Circular RNAs (circRNAs) are covalently closed circular non-coding RNAs (ncRNAs) without a 5' cap and a 3' poly(A) tail (1). For decades, circRNAs were considered to be non-functional byproducts of mis-splicing (2,3). In recent years, circRNAs have been discovered in eukaryotes (4-8). With the maturity of sequencing technologies and algorithms, thousands of circRNAs have been identified, and experiments have confirmed that such RNAs are no longer ‘splicing noise’, but functional molecules (7,9-12). CircRNAs can share the microRNA (miRNA/miR) response elements targeted by mRNA, thereby regulating the expression of mRNA (13).

In addition, circRNAs have other biological functions, such as regulating gene transcription and translation, and binding to RNA-binding proteins (9). CircRNAs can regulate transcriptional and post-transcriptional gene expression in various diseases such as lung cancer and gastric carcinoma (14). These functional circRNAs are of significance for the maintenance of normal cell functions and the occurrence and development of abnormal biological functions. In addition to participating in the regulation of epigenetic, transcriptional or post-transcriptional biological processes of various cells, circRNAs also play important roles in the signal pathways of cellular processes (15,16).

Preeclampsia (PE) is defined as new-onset hypertension after 20 weeks of gestation, accompanied by proteinuria, headache, dizziness, nausea, vomiting and epigastric discomfort (17). PE is a serious obstetric emergency worldwide, with an annual incidence of 3-8%, and is a major cause of increased maternal and neonatal morbidity and mortality (18). Therefore, understanding the pathogenesis of PE remains imperative for obstetricians. A growing body of evidence supports that the pathogenesis of PE is multifactorial, including insufficient invasive ability of trophoblasts (19), failure of spiral artery remodeling (18), abnormal immune responses (18), inflammatory responses (20) and genetic factors (21). These pathogenic mechanisms can be regulated by epigenetics (21). As a type of ncRNA, circRNAs are widely involved in gene expression, protein/RNA splicing or modification or protein-coding process (22). Abnormally expressed ncRNAs associated with PE have been identified by genome-wide analysis of placental-derived circRNA (23). Studies have confirmed that circRNA plays an important role in regulating the development and function of the placenta and the pathogenesis of PE (24,25). The abnormal placental transcriptome of PE is affected by epigenetic regulation. However, the regulation of differentially expressed genes and transcripts on the occurrence and development of PE has not been fully clarified (26). The present review summarized studies on the role of circRNAs in the pathogenesis of PE.

At present, in the field of PE, the study of circRNAs is mainly limited to diagnostic markers and pathogenesis. Risk factors for PE include a history of PE in a previous pregnancy, chronic kidney disease, hypertension, diabetes, autoimmune diseases such as systemic lupus erythematosus, initial onset PE, age >40 years, inter-pregnancy interval over 10 years, a body mass index >35 kg/m², polycystic ovary syndrome and multiple pregnancies (20). However, only 30% of individuals predisposed to PE can be detected based on these risk factors (27). Due to the heterogeneous presentation of PE, potential biomarkers are required for its early detection.

Shao et al (28) studied the expression of circRNA in the blood of 82 pregnant individuals at 8-20 weeks of gestation and revealed that the blood concentration of circ_101222 in patients with PE was significantly higher compared with that in healthy pregnant individuals. Other studies have analyzed the expression of several mRNAs, ncRNAs and circRNAs in the plasma or placentas of individuals with PE and healthy pregnant individuals to identify potential predictive markers of PE. The expression pattern of current potential circRNAs that may serve as diagnostic markers for PE is summarized in Table I.

Among all detected markers, hsa_circ_0036877 is recognized as a potential plasma biomarker for PE (29). Two biomarkers, hsa_circ_0004904 and hsa_circ_0001855, are involved in the pathogenesis of PE by activating miRNA sponges that directly target pregnancy-associated plasma protein A (PAPP-A). This indicates that PAPP-A is present in the plasma of individuals with PE (30). However, the limitations of circRNAs as diagnostic markers must be considered. Pregnancy is a process, and the level of molecular expression in the placenta dynamically changes. More detailed grouping and long-term studies are needed to determine the time point for screening circRNA as a diagnostic marker.

PE is a hypertensive disorder (31,32). Unfortunately, the etiology and pathogenesis of PE are still far from clear. However, accumulating evidence confirms that impaired spiral artery remodeling, placental dysfunction and insufficient trophoblast invasion may play critical roles in the development and progression of PE (33-35). Furthermore, extensive or shallow invasion of extravillous trophoblasts (EVTs) at the maternal-fetal interface has been identified as a major cause of placental failure, ultimately leading to PE (36,37). Restricted migratory activity of EVTs in the maternal decidua has been shown to impede trophoblast function, leading to PE (38). Some studies have investigated the pathogenesis of PE from the perspective of placenta (39,40). A large number of recent studies have shown that a variety of ncRNAs are associated with the pathogenesis of pregnancy disorders (41-45).

From the perspective of pathogenesis, abnormal placental development is one of the main causes of PE. Zhou et al (46) have revealed that circRNA_3286, circRNA_5593 and circRNA_3800 are downregulated in placental tissues of individuals with PE compared with healthy pregnant individuals. Melchiorre et al (39) have investigated the distribution of circRNAs in the placental tissues of individuals with PE and explored the potential impact of circRNA dysregulation on the progression of PE. A total of 300 circRNAs that are differentially expressed between individuals with PE and healthy pregnant individuals were identified. Reverse transcription-quantitative PCR results showed that hg38_circ_0014736 and hsa_circ_0015382 are highly upregulated, and hsa_circ_0007121 is downregulated in all patients with PE. The data showed that these three circRNAs are significantly associated with the regulation of transcription, proliferation, hypoxia response and protein binding (47). The studies on the role of circRNAs in the pathogenesis of PE are summarized in Table II, which is helpful to explore novel strategies for the treatment of PE.

In recent decades, researchers have confirmed that circRNAs are involved in a variety of diseases (48). However, to the best of our knowledge, there have been few studies on the role of circRNAs in the pathogenesis of PE (49-51). The different mechanisms are described in detail below.

Although the functions of most circRNAs remain unclear, only a small fraction of identified circRNAs have been studied for their biological significance (81-85). A study has shown that circRNAs have binding sites for microRNAs and RNAs, and can act as RNA sponges to regulate the expression levels of target genes (86). The most representative circRNA is ciRS-7, which contains >70 conserved binding sites for miR-7 (87-89). A study has shown that circCDR1as and circMTO1 bind to miR-7 and miR9, respectively, and affect gene regulation, thereby indirectly suppressing or stimulating tumors (90). In addition, subsequent studies have also demonstrated the presence and importance of ciRS-7 as a miR-7 sponge in a number of pathophysiological processes, such as insulin secretion, myocardial infarction, hepatocellular carcinoma and gastric cancer progression (88,91-93). Based on the aforementioned theory, artificial sponge technology is a method of manufacturing molecules that can specifically bind to target miRNAs so that they can specifically adsorb target miRNAs. According to the partial base sequence of the target miRNA, circRNAs are artificially processed, then packaged with plasmids and transfected into cells or tissues, while circRNA acts as a miRNA ‘sponge’ to adsorb a large number of target miRNAs (64). Fan et al (94) found that the expression of circNR3C2 significantly enhances the tumor suppressive effect of HRD1 by sponging miR-513a-3p.

CircRNAs can also regulate biological processes by binding to proteins such as transcription factors. After binding to peccadillo homolog 1, circANRIL affects exonuclease-mediated pre-ribosomal (r)RNA processing and ribosome biogenesis (95). Circ-Foxo3 binds to CDK2 and cyclin-dependent kinase inhibitor 1 to form a ternary complex, thereby inhibiting the function of CDK2 and blocking cell cycle progression (10,96). Circ-Foxo3 also has a high binding affinity to anti-aging inhibitor of DNA binding 1, transcription factor E2F1 and anti-stress proteins FAK and HIF1a, and retains them in the cytoplasm, leading to increased cellular senescence (10). Circ-poly(A)-binding protein nuclear 1 (PABPN1) binds to HuR, thus preventing HuR from binding to PABPN1 mRNA to reduce PABPN1 translation (61,97). However, not all circRNAs that interact with proteins inhibit protein function. Ectopic circ-Amotl1 interacts with and stabilizes the nuclear oncogene c-myc, thereby upregulating c-myc targets and promoting tumorigenesis (98,99).

An early study by Chen and Sarnow in 1995(100) demonstrated that synthetic circRNAs can recruit the 40S ribosomal subunit and initiate the translation of detectable peptides in human cells through internal entry sites. Studies have shown that, if an internal ribosome entry site (IRES) is inserted upstream of the start codon, whether in vivo or in vitro, circRNAs that are similar to certain RNAs without a 5' cap structure and a 3' (polyA) tail structure can be translated into proteins (101,102). Although IRES-mediated translation was first discovered in RNA and DNA viruses, it has subsequently been discovered in mRNAs such as immunoglobulin heavy chain binding protein mRNA, fibroblast growth factor and VEGF mRNA (103-105).

At present, several translated circRNAs have been identified to play key roles in human diseases, especially cancer. In gliomas, circSHPRH, produced by the SNF2 histone linker PHD RING helicase (SHPRH) gene, encodes a novel 146-amino acid protein (SHPRH-146aa), which exhibits inhibitory activity during tumorigenesis and glioma activity and also serves as a biomarker (106-108). Another glioma study revealed that circLINC-PINT is derived from a long intergenic non-coding RNA p53-induced transcript (LINC-PINT), which encodes an 87-amino acid peptide (PINT87aa). It interacts with the PAF1 complex in the nucleus to inhibit the transcriptional elongation of multiple oncogenes, thus playing a tumor suppressor role in the control of cell proliferation and tumorigenesis (106).

There are also studies showing that circ-F-box and WD-repeat domain containing 7 (FBXW7) can inhibit the development of glioma (109,110). CircFBXW7 is produced by the tumor suppressor E3 ligase FBXW7, which encodes a 185 amino acid peptide (FBXW7-185aa), that plays the role of a tumor suppressor in glioma (111). Another recent study showed that FBXW7-185aa can inhibit the proliferation and migration of triple-negative breast cancer cells by increasing FBXW7 abundance and inducing c-Myc degradation (112).

In hepatocellular carcinoma, GSK3β-induced phosphorylation and degradation of β-catenin lead to activation of the Wnt pathway, which is associated with poor hepatocellular carcinogenesis and prognosis (113,114). In circBase, circβ-catenin is the only isoform that can be expressed in hepatocellular carcinoma (115). Circβ-catenin produces a 370 amino acid peptide (β-catenin-370aa), and can promote the growth of liver cancer cells by activating the Wnt pathway (116,117).

Other studies have confirmed that circRNAs can activate the encoded protein through the m6a mechanism (118,119). The modification of m6A is completed by methyltransferase complexes such as methyltransferase-like (METTL)-3, METTL-14, Wilms Tumor 1-associated protein, RNA-binding motif protein 15 and zinc finger CCCH domain-containing protein 13 (4,120-124). Various internal or external factors, such as cell type, developmental stage, nutrient supply, circadian rhythm and environmental stresses initiate m6a translation (125). The 5'UTR m6A residue can directly recruit eukaryotic initiation factor 3, which is sufficient to recruit the 43S pre-initiation complex and bypass the m7G capping requirement to initiate translation, thus enabling translation initiation in the absence of the cap-binding factor eIF4E model (126).

Unlike conventional forward splicing, circRNAs originate from the same precursor as linear RNA transcripts, which is formed by a process called back splicing. Back-splicing creates a covalently closed loop that is characterized by a non-linear back-splicing junction between the splice donor and upstream splice acceptor, and it lacks a 5'cap and a 3' poly(A) tail (127). Due to this structural feature, circRNA can resist digestion by nucleic acid ribozymes (such as RNase R) and is more difficult to be degraded by exonuclease, so it is more stable compared with linear RNAs, with a longer half-life of up to 10 times that of linear RNA (128). These attributes make circRNA a potential biomarker for disease diagnosis and prognosis (85). CircRNA is stable and not easy to be degraded in blood and exosomes. It can be quantitatively detected by reverse transcription followed by qPCR (85). At the same time, compared with the complex antigen-antibody reaction and unclear parameters of protein detection, circRNAs are always expressed in a tissue- or cell-specific manner and can be detected by qPCR and in situ hybridization, which makes circRNA an ideal molecule for clinical diagnosis or prognosis detection of diseases, with landmark significance (1,7,129-131).

CircRNAs have been implicated in a variety of diseases, such as bone-osteosarcoma, colon-colorectal adenocarcinoma, kidney-renal cell carcinoma, liver-hepatocellular carcinoma, lung-lung adenocarcinoma and stomach-gastric adenocarcinoma (132). Due to the observed association between circRNA abundance and cancer, circRNA may serve as a cancer biomarker with good diagnostic performance (133). A study has also shown that circRNAs are present in human body fluids such as saliva, plasma, plasma and exosomes at relatively high steady-state levels, making them candidate biomarkers for non-invasive liquid biopsies (127). Zhang et al (79) found that circSATB2 is highly expressed in non-small cell lung cancer cells and tissues. CircSATB2 is highly expressed in plasma exosomes of patients with lung cancer with high sensitivity and specificity for clinical detection, and is associated with lung cancer metastasis (134). Wang et al (134) found that circRNA-002178 is detectable in the plasma exosomes of patients with lung adenocarcinoma (LUAD) and can be used as a biomarker for early diagnosis of LUAD. These studies provide a certain basis for the use of circRNAs as molecular markers for disease diagnosis and provide a new method for clinical screening of diseases.

PE is defined as new-onset hypertension after 20 weeks of gestation, so early diagnosis is crucial for PE. Due to the current lack of sufficient data or the heterogeneity of the recruited population, circRNA is not sufficient as a marker for PE monitoring and screening. Given the molecular advantages of circRNAs over linear RNAs, studies on circRNAs are more focused on possible screening purposes. Although some studies have reported the possible screening performance of circRNAs in the first or second trimester of pregnancy, a single circRNA has not been successfully used in any PE screening program (135-137). One study has found that the area under the curve (AUC) of plasma hsa_circ_0001855 is 0.62. While using the plasma protein PAPP-A in combination with hsa_circ_0001855 and hsa_circ_0004904, the AUC increases to 0.94, with a sensitivity of 0.87 and a specificity of 0.97(30). In another study combining plasma hsa_circ_0007885 level, plasma sFLT1 level and abnormal uterine artery pulsatility index (UtA-PI), the AUC is 0.85, and the sensitivity and specificity are 0.80 and 0.86, respectively (138). The predictive power for PE is far stronger compared with any previous single molecular or ultrasound data (61). In summary, circRNAs can be combined with some specific molecules or clinical examination data such as PPAP-As, sFlt-1 and UtA-PI for prediction as a new strategy for early clinical diagnosis of PE.

The present review summarizes the study progress of circRNAs in PE in recent years. The endogenous competitive mechanism of circRNAs occupies the majority, which is of help for understanding the pathogenesis of PE. However, the research on circRNAs has only revealed the tip of the iceberg, such as RNA-binding proteins and encoded proteins, which have potential for the prevention and treatment of PE. With the development of sequencing technology, more circRNAs will be discovered and new methods will be used to study PE. We hope that this review has provided help for the diagnosis and treatment of PE.