up-regulates the expression of p21 through sponge miR-1178-3p, which inhibits the proliferation, migration, and invasion of bladder cancer (185). In addition, some circRNAs have also been found to play an important role in prostate cancer and cervical cancer, as shown in Table 3 .

In short, various studies have shown that circRNAs are involved in the occurrence and development of various cancers. However, the role of circRNAs in the diagnosis and treatment of cancers needs to be further studied.