The potential of using blood circular RNA as liquid biopsy biomarker for human diseases

Circular RNA (circRNA) is a novel class of single-stranded RNAs with a closed loop structure. The majority of circRNAs are formed by a back-splicing process in pre-mRNA splicing. Their expression is dynamically regulated and shows spatiotemporal patterns among cell types, tissues and developmental stages. CircRNAs have important biological functions in many physiological processes, and their aberrant expression is implicated in many human diseases. Due to their high stability, circRNAs are becoming promising biomarkers in many human diseases, such as cardiovascular diseases, autoimmune diseases and human cancers. In this review, we focus on the translational potential of using human blood circRNAs as liquid biopsy biomarkers for human diseases. We highlight their abundant expression, essential biological functions and significant correlations to human diseases in various components of peripheral blood, including whole blood, blood cells and extracellular vesicles. In addition, we summarize the current knowledge of blood circRNA biomarkers for disease diagnosis or prognosis.

Due to the importance of liquid biopsy biomarkers in precision medicine (Vargas and Harris, 2016) and the superior characteristics of circRNAs as disease biomarkers (Zhang et al., 2018j), recent studies have put great efforts into researching the use of circRNAs as liquid biopsy biomarkers for human diseases. Here, we review current progress that has been made on this topic. We focus on circRNAs in the peripheral blood, although circRNAs are abundantly expressed in other body fluids, such as saliva (Bahn et al., 2014;Ghods, 2018) and urine (Kölling et al., 2019;Lam and Lo, 2019;Vo et al., 2019). In this review, we first briefly introduce the biogenesis, function, and expression patterns of circRNAs and their close correlations to human diseases. Then, we emphasize the functional roles they may play in different blood components, including blood cells, serum, plasma, platelets and EVs in the blood. We summarize peripheral blood circRNA biomarkers that have been constructed for the management of human diseases.
Finally, we discuss future opportunities and challenges of translating blood circRNAs into clinical practice.

THE BIOGENESIS OF CIRCRNAS
CircRNAs are derived from pre-mRNAs and formed by backsplicing, in which a downstream 5-prime splice site (ss) is covalently joined with an upstream 3-prime ss ( Fig. 1) (Li et al., 2018d). There are three major types of circRNAs, exonic circRNAs (ecircRNAs), exon-intron circRNAs (EI-circRNAs) and circular intronic RNAs (ciRNAs) (Fig. 1). EcircRNAs only contain exons, and most ecircRNAs are located in the cytoplasm. Two models have been proposed to explain the production of ecircRNAs (Fig. 1A) (Chen, 2016;Wu et al., 2017a;Kristensen et al., 2019;Xiao et al., 2020). The first model suggests that when the pre-mRNA is partially folded during transcription, canonical splicing will cause an "exon jump" and generate a linear RNA containing skipped exons (Starke et al., 2014;Kelly et al., 2015). A subsequent back-splicing event will turn the 'lariat intermediate' into a closed circular transcript (Barrett et al., 2015;Li et al., 2018d). The second model proposes that when the 5-prime ss is pulled closer to the 3-prime ss by base pairing of flanking intronic complementary sequences (Ivanov et al., 2014;Vicens and Westhof, 2014;Zhang et al., 2014;Wilusz, 2015) or specific bindings of intronic sequences to RBPs (Conn et al., 2015), a back-splicing event may occur, and a circRNA may be generated (Jeck et al., 2013). These models are supported by accumulating evidence showing that cis-acting elements (Jeck et al., 2013;Zhang et al., 2014;Yoshimoto et al., 2019) and trans-acting factors (Ashwal-Fluss et al., 2014;Conn et al., 2015;Agirre et al., 2019) play crucial regulatory roles in circRNA formation. EIcircRNAs are generated when an intron or several introns are alternatively retained in splicing (Li et al., 2015c). Therefore, EIcircRNAs are mainly considered intermediates of ecircRNAs (Li et al., 2015c). CiRNAs, which only contain introns, are produced through escape from debranching of intron lariats (Zhang et al., 2013). This process depends on a GU-rich motif near the 5-prime ss and a C-rich motif close to the branch-point site (Zhang et al., 2013). Unlike ecircRNAs, EIcircRNAs and ciRNAs are mainly located in the nucleus.

THE FUNCTION OF CIRCRNAS
With the rapid application of high-throughput RNA-sequencing (RNA-seq) technology (Stark et al., 2019) and related bioinformatics tools that identify circRNAs from RNAseq datasets (Jeck and Sharpless, 2014;Szabo and Salzman, 2016;Li et al., 2017a;Gao and Zhao, 2018), more than one million circRNAs have been annotated in model organisms (Glažar et al., 2014;Vo et al., 2019;Cai et al., 2020;Wu et al., 2020a). However, the functional characterization of circRNAs is still at its early stage (Chen, 2016(Chen, , 2020Li et al., 2018d). Their circular conformation and the large sequence overlap with their linear mRNA counterparts have posed interact with RNA Pol II and U1 snRNP to regulate gene transcription in the nucleus (i). Second, ecircRNAs can accumulate in the cytoplasm and act as miRNA decoys (ii), protein regulators (iii), and translation templates (iv). Third, circRNAs can be secreted into EVs by many cell types and transported to recipient cells by EVs. EV circRNAs can also act as important gene regulators. substantial challenges in investigating the functions of cir-cRNAs (Li et al., 2018d). Based on our current knowledge, circRNAs have diverse functions as miRNA decoys, protein regulators and translation templates (Fig. 1B, see (Chen, 2016(Chen, , 2020Li et al., 2018d) for excellent reviews). (B) Some exosomal circRNAs, such as circ-DB (i), circSHKBP1 (ii), and circUHRF1 (iii), are important regulators in oncogenic pathways, while some circRNAs in exosomes, such as circHIPK3 (iv), play key roles in the release of inflammatory cytokines.
(C) Peripheral blood circRNAs, both cell-free circRNAs and intracellular circRNAs in blood cells, have potential clinical applications as liquid biopsy biomarkers in many human diseases, such as the diagnosis, prognosis and treatment guidance of many human diseases, including autoimmune diseases, cancers, cardiovascular diseases, immuno-deficiency diseases, infectious diseases, and neurodegenerative diseases.
Translational potential of blood circular RNAs

MiRNA decoys
The most well-known function of circRNAs is that they can act as a sponge to inhibit the function of miRNAs and indirectly regulate the expression of miRNA target genes at the posttranscriptional level (

Protein regulators
In addition to miRNAs, circRNAs can interact with proteins or protein complexes to regulate protein expression and function ( Fig. 1B) (Huang et al., 2020a). First, circRNAs can bind to some RBPs and act as RBP sponges. CircMbl, a circRNA originating from the second exon of the splicing factor (B) Exogenous circRNAs released by viruses can be recognized by RIG-I, thus activating the host innate immunity to viruses (i). Moreover, NF90/NF110 not only promotes circRNA production in the nucleus but also interacts with mature host circRNAs to form circRNP complexes in the cytoplasm. Upon viral infection, NF90/NF110 can be released from circRNP complexes and bind viral mRNAs to inhibit viral replication (ii). In addition, circRNAs can form RNA duplexes and act as inhibitors of PKR under normal conditions. When a virus invades the cells of its host, RNase L is activated to efficiently degrade circRNAs, and PKR is released and activated to initiate the early cellular innate immune response (iii).

REVIEW
Guoxia Wen et al.

REVIEW
Guoxia Wen et al. at the promoter region of their parental genes and enhances their transcription (Zhang et al., 2013;Li et al., 2015c). Similarly, some ciRNAs, such as ci-ankrd52, can also associate with Pol II as well (Zhang et al., 2013;Li et al., 2015c). They can also regulate the expression of their parental gene by modulating the elongation of Pol II (Zhang et al., 2013;Li et al., 2015c). Many circRNAs, rather than a specific circRNA, can interact with NF90/NF110 as a group to form circRNP complexes in the cytoplasm, which may function as a reservoir of NF90/NF110 (Li et al., 2017e). Upon viral invasion, NF90/NF110 could be released from circRNP complexes and bind to viral mRNAs to perform its antiviral functions (Li et al., 2017e). Third, the interaction of circRNAs with proteins can facilitate or block the function of their binding proteins. For instance, increased expression of circFoxo3 in the cytoplasm may arrest the anti-senescent protein ID1, transcription factor E2F1, anti-stress proteins FAK and HIF1α and prevent the nuclear translocation of these transcription factors and their function, thus leading to increased cellular senescence (Du et al., 2017). CircFoxo3 can also form a tertiary circRNA-protein complex with cyclindependent kinase 2 (CDK2) and cyclin-dependent kinase inhibitor 1 (p21), which facilitates the inhibition of CDK2 by The spatiotemporal expression of circRNAs is mediated by the balance between circRNA generation and circRNA turnover, which can be regulated at both the transcriptional and posttranscriptional levels (Li et al., 2018d;Chen, 2020). On the one hand, the production of circRNAs can be regulated by intronic complement sequences (ICSs) (Jeck et al., 2013;Liang and Wilusz, 2014;Zhang et al., 2014), cissplicing elements (Ashwal-Fluss et al., 2014;Starke et al., 2014;Wang and Wang, 2015) and trans-acting splicing Translational potential of blood circular RNAs REVIEW factors (Ashwal-Fluss et al., 2014;Conn et al., 2015;Kramer et al., 2015;Khan et al., 2016;Aktaş et al., 2017;Errichelli et al., 2017;Fei et al., 2017;Li et al., 2017e). On the other hand, the degradation of circRNAs can be affiliated by miRNA-initiated AGO2 cleavage (Kleaveland et al., 2018) and nuclease-mediated cleavage, including RNase L endonuclease , RNase P/MRP endonuclease complex (Park et al., 2019) and G3BP1 endonuclease (Fischer et al., 2020). Dysregulation of circRNA generation or turnover may lead to aberrant circRNA expression in cells or tissues, which may be related to many human diseases . Cdr1as, one of the most studied circRNAs, harbors more than 70 binding sites for miR-7, and its overexpression strongly inhibits the activity of the tumor suppressor miR-7 (Hansen et al., 2013). Accumulating evidence has demonstrated its significant increase in colorectal cancer samples, which can inhibit miR-7 function and activate EGFR and RAF1 oncogenes (Weng et al., 2017). Moreover, the dysregulation of the miR-7/Cdr1as axis is involved in some other diseases, such as Alzheimer's diseases (Lukiw, 2013), diabetes (Xu et al., 2015) and cardiometabolic diseases (Geng et al., 2016). Furthermore, the loss of the Cdr1as locus causes miRNA deregulation and affects synaptic transmission in knockout mice (Piwecka et al., 2017). In addition to Cdr1as, other circRNAs are also implicated in many different human diseases, including cancers (Zhang et al., 2018j;Smid et al., 2019;Su et al., 2019;Vo et al., 2019), cardiovascular diseases (Zhang et al., 2018j;Aufiero et al., 2019), neurodevelopment and neurodegenerative diseases (Kumar et al., 2017;Zhang et al., 2018j;Dube et al., 2019;Chen et al., 2020d;Hanan et al., 2020) and immune diseases (Chen et al., 2019c;Zhou et al., 2019b).

CIRCRNAS ARE PROMISING BIOMARKERS FOR HUMAN DISEASES
In general, biomarkers are physiological indexes or biological molecules that can be objectively measured and can indicate either a normal or a pathogenic state (Lesko and Atkinson, 2001). Technological advances in genomics, transcriptomics, proteomics and metabolomics have led to the discovery and validation of many biomarkers that are pushing personalized medicine forward (Chen et al., 2012;Karczewski and Snyder, 2018;Ahadi et al., 2020). A good biomarker with clinical significance must meet the criteria of analytical validity, clinical validity and clinical utility (Byron et al., 2016). As a novel type of noncoding RNA, circRNA has several distinct advantages over canonical linear RNAs as a disease biomarker (Zhang et al., 2018j). First, circRNA is more stable than linear RNAs because it has a closed loop structure without 5-prime and 3-prime ends (Enuka et al., 2015;Li et al., 2015b). In the MCF10A human mammary epithelial cell line, Enuka et al. observed at least 2.5 times longer half-lives of circRNAs compared to linear RNAs, including mRNAs and miRNAs (Enuka et al., 2015). Second, circRNAs are abundantly expressed in many tissue samples. For example, brain samples express more than 10,000 circRNAs in the rat BodyMap dataset (Zhou et al., 2018a). In some cases, the expression values of circRNAs are much higher than their linear counterparts (Westholm et al., 2014;Rybak-Wolf et al., 2015;You et al., 2015;Liang et al., 2017). Third, circRNAs are expressed in a tissue-specific (Salzman et al., 2013;Guo et al., 2014;Szabo et al., 2015;Zhou et al., 2018a) and developmental stage-specific manner (Rybak-Wolf et al., 2015;Szabo et al., 2015;You et al., 2015;Zhou et al., 2018a). Importantly, the tissue specificity of circRNA is higher than that of mRNA of its host gene (Guo et al., 2014;Zhou et al., 2018a). These features suggest that circRNAs may have better analytical validity (Zhang et al., 2018j), including analytical specificity, robustness, reproducibility and repeatability (Byron et al., 2016), when used as biomarker molecules. In a recent study, Maass et al. investigated circRNA expression in 20 clinically relevant tissue samples, which underscored the feasibility of using cir-cRNAs as potential disease biomarkers (Maass et al., 2017).
To date, many circRNAs have been identified as biomarkers for human diseases (Zhang et al., 2018j), especially human cancers (Li et al., 2015a;Meng et al., 2017;Su et al., 2019;Sheng et al., 2020). For example, Cdr1as has been revealed as a prognostic biomarker in colorectal cancer patients (Weng et al., 2017). In a training cohort comprising 153 primary colorectal cancer tissues and 44 matched normal mucosae, significantly increased Cdr1as expression was observed in colorectal cancer tissues, and its overexpression was associated with poor patient survival. The prognostic power of Cdr1as was further validated in an independent validation cohort comprising 165 colorectal cancer patients (Weng et al., 2017). A four-circRNA signature, consisting of hsa_circ_101308, hsa_circ_104423, hsa_circ_104916 and hsa_circ_100269, has been constructed that can predict the early recurrence of stage III gastric cancer (GC) patients after surgery . Moreover, Vo et al. constructed a comprehensive catalog of circRNAs in human cancer tissues in a systematic analysis of more than 2,000 cancer samples (Vo et al., 2019). They identified two circRNAs, circ-AURKA and circ-AMACR, as potential diagnostic biomarkers for neuroendocrine prostate cancer (Vo et al., 2019). Although these circRNA biomarkers are potentially useful in cancer management, most of them are derived from tissue samples (Li et al., 2015a;Meng et al., 2017;Su et al., 2019;Sheng et al., 2020). To improve biomarker accessibility, especially for biomarkers of cancer screening and diagnosis, circRNA biomarkers in body fluids are ideal for clinical application (Anfossi et al., 2018). Hence, the authors further investigated and validated the reliability of detecting circRNA biomarkers in urine samples of prostate cancer patients (Vo et al., 2019), which suggests that circRNA in urine samples is a promising strategy for prostate cancer screening.

BLOOD CIRCRNAS AND LIQUID BIPOSY BIOMARKERS
Liquid biopsy has been a revolutionary tool in disease management, supporting the diagnosis, prognosis and treatment guidance of human diseases (Heitzer et al., 2019;Mattox et al., 2019;Rubis et al., 2019;Luo et al., 2020a). In comparison to urine, saliva or cerebrospinal fluid, peripheral blood has been used as the major body fluid in liquid biopsy (Rubis et al., 2019;Luo et al., 2020a). In the circulating blood, aberrantly expressed RNAs or fusion transcripts in different blood components have been associated with human cancers (Byron et al., 2016;Zaporozhchenko et al., 2018;Sole et al., 2019) and infectious diseases (Byron et al., 2016;Correia et al., 2017). These RNA biomarkers include cfRNAs in plasma or serum (Mitchell et al., 2008;Fehlmann et al., 2020), exosome-derived RNAs (Maas et al., 2017), EV-incorporated cfRNAs (Maas et al., 2017) and RNA transcripts in tumor-educated platelets (Best et al., 2017(Best et al., , 2018. These cell-free or EV-incorporated RNA biomarkers represent the changes in expression that occur in abnormal cells, such as dysregulated genes in cancer cells (Byron et al., 2016;Rubis et al., 2019;Luo et al., 2020a). Other than cfRNAs, gene expression profiles of PBMCs or whole blood have been proven to be good indicators of many human diseases (Chaussabel and Baldwin, 2014;Chaussabel, 2015), since they can assess the immune status (Chaussabel, 2015). Unlike the expression of cfRNAs in serum, plasma or EVs, RNA expression levels in PBMCs or whole blood are measures of the host response to exogenous pathogens or autoantigens (Schnell et al., 2018;Shaked, 2019). Several whole blood or PBMC gene expression signatures have been developed for cancer management, including the early diagnosis of colorectal cancer (Marshall et al., 2010;Ciarloni et al., 2015Ciarloni et al., , 2016 and lung cancer (Showe et al., 2009;Kossenkov et al., 2018), the prognosis of adult acute myeloid leukemia (Bullinger et al., 2004;Valk et al., 2004) and prostate cancer patients (Ross et al., 2012), and the monitoring of renal cell carcinoma relapse (Giraldo et al., 2017). Moreover, whole blood or PBMC RNA signatures have been widely investigated as liquid biopsy diagnostic tools for infectious diseases (Ramilo and Mejias, 2017), such as discriminating influenza from other respiratory viral infections (Zaas et al., 2009), differentiating viral and bacterial infections (Tsalik et al., 2016), diagnosing septic patients (Sweeney et al., 2018;Gunsolus et al., 2019;Mayhew et al., 2020), and discriminating active tuberculosis (TB) patients from patients with latent TB or other lung diseases (Bloom et al., 2013;Anderson et al., 2014;Qian et al., 2016;Zak et al., 2016;Sambarey et al., 2017;MacLean et al., 2019;Warsinske et al., 2019;Esmail et al., 2020;Gupta et al., 2020). In addition, the transcriptome of blood cells has been implicated in assessing the likelihood of developing obstructive coronary artery disease in symptomatic nondiabetic patients (Vargas et al., 2013) and predicting antibody-mediated kidney allograft rejection in kidney transplant patients (Loon et al., 2019).
With regard to RNA molecules, both protein-coding mRNAs (Byron et al., 2016;Sole et al., 2019) and several classes of noncoding RNAs (Byron et al., 2016;Anfossi et al., 2018;Pardini et al., 2019;Sole et al., 2019) have been used as blood disease biomarkers. In comparison to mRNAs or long noncoding RNAs, small noncoding RNAs, such as miRNAs (Mitchell et al., 2008;Max et al., 2018;Fehlmann et al., 2020) and noncanonical small RNAs (Fritz et al., 2016;Pardini et al., 2019), have the advantage of high stability in the circulating blood (Mitchell et al., 2008;Anfossi et al., 2018). This superiority is especially important when translating RNA biomarkers into clinical practice (Byron et al., 2016;Anfossi et al., 2018), since fast mRNA degradation in blood sample processing may affect the performance of mRNA biomarkers (Dvinge et al., 2014). Given the high stability of circRNAs, great efforts and substantial progress have been made to investigate the possibility of using cir-cRNA biomarkers in liquid biopsy in recent years.
Next, we summarize circRNAs in the peripheral blood, their correlations with human diseases and their potential application in liquid biopsy as disease biomarkers (Fig. 2). We classified blood circRNAs into blood cell-free circRNAs and circRNAs in blood cells since they have distinct biological meanings in the context of human diseases. Blood cellfree circRNAs, including circulating cell-free circRNAs in plasma, serum and blood EVs, are secreted from different tissue cells into the blood. Therefore, cell-free circRNAs have corresponding tissue origins, and cell-free circRNA biomarkers represent their clinical significance in the original tissue (Li et al., 2019d. On the other hand, blood cell circRNAs consist of circRNAs in various blood cells, such as monocytes, erythrocytes, neutrophils and platelets. Cir-cRNAs in mixtures of different blood cells, such as circRNAs in lymphocytes, PBMCs and whole blood, are classified in this group as well. The circRNA expression profiles of peripheral blood cells or a mixture of blood cells are important indicators of the host's immune status (Chaussabel, 2015), which may undergo dynamic changes during acute events such as viral infections Rose et al., 2019;Zhou et al., 2019a). Hence, circRNA biomarkers in blood cells, PBMCs or whole blood represent the specific immune response of an individual to different physiological aspects.

Cell-free circRNAs in peripheral blood
Mounting evidence has demonstrated abundant circRNA expression in blood plasma (Maass et al., 2017;Yi et al., 2018;Smid et al., 2019) and serum (Gu et al., 2017;Maass et al., 2017;Sonnenschein et al., 2019; ( Fig. 2A). In the circulating blood, EVs, including exosomes and microvesicles, are important blood components that have great diagnostic and prognostic value (Revenfeld et al., 2014). Li et al. investigated circRNA expression in MHCC-Translational potential of blood circular RNAs REVIEW LM3 liver cancer cells and cell-derived exosomes and found at least 2-fold circRNA enrichment in exosomes compared to producer cells (Li et al., 2015b). They further found that the expression level of exosomal circRNAs is largely the same after serum incubation at room temperature for up to 24 h (Li et al., 2015b). Using the extracellular vesicle long RNA (exLR) sequencing method, the same group explored the expression profiles of circRNAs in 352 plasma EV samples (Li et al., 2019d). They found 137,196 circRNA candidates that were expressed in normal plasma EV samples, and the circular to linear ratio was significantly higher in EVs than in PBMCs (Li et al., 2019d). A recent study also isolated platelet-derived EVs and demonstrated the selective release of platelet-specific circRNAs in exosomes and microvesicles (Preußer et al., 2018). Based on these data, a database of blood exosomal circRNAs, exoRBase, has been developed to facilitate the development of circRNA signatures in blood exosomes . In addition, two previous studies have observed that the majority of circulating cell-free miR-NAs were associated with AGO2 protein complexes rather than with vesicles in blood plasma (Arroyo et al., 2011;Turchinovich et al., 2011). Like cell-free miRNAs, blood cellfree circRNAs may bind to several RNA-binding proteins or protein complexes as well (Huang et al., 2020a). Therefore, some cell-free circRNAs may not be associated with blood EVs. In a pilot study of circRNA expression profiles in 348 primary breast cancer tissue samples, Smid et al. found that circCNOT2 is a prognostic biomarker of aromatase inhibitor therapy in advanced breast cancer patients (Smid et al., 2019). To investigate the possibility of using circCNOT2 as a noninvasive biomarker, they further amplified circCNOT2 in plasma samples of four breast cancer patients and found variable circCNOT2 expression in all plasma samples (Smid et al., 2019). All these results suggest that cell-free circRNAs in peripheral blood are stable, enriched and detectable.
Blood EVs can be secreted from cells of many different biological systems and can be circulated to recipient cells through the bloodstream (Revenfeld et al., 2014;Lasda and Parker, 2016). Therefore, blood cell-free circRNAs can perform vital roles in many biological processes, such as cancer cell proliferation, cancer metastasis, drug resistance, hemostasis and inflammation (Fig. 2B) (Wang et al., 2019e;Cui et al., 2020;Shang et al., 2020;Xu et al., 2020b). For example, circ-deubiquitination (circ-DB) in adipose-secreted exosomes was found to regulate deubiquitination via the suppression of miR-34a and the activation of deubiquitination-related USP7 in plasma samples of hepatocellular carcinoma (HCC) patients, which could reduce DNA damage and promote HCC cell growth . Exosomal circSHKBP1 was found to promote the progression of GC by regulating the miR-582-3p/HUR/VEGF pathway and suppressing HSP90 degradation . CircUHRF1 in plasma exosomes can inhibit the functions of natural killer cells by upregulating the expression of TIM-3 via miR-449c-5p degradation and drive resistance to anti-PD1 immunotherapy in HCC patients (Zhang et al., 2020a). Moreover, exosomal circHIPK3 can prevent ischemic muscle injury by downregulating miR-421 expression, increasing FOXO3a expression, inhibiting pyroptosis and releasing IL-1β and IL-18 .
Given the biological functions that blood cell-free cir-cRNAs can perform (Wang et al., 2019e;Cui et al., 2020; and their implications in many human diseases, cell-free circRNAs in the peripheral blood have great potential as liquid biopsy biomarkers of human diseases (Fig. 2C) (Anfossi et al., 2018;Zaporozhchenko et al., 2018;Zhang et al., 2018j). To date, many blood cell-free circRNAs have been introduced for cancer management, including early cancer diagnosis, cancer prognosis and prediction of cancer treatment (Preußer et al., 2018;Aufiero et al., 2019;Fraipont et al., 2019;Lu et al., 2019c;Pardini et al., 2019;Sole et al., 2019;Su et al., 2019;Wang et al., 2019e;Beltrán-García et al., 2020;Cui et al., 2020;. In a recent study, Luo et al. measured the expression levels of two circRNAs in plasma samples of 231 lung cancer patients and 41 healthy controls using reverse transcription droplet digital PCR (RT-ddPCR) (Luo et al., 2020c). They identified hsa_circ_0000190 as a circRNA biomarker in human blood plasma that can predict the survival outcomes of lung cancer patients (Luo et al., 2020c). Furthermore, the increased expression of plasma hsa_circ_0000190 was also correlated with poor response to systemic therapy and immunotherapy in lung cancer patients (Luo et al., 2020c). Similarly, Li et al. investigated the clinical relevance of serum exosomal cir-cFLI1 in lung cancer patients in a cohort of 61 small cell lung cancer (SCLC) patients and 55 normal subjects. They found that serum exosomal circFLI1 levels were significantly higher in SCLC patients, especially in SCLC patients with distant metastasis . Notably, they observed that SCLC patients with lower exosomal circFLI1 expression levels experienced longer disease remissions, indicating its prognostic power in SCLC. The authors also suggested that serum exosomal circFLI1 may be used as a biomarker that can monitor the clinical response to chemotherapy in SCLC patients . By analyzing blood plasma samples of 62 GC patients and 25 healthy controls, Tang et al. proposed a novel circulating diagnostic biomarker of GC, plasma exosomal circKIAA124, that was correlated with clinical TNM stage, lymphatic metastasis and overall survival time of GC patients (Tang et al., 2018). In addition to the aberrant expression of blood cell-free circRNAs, the presence of some fusion circRNAs (f-circRNAs) in the peripheral blood has also been used as a liquid biopsy biomarker. Guarnerio et al. found that f-circRNAs could be produced from cancer-associated chromosomal translocations in cancer cells (Fig. 1), and f-circRNAs could promote cellular transformation, cell viability and resistance upon therapy (Guarnerio et al., 2016). In a systematic analysis of f-cir-cRNAs in localized prostate cancer tissues, Chen et al. observed more f-circRNAs in tumors with worse prognosis . Due to the lack of recurrence of these f-circRNAs, the authors suggested that f-circRNAs are good Guoxia Wen et al. biomarker candidates . In NSCLC patients, Tan et al. found that F-circEA, an f-circRNA originating from the EML4-ALK fusion gene, was exclusively expressed in the plasma of patients with the EML4-ALK fusion . Therefore, plasma F-circEA may serve as a liquid biopsy biomarker to diagnose NSCLC patients with EML4-ALK translocation and guide targeted therapy for NSCLC patients in this subgroup . More cell-free circRNA biomarkers in the blood are summarized in Table 1.

CircRNAs in blood cells and whole blood
CircRNA expression in blood cells and whole blood, a major source of liquid biopsy samples, has been extensively investigated ( Fig. 2A). In a pilot study, Memczak et al. detected thousands of circRNAs in peripheral whole blood samples using RNA-seq (Memczak et al., 2015). They found that the expression levels of these blood circRNAs were comparable to those in circRNA-rich cerebellar tissue (Memczak et al., 2015). In separate blood cell populations, circRNAs were observed to be enriched approximately 100-fold in platelets and anucleate erythrocytes relative to nucleated tissues, such as lung, brain and colon (Alhasan et al., 2016). Gaffo et al. investigated circRNA expression in T cells, B cells and monocytes of healthy subjects and found abundant circRNA expression in these mature blood cells (Gaffo et al., 2019). We also explored the expression landscape of circRNAs in PBMCs and found that the expression level of circRNAs in PBMCs, together with that in platelets, red blood cells (RBCs) and whole blood, is high enough to be detected (Qian et al., 2018). All these results suggest that whole blood (Memczak et al., 2015;Qian et al., 2018), PBMCs (Qian et al., 2018), and several blood cells, including neutrophils (Maass et al., 2017) (Alhasan et al., 2016;Qian et al., 2018) and platelets (Maass et al., 2017;Qian et al., 2018;Gaffo et al., 2019), are reliable clinical samples for circRNA profiling in liquid biopsy.
Accumulating evidence has suggested that circRNAs play crucial roles in the immune response and its regulation (Fig. 3) (Chen et al., 2017b(Chen et al., , 2019cXu et al., 2018bXu et al., , 2020aYang et al., 2018;Zhou et al., 2019b;Awan et al., 2020). First, circRNAs are important regulators of blood cell biogenesis, differentiation and activation (Chen et al., 2019c;Zhou et al., 2019b;Xu et al., 2020a). In a comprehensive study of circRNA expression in hematopoietic progenitors, differentiated lymphoid and myeloid cells, Nicolet et al. observed a cell-type specific pattern of circRNA expression profiles in blood cells, and the type and number of circRNAs increased upon hematopoietic maturation (Nicolet et al., 2018). Moreover, Holdt et al. found that circANRIL can bind to PES1 to impede the generation of pre-rRNA and ribosomes, resulting in the biogenesis of macrophages by p53 activation (Holdt et al., 2016). Second, circRNAs are actively involved in antiviral immune responses (Fig. 3B) (Cadena and Hur, 2017;Awan et al., 2020). For example, Chen et al. showed that exogenous circRNAs released by viruses can be recognized by the pattern recognition receptor RIG-I of host cells, which activates host innate immunity (Chen et al., 2017a). In their subsequent work, the authors showed that m 6 A RNA modification of human circRNAs inhibits innate immunity, while unmodified circRNAs and K63-polyubiquitin can activate RIG-I and innate immune response . Li et al. found that two immune factors, NF90/NF110, not only promote circRNA production in the nucleus but also bind to mature host circRNAs to form circRNP complexes in the cytoplasm. Upon viral infection, circRNP complexes in the cytoplasm can release NF90/NF110, which binds viral mRNAs to inhibit viral replication (Li et al., 2017e). Moreover, Liu et al. presented that circRNAs can form RNA duplexes and act as inhibitors of innate immunity-related protein kinase (PKR) under normal conditions . Upon poly(I:C) treatment or viral infection, RNase L is activated to efficiently degrade circRNAs, and PKR is thus released from circRNA inhibition to initiate the early cellular innate immune response . Third, circRNAs are closely associated with the antibacterial immune response as well. Ng et al. characterized circRNAs induced by lipopolysaccharide (LPS) and identified circRasGEF1B as a conserved positive regulator of the LPS response (Ng et al., 2016). Their functional analysis revealed that cir-cRasGEF1B can induce the expression of ICAM-1 in the TLR4/LPS pathway, which activates pathogen recognition and the inflammatory response upon microbial infection (Fig. 3A) (Ng et al., 2016).
The important functions of circRNAs in blood cells suggest that the dysregulation of circRNA expression in blood cells is likely to contribute to the occurrence and progression of immune-related diseases, including autoimmune diseases, infectious diseases and cardiovascular diseases (Chen et al., 2019c;Gaffo et al., 2019;Zhou et al., 2019b;Xie et al., 2020b;Xu et al., 2020a). For instance, we determined that the expression level of PBMC circRNAs is higher in TB patients than healthy controls, and five immune-related pathways were upregulated upon Mycobacterium tuberculosis infection, including "cytokine-cytokine receptor interactions", "chemokine signaling pathways", and "neurotrophic signaling pathways" (Qian et al., 2018). Similarly, Huang et al. identified 13 upregulated and 24 downregulated cir-cRNAs in PBMCs of TB patients (Huang et al., 2018c). Among them, hsa_circRNA_001937 is likely to participate in the inflammatory response by targeting miR-26b and modulating the NF-κB pathway (Huang et al., 2018c). In addition, Zhang et al. investigated the roles that circRNAs play in early human immunodeficiency virus (HIV) infection (EHI), especially in regulating HIV replication (Zhang et al., 2018i). EHI represents a stage where viral replication increases to a peak level and intense antiviral immune response and immune injury occur (Powers et al., 2011;Richey and Halperin, 2013). The authors characterized the expression profiles of circRNAs, mRNAs and miRNAs in PBMCs of EHI Translational potential of blood circular RNAs REVIEW patients and constructed a circRNA-associated competing endogenous network in EHI patients. They revealed 67 differentially expressed circRNAs, such as CCNK, CDKN1A and IL-15, that are potentially involved in HIV replication by regulating the expression of genes in the immune response, inflammatory response and defense response to the virus (Zhang et al., 2018i). Regarding autoimmune diseases, Liu et al. detected a global reduction in circRNAs and activation of RNase L in PBMCs of systemic lupus erythematosus (SLE) patients . They further found that circPOLR2A overexpression can lead to reduced PKR activation, EIF2α phosphorylation and type I IFN-induced gene suppression. This highlights the link between circRNAs and innate immunity regulation and provides the potential for circRNA manipulation in SLE treatment . In addition to the above examples, the abnormal expression of blood circRNAs is related to several other diseases, such as the immune response to Ebola virus  and hepatitis C virus (Jost et al., 2018), rheumatoid arthritis (RA) , type 2 diabetes mellitus (T2DM) (Fang et al., 2018), heart failure (Han et al., 2020b) and adenosine deaminase deficiency (Maass et al., 2017).
With increasing knowledge of blood cell circRNAs and their function, many circRNAs in blood cells or whole blood have been proposed as liquid biopsy biomarkers for human diseases (Fig. 2C) (Aufiero et al., 2019;Beltrán-García et al., 2020;Kumar et al., 2017;Ravnik-Glavač and Glavač, 2020;Sun et al., 2020a). For instance, we developed a PBMC circRNA-based molecular signature that discriminates active TB patients from healthy controls in our previous study (Qian et al., 2018). The classification power of this PBMC circRNA signature was further validated in an independent cohort with an area under the receiver operating characteristic curve (AUC) of 0.946 (Qian et al., 2018). In another study, Huang et al. found that the expression of hsa_circ_001937 in PBMCs was significantly higher in TB patients than in pneumonia, chronic obstructive pulmonary disease (COPD) and lung cancer patients (Huang et al., 2018c). In a cohort consisting of 115 TB, 40 pneumonia, 40 COPD, and 40 lung cancer patients and 90 healthy control subjects, hsa_circ_001937 had good discriminative power with an AUC of 0.873. After anti-TB treatment, the expression level of hsa_circ_001937 was significantly decreased compared to that of healthy controls. These results suggest that PBMC hsa_circ_001937 may be a TB diagnostic biomarker (Huang et al., 2018c). Furthermore, Lei et al. found an upregulation of circ_0000798 expression in PBMCs of HCC patients, which was associated with poor overall survival (Lei et al., 2019). In a cohort of 72 HCC patients and 30 healthy control subjects, circ_0000798 expression could distinguish HCC patients from healthy controls with an AUC of up to 0.703. The authors suggested that PBMC circ_0000798 has potential as a blood biomarker for HCC diagnosis and prognosis (Lei et al., 2019). In addition, Li et al. measured circRNA expression changes between children with SLE and healthy children and investigated the significance of blood circRNAs in SLE diagnosis . They identified and validated the diagnostic power of two circRNAs in whole blood, hsa_circ_0057762 and hsa_circ_0003090, that can differentiate children with SLE from healthy controls . Zhao et al. also identified and validated hsa_-circ_0054633 in peripheral whole blood as a sensitive and specific diagnostic biomarker for prediabetes and T2DM . In addition to the above examples, the potential of using blood circRNAs as disease biomarkers has been explored for many other human diseases, such as coronary artery disease (Zhao et al., 2017a;Wang et al., 2019a;Liang et al., 2020), community-acquired pneumonia , and schizophrenia (Yao et al., 2019). A list of current proposed potential blood circRNA biomarkers is listed in Table 2.

CONCLUSIONS AND FUTURE PERSPECTIVES
The high stability, abundance and spatiotemporal specific expression of blood circRNAs make them ideal biomarkers for liquid biopsy. In the past several years, many studies have shown that blood circRNAs, both cell-free blood cir-cRNAs and circRNAs in blood cells, have great potential as biomarkers of many human diseases in liquid biopsy. A biomarker with broad clinical application must have demonstrated analytical validity, clinical validity and clinical utility (Byron et al., 2016). Therefore, several issues need to be considered and investigated before peripheral blood cir-cRNA biomarkers can be translated into clinical practice. First, a blood circRNA-based gene test should prove its analytical validity within clinically relevant conditions. Although substantial advances have been made in the past several years (Szabo and Salzman, 2016;Kristensen et al., 2019), the methods to discover and profile circRNAs are far from optimal. Future studies need to test the analytical performance of different circRNA profiling methods in clinical blood samples, such as RNA-seq, circRNA microarray, reverse transcription quantitative PCR (RT-qPCR) and RT-ddPCR (Kristensen et al., 2019). In estimating analytical sensitivity and specificity, reference standards that can be specifically applied in circRNA discovery and profiling are needed (Hardwick et al., 2017). Furthermore, the procedure to discover and validate blood circRNA biomarkers needs to be standardized, including blood collection and preservation, circRNA extraction, library construction, and computational analysis (Byron et al., 2016;Anfossi et al., 2018). With the use of a standardized procedure, future studies need to estimate the technical robustness and reproducibility of the proposed biomarkers within and between laboratories. Second, the blood circRNA biomarkers identified in current studies (Tables 1 and 2) are only preliminary biomarker signatures for human diseases. The designed experiments in most studies are case-control studies of samples with well-defined phenotypes, and the sample size is relatively small. To test their clinical validity, more clinical samples are required to validate their sensitivity and specificity in a larger Guoxia Wen et al. cohort, especially their performance in discriminating patients with similar clinical phenotypes. Moreover, their performance in diagnosing the disease or predicting the disease outcome also needs to be tested in a prospective cohort in a clinical practice setting. Finally, further studies are needed to test and validate the usefulness of these biomarkers in clinical practice, such as their ability to inform clinical decisions and improve outcomes. Although many challenges and problems need to be solved, the promising potential of translating blood circRNA biomarkers into the clinic brings us new and inspiring options for liquid biopsy.