Metastasis is one of the ten essential characteristics of malignant tumors and a hot research topic (Hanahan and Weinberg, 2011). It is a process that enables malignant cells to escape from the primary tumor site, migrate through the lymphatic and/or blood circulation, and ultimately spread to remote sites (Lazebnik, 2010; Tarin, 2011; Sun et al., 2015). The onset of tumor metastases is often indicative of poor prognosis, and >90% of the cancer-related deaths result from metastases (Chambers et al., 2002; Maishi and Hida, 2017; Jiang et al., 2019; Mattiuzzi and Lippi, 2020; Wang et al., 2020d). Most metastatic lesions cannot be surgically eradicated because such lesion is often indicative of more widespread systemic disease (Gupta and Massagué, 2006). Although an increasing number or tumor treatments are being developed with advances in medical technology, tumor metastasis remains one of the major causes of the extremely high mortality rate in a variety cancers. Therefore, the search for tumor markers and therapeutic targets remains an important strategy for improved cancer treatment (Sun L. et al., 2019; Stoletov et al., 2020; Wang J. et al., 2020).

Exosomes are extracellular vesicles that originate from the multivesicular bodies (MVBs) and are present in intercellular space or circulate in biological fluids (Kalluri, 2016; Zhang and Yu, 2019; He et al., 2021a). These vesicles can be internalized by neighboring cells or by remote receptor cells through fusion with the target cell membrane, thereby altering the behavior of the target cell (Hessvik and Llorente, 2018; Gonda et al., 2019). Exosomes play integral roles in mediating intercellular communication, regulating immune system function (Gao et al., 2018; Yu et al., 2018), promoting cell development and differentiation (Zhou et al., 2021), influencing viral replication, and other physiological or pathological disease processes (Alenquer and Amorim, 2015) that affect the progression of many diseases, including tumors (Zhang and Yu, 2019). Recent studies have revealed that exosomes are involved in regulating several malignant biological behaviors of tumors by transporting various growth factors, proteins, lipids, nucleic acids, non-coding RNAs, and other molecules (Kalra et al., 2016), including promoting malignant proliferation, metastasis, and immune escape by tumor cells (Wortzel et al., 2019; Kugeratski and Kalluri, 2021), and contributing to tumor microenvironment (TME) (Meng et al., 2019; Wu et al., 2019). The roles played by non-coding RNAs (ncRNA)—a class of molecules present in high concentrations in exosomes, i.e., microRNAs (miRNAs), long non-coding RNAs (lncRNAs), and circular RNAs (circRNAs)—in tumor metastasis are gradually gaining attention (Zhao et al., 2015; Chen et al., 2020a; Guo et al., 2020). CircRNAs are covalently closed ncRNA molecules, comprising 3′ and 5′ ends joined in a non-collinear manner by reverse splicing (Zhang et al., 2016), which exhibit properties and functions different from those of linear RNA. Many studies have demonstrated that intracellular circRNAs can regulate tumor metastasis in multiple ways. However, the effect of exosomal circRNAs on tumor metastasis cannot be fully explained. Herein, we performed a systematic literature review of exosomal circRNAs in the context of development and progression of tumor metastases.

Metastasis of tumor cells is a multi-step process (He et al., 2021b) that includes the following steps: (I)tumor cells lose adhesion to neighboring cells and detach from the primary tumor, a phenomenon that results in the degradation of the extracellular matrix (ECM), and the migration and invasion of the cells into the surrounding tissues (Friedl and Wolf, 2003); (II)infiltration of the tumor cells into the bloodstream, followed by adaptation and escape from anoikis to survive in the circulation (Zhan et al., 2004), (III) exudation of the tumor cells outside the blood vessels (IV), and finally entry of the tumor cells into the metastatic site, followed by adaptation and growth, resulting in the eventual colonization of the site (Fidler, 2003; Nguyen et al., 2009; Scully et al., 2012) (Supplementary Figure S1). Epithelial-to-mesenchymal transition (EMT) is an important form of tumor metastasis. EMT is a complex cellular pathway in which epithelial cells lose intercellular adhesion (characterized by loss of membrane E-Cad) and gain mesenchymal features (characterized by increased N-cadherin expression and migration capacity) (Bakir et al., 2020). Stephen Paget first proposed the classic hypothesis of “seed and soil” in 1989. He compared the primary tumor to a “plant”, the tumor cells to “seeds”, and the host environment to “soil” (Paget, 1989), and boldly hypothesized, “When a plant has seeds, its seeds can be taken anywhere; but they can only survive and grow if they fall on suitable soil”. This assumption is accepted as the fundamental theory for explaining tumors and metastasis (Langley and Fidler, 2011; Ribelles et al., 2014). In this doctrine, it is believed that the autonomous mechanisms of tumor cells are insufficient to accomplish metastasis and that tumor metastasis is regulated by other factors, including tumor microenvironment (TME) (Quail and Joyce, 2013; McAllister and Weinberg, 2014). Interactions between the TME and tumor tissue is gradually gaining attention as a new field (Quail and Joyce, 2013; Liao et al., 2021; Xiao and Yu, 2021). TME refers to the cellular environment in which the tumor exists, and its composition includes tumor cells as well as surrounding blood vessels, ECM, signaling molecules, and non-malignant cells such as fibroblasts and immune cells (Luo et al., 2016). Several studies have shown that the TME has an inhibitory effect on the growth of malignant tumors (Holmgren et al., 1995; Suzuki et al., 2006). However, in most malignant tumors, these inhibitions are overcome, resulting in the use of support cells by malignant tumors to increase their metastatic potential and promote their own growth and relocation to remote sites (Marx, 2013; Massagué and Obenauf, 2016). Tumor cells in highly aggressive primary tumors are more adept at exploiting this particular tissue microenvironment. Moreover, stromal cells and fibroblasts can also secrete growth factors such as hepatocyte growth factor (HGF), chemokines (e.g., CXCL12), and exosomes, which can promote the forming of pre-metastatic niche (PMN) (Filipazzi et al., 2012; Liu and Cao, 2016; Whiteside, 2016). These growth factors not only directly promote the growth and survival of malignant cells, but also act as decoys to stimulate other cells to migrate to the TME and indirectly promote tumor invasion and metastasis (Spaeth et al., 2008; Hanahan and Coussens, 2012). Metastasis is “a long journey” for the tumor cells themselves, as there are many rate-limiting steps in the formation of metastatic cancer, including extravasation, distal organ survival, and the establishment of sustained growth (Psaila and Lyden, 2009; Acharyya and Massague, 2016). Microenvironmental cues play important roles in all steps of metastasis. Thus, successful metastasis depends on the ability of cancer cells to adapt to different microenvironments at each step of the metastatic cascade (primary tumor, body circulation and final metastatic site) (Zhuang et al., 2019).

Exosomes are lipid bilayer nano-vesicles with a “spherical” morphology (30–150 nm) that are thought to be released by almost all cell types (Théry et al., 2002). These vesicles display a number of surface molecular markers, such as CD9, CD63, and CD81 (Mathivanan et al., 2010; Kowal et al., 2016). Exosome formation comprises four stages, i.e., initiation, endocytosis, formation of MVBs, and exosome secretion. In this process, the endosomal sorting complex required for transport (ESCRT) can select ubiquitin-tagged proteins, lead them to MVBs, and separate and release them from the peripheral membrane by a process similar to cytoplasmic division and viral outgrowth (Hanson and Cashikar, 2012; van Niel et al., 2018; Xu et al., 2018). Exosomes are widely available and have become essential mediators of intercellular communication in physiological and pathological states (Meldolesi, 2018; Delpech et al., 2019; Asghar et al., 2020). As a signal vector for intercellular communication, the content of exosomes varies with different physiological and pathological conditions and primary cell types (Bebelman et al., 2018). Many studies have reported that exosomes regulate tumor progression by carrying or delivering multiple biomodulator “cargoes” (including ncRNAs and proteins) (He et al., 2019; Li R. et al., 2019). Exosomes are key contributors to a wide range of biological processes during tumor growth and progression.

Exosomes contain a variety of RNA molecules, including mRNAs, miRNAs, lncRNAs, and circRNAs (Braicu et al., 2015; Yang and Li, 2018). Compared to other types of RNA, circRNAs are abundant and specific non-coding RNAs that are still not fully understood (Nie et al., 2020). CircRNAs was first discovered in Sendai virus and plant-like viruses by electron microscopy in 1976 (Kolakofsky, 1976). It has long been believed that these closed-loop covalent RNA molecules are a byproduct of rare error responses and have no specific functions (Cocquerelle et al., 1993). However, in recent years, many advances have been made in the study of circRNAs: circRNAs are produced by reverse splicing a 3’ splice donor to an upstream 5’ splice acceptor (Wilusz and Sharp, 2013). Because of this specific structure, circRNAs are resistant to exonucleases and exhibit greater stability than linear non-coding RNAs (such as miRNAs and lncRNAs) (Salzman et al., 2012; Jeck et al., 2013). CircRNA can be found in the cytoplasm, nucleus, or extracellular vesicles external to the cell. They usually perform different functions depending on their localization and distribution (Lasda and Parker, 2016; Ou et al., 2020). Although we have not been able to provide a complete explanation for the biological origins and functions of circRNAs, a large number of studies have confirmed that the main function of circRNAs is to effectively sponge miRNAs through a competitive endogenous RNA (ceRNA) mechanism, reduce their inhibitory effect on target genes (Hansen et al., 2013; Hu Y. et al., 2019; Liang et al., 2020), and activate or inhibit downstream signaling pathways by interacting with proteins (Du et al., 2017; Zang et al., 2020). Endogenous circRNAs have been reported to have the potential for translation (Legnini et al., 2017; Lei et al., 2020), and the products of translation may play important roles in disease progression.

CircRNAs can be encapsulated into exosomes that can be shared between cells. The entry of circRNAs into exosomes is influenced in part by the levels of relevant miRNAs in the parent cells (Li et al., 2015). The sorting of circRNAs species into exosomes may be positively regulated. This suggests that circRNAs are selectively encapsulated within exosomes (Hou et al., 2018). Interestingly, related studies have shown that circRNAs are much more enriched in exosomes than in the cells producing them (Dignat-George and Boulanger, 2011; Dou et al., 2016), and that exosomal circRNAs levels are only moderately correlated with the cellular circRNAs levels (Li et al., 2015). Many questions still remain about the mechanisms involved in exosomal circRNAs enrichment. Emerging data indicate that exosomal circRNAs have multiple functions, such as promoting inflammatory responses, regulating hormone levels in the body, and modulating immunity (Table 1). For example, Wang et al. (Wang S. et al., 2020) found that circRNA-0077930 in exosomes released from endothelial cells induces vascular smooth muscle cell senescence in high-glucose environments. This study provides new insights into the mechanism of smooth muscle cell aging in a high-glucose environment. This finding has significant implications for the prevention of diabetic vascular disease and the exploration of new drug targets. Exosomes released from astrocytes deliver circSHOC2 (Chen et al., 2020b) to neuronal cells, thereby ameliorating ischemia-induced apoptosis and protecting neurons from ischemic injury. In case of hypoxia, circHIPK3 (Wang et al., 2020e) is transferred to myocardial endothelial cells via exosomes released from cardiomyocytes. CircHIPK3-mediated VEGFA overexpression significantly increases cell proliferation and migration, which preserves the function and integrity of post-infarction myocardial endothelial cells and exhibits cardioprotective effects.

Recently, Li et al. (Li et al., 2015) first revealed the presence and enrichment of circRNAs in cancer-derived exosomes by RNA-seq analysis. CircRNAs play significant roles in regulating tumorigenesis, tumor cell progression, metastasis, and drug resistance development. For instance, Xiao et al. (Dai et al., 2018) transferred circRNA_100284 secreted by arsenite-transformed human liver epithelial cells (L-02) to normal L-02 cells via exosomes. CircRNA_100284 acted as a sponge for miRNA-217 to accelerate the cell cycle and promote cell proliferation, thereby inducing malignant transformation of L-02 cells. This mechanism provides a new explanation for arsenite-induced cancer. Glioma-derived exosomal circRNA-0001445 (Han et al., 2021) significantly promotes proliferation and inhibits the apoptosis of glioma cells via the miRNA-1275p/SNX5 signaling pathway. Shi et al. (Xiao and Shi, 2020) found that exosomal circRNA_400068 produced by renal cell carcinoma cells promotes the proliferation of healthy kidney cells, significantly inhibits their apoptosis, and promotes their transformation to malignancy, which may be due to signaling through regulation of the miRNA-210-5p/SocS1 axis pathway. Circ_UBE2D2 was found to be upregulated in exosomes isolated from triamcinolone-resistant breast cancer cells (Hu et al., 2020). Exosomes enhance the resistance of breast cancer cells to triamcinolone by mediating the transfer of circRNA_UBE2D2 into non-triamcinolone-resistant breast cancer cells. This mechanism may suggest the prospect of a promising candidate biomarker and therapeutic target for drug resistance in breast cancer.

Interestingly, researchers have identified that tumor cells produce more than 10 times more exosomes than normal cells. The enrichment and stable presence of circRNAs in exosomes and their intercellular transmission strongly implicate them in tumor metastasis (Li et al., 2015; Bao et al., 2016; Bai et al., 2019). There is increasing evidence that exosomal circRNAs play a key role in the metastasis of various cancers. Exosomal circRNAs may regulate tumor metastasis through a variety of different mechanisms, 1) cytoplasmic circRNAs adsorb miRNAs by sponging and de-repress miRNA-regulated genes, 2) circRNAs delivered to recipient cells can act as a protein sponge or decoy by adsorbing one or more proteins through specific binding sites, 3) exosomal circRNAs are able to regulate the TME through interactions with the immune system cells (Figure 1).

Figure 1 The potential mechanisms of exosomal circRNAs regulate tumor metastasis. After taken up by cancer cells, exosomal circRNAs can regulate the tumor metastasis by interacting with miRNAs or proteins. Exosomal circRNAs regulate tumor metastasis by sponging miRNAs.

CircRNAs represent upstream regulatory molecules with a large number of miRNA response elements (MREs) (Memczak et al., 2013; Lasda and Parker, 2014). CircRNAs can be delivered to recipient cells via exosomes and bind to miRNAs, acting as an “miRNA sponge” by blocking the binding of miRNAs to target gene 3’- untranslated regions (UTRs) to disable miRNAs and restore the translation of proteins inhibited by specific miRNA-mRNA binding events, thereby regulating tumor metastasis (Zhong et al., 2018; Verduci et al., 2019). CircRNAs exhibit more preferential binding to miRNAs than other endogenous RNAs (e.g., lncRNAs), hence the name (Shi et al., 2020) “super sponge”. Xie et al. (Xie et al., 2020) found that exosome circSHKBP1 increased HUR expression in gastric cancer (GC) tissues through sponging of miR-582-3p. HUR is reported to be part of the VEGF signaling pathway that promotes VEGF secretion and induces angiogenesis, which promotes GC metastasis. In triple-negative breast cancer (TNBC) (Yang S.-j. et al., 2021, 1), breast cancer cell metastasis is promoted through the direct targeting of AKT1 via the exosome circPSMA1 that functions as a sponge for miR-637, which activates the AKT1/β-catenin signaling pathway to regulate cell proliferation and migration. High expression of AKT1 and low expression of mir-637 are highly correlated with poor prognosis in TNBC patients with lymph node metastasis. CircPSMA1 overexpression significantly enhances the metastatic capacity in the liver and lungs of mice. Zeng et al. (Zeng et al., 2020) found that overexpression of circFNDC3B severely inhibits angiogenesis in a mouse model of colorectal cancer (CRC). Treatment with exosomal circFNDC3B inhibits CRC cell growth, angiogenesis, and liver metastasis in vivo. Liu et al. (Liu et al., 2020) found that circ_MMP2 could be delivered to less invasive hepatocellular carcinoma (HCC) cells via exosomes derived from highly invasive HCC cells. Circ_MMP2 can upregulate the expression of its host gene matrix metallopeptidase 2 (MMP2) by acting as a sponge for miR-136-5p, a metastasis-associated RNA that promotes HCC cell metastasis.

EMT is one of the initiating steps of primary tumor invasion (Nishiyama et al., 2018; Shang et al., 2019). Exosomal circRNAs can promote tumor metastasis by sponging miRNAs, thereby promoting the EMT phenotype. For example, circPRMT5 (Chen et al., 2018) is expressed at abnormally high levels in urothelial carcinoma of the bladder (UCB) cells, and can be secreted into the blood and urine via exosomes. High levels of exosomal circPRMT5 in the serum and urine are positively correlated with lymph node metastasis and tumor progression. CircPRMT5 contributes to a significant reduction in SNAIL1 downregulation by reducing the inhibitory effect of miR-30c, which in turn promotes UCB cell invasiveness and EMT. Statistical analyses showed that high expression of circPRMT5 in UCB tissues is positively correlated with late T and N stages, and is associated with low disease-free survival (DFS). CircNRIP1 is a promoter of EMT in GC (Zhang et al., 2019). The exosomal circNRIP1 sponges miR-149-5p to regulate the expression of AKT1 in GC, which in turn exerts a tumor-promoting effect in the context of EMT. Circ_PVT1 (Wang H. et al., 2020) can enter exosomes originating in cervical cancer cells and function as a tumor promoter by inducing EMT in cervical cancer cells by targeting miR-1286, which in turn promotes tumor metastasis (Table 2).

RNA binding proteins (RBPs), a class of proteins involved in gene transcription and translation, are essential elements of circRNA function. Bioinformatic analysis of circRNA sequences reveals a low enrichment of RBP-binding sites compared to the corresponding linear RNAs. However, the unique tertiary structure of circRNA leads to a greater protein binding capacity than linear RNA sequences, allowing better interaction with proteins (Hentze and Preiss, 2013; You et al., 2015; Liang et al., 2018; Huang A. et al., 2020). CircRNAs can function as protein sponges (Yang et al., 2017), decoys (Abdelmohsen et al., 2017), scaffolds, or recruiters (Zeng et al., 2017; Sun Y.-M. et al., 2019) in different physiological and pathological environments. CircRNA-protein interactions also play pivotal roles in the regulation of tumor metastasis.

In HCC, the exosomal circ-0004277 (Zhu et al., 2021) derived from HCC cells blocks the binding of HuR to ZO-1 mRNA by competitive interactions with HuR protein, which in turn stimulates EMT progression by inhibiting ZO-1. Moreover, exosome-delivered circ-0004277 induces EMT in adjacent normal cells, further promoting the invasion of HCC cells into the surrounding normal tissues. Similarly, Xie et al. (Xie et al., 2020) found that exosomal circSHKBP1 could not only promote GC progression by regulating the miR-582-3p/HUR/VEGF pathway, but also promotes tumor growth and lung metastasis by sequestering HSP90 away from STUB1. Xu et al. (Xu Y. et al., 2021) explored the function of circ-CCAC1 in cholangiocarcinoma (CAA), and found that circ-CCAC1 could translocate into vascular endothelial cells via exosomes and bind to EZH2 in the cytoplasm to inhibit the expression of intercellular junctional proteins (ZO-1 and occludin) that control endothelial cell permeability. Decreased expression of intercellular linker proteins disrupts the vascular endothelial barrier and induces angiogenesis, thereby promoting the formation of pre-metastatic ecological niches and providing a supportive microenvironment for the spread of cancer cells.

The TME is complicated and ever-evolving. In addition to stromal cells, fibroblasts, and endothelial cells, the TME includes both innate and adaptive immune cells (Hinshaw and Shevde, 2019). Adaptive immune cells are mainly T lymphocytes, which can directly contribute to, or stimulate other cells in the TME to influence tumor growth. They can be classified as “Th1” and “Th2” cells based on their differentiation status. Th1 cells control the pro-inflammatory phenotype, and Th2 cells coordinate the immunosuppressive phenotype (McGuirk and Mills, 2002). The innate immune cell types include macrophages, dendritic cells (DCs), neutrophils, myeloid-derived suppressor cells (MDSCs), natural killer cells (NK), and innate lymphocytes (ILC). The innate immune response can have a significant impact on the TME (Joyce and Pollard, 2009) either directly or indirectly (through control of T-cell fate) (Hinshaw and Shevde, 2019). In addition, immune cells interact with stromal cells, thus influencing tumor development (Tlsty and Coussens, 2006). Here, we focus on mesenchymal stem cells (MSCs), macrophages, and tumor-associated fibroblasts (Figure 2).

With the rapid development of high-throughput sequencing technology, an increasing number of exosomal circRNAs have been discovered and identified, and have been studied and proven to play important roles in tumorigenesis. The covalently closed loop structure of circRNAs prevents them from being easily degraded by nucleases. Compared with exosomal proteins and ncRNAs (miRNA, lncRNA), exosomal circRNAs from tissues and blood are more conserved, stable, and exhibit greater target specificity, which is characteristic of tumor metastasis-related markers. In addition, as exosomal circRNAs play important regulatory roles in tumor metastasis, they have significant potential as important candidate targets for tumor metastasis-related therapies, which opens new avenues for curing tumorigenic diseases. As the study of exosomal circRNAs is still in its infancy, many aspects remain limited. 1) The technology for the isolation and purification of exosomes is not yet sufficiently developed. Currently, there are four main methods for isolating and purifying extracellular vesicles, i.e., ultra-high speed centrifugation, ultrafiltration, precipitation, and immune-enrichment (van Niel et al., 2018). It is difficult to distinguish between exosomal and non-vesicular components using established purification methods, and this may affect subsequent experimental procedures involving exosomal circRNA, both in vivo and in vitro; 2) Although high-throughput sequencing techniques have identified many exosomal circRNAs that are aberrantly expressed in tumor tissues, their specific mechanisms of action and biological functions are still not fully understood; 3) The factors that determine the endogenous and exogenous nature of exosomes remain undetermined, which in part leads to the difficulty in using exosomal circRNAs as clinical markers of tumor metastasis; 4) Although an increasing number of studies have focused on exploring the use of exosomal circRNAs as biological markers for the diagnosis of certain tumors, only a few clinical trials have confirmed their feasibility.

In conclusion, this review has discussed the regulation of the occurrence and transport of exosomal circRNAs in various biological, physiological, or pathological processes by describing them as stellar molecules that have attracted much attention over recent years. Exosomal circRNAs are enriched in tumors and regulate tumor metastasis through mechanisms such as by acting as sponges for miRNAs, binding to proteins, and interacting with the TME. Although our current understanding of the functions of exosomal circRNAs is undoubtedly only the tip of the iceberg, the development of new technologies and assays would enable us to understand the regulatory mechanisms involving exosomal circRNAs, which will provide superior evidence for their use as early, novel markers of tumor metastasis, and as promising candidate therapeutic targets.