Your privacy, your choice

We use essential cookies to make sure the site can function. We also use optional cookies for advertising, personalisation of content, usage analysis, and social media.

By accepting optional cookies, you consent to the processing of your personal data - including transfers to third parties. Some third parties are outside of the European Economic Area, with varying standards of data protection.

See our privacy policy for more information on the use of your personal data.

for further information and to change your choices.

You are viewing the site in preview mode

Skip to main content

Circular RNA landscape in extracellular vesicles from human biofluids

Abstract

Background

Circular RNAs (circRNAs) have emerged as a prominent class of covalently closed single-stranded RNA molecules that exhibit tissue-specific expression and potential as biomarkers in extracellular vesicles (EVs) derived from liquid biopsies. Still, their characteristics and applications in EVs remain to be unveiled.

Methods

We performed a comprehensive analysis of EV-derived circRNAs (EV-circRNAs) using transcriptomics data obtained from 1082 human body fluids, including plasma, urine, cerebrospinal fluid (CSF), and bile. Our validation strategy utilized RT-qPCR and RNA immunoprecipitation assays, complemented by computational techniques for analyzing EV-circRNA features and RNA-binding protein interactions.

Results

We identified 136,327 EV-circRNAs from various human body fluids. Significantly, a considerable amount of circRNAs with a high back-splicing ratio are highly enriched in EVs compared to linear RNAs. Additionally, we discovered brain-specific circRNAs enriched in plasma EVs and cancer-associated EV-circRNAs linked to clinical outcomes. Moreover, we demonstrated that EV-circRNAs have the potential to serve as biomarkers for evaluating immunotherapy efficacy in non-small cell lung cancer (NSCLC). Importantly, we identified the involvement of RBPs, particularly YBX1, in the sorting mechanism of circRNAs into EVs.

Conclusions

This study unveils the extensive repertoire of EV-circRNAs across human biofluids, offering insights into their potential as disease biomarkers and their mechanistic roles within EVs. The identification of specific circRNAs and the elucidation of RBP-mediated sorting mechanisms open new avenues for the clinical application of EV-circRNAs in disease diagnostics and therapeutics.

Background

Circular RNAs (circRNAs) have emerged as a prominent class of covalently closed single-stranded RNA molecules, primarily derived from protein-coding genes through back-splicing events mediated by the canonical splicing machinery. Human cells produce a significantly greater number of unique circRNAs compared to protein-coding genes. While most circRNAs are expressed at low levels in most tissues, specific highly abundant and conserved circRNAs play essential physiological and/or pathophysiological roles [1]. Furthermore, circRNAs exhibit tissue-restricted and disease-specific expression patterns, indicating their potential clinical relevance and utility [2]. Notably, circRNAs have shown promise as diagnostic, prognostic, and predictive biomarkers, as they can be detected in extracellular vesicles (EVs) obtained from liquid biopsy samples, such as plasma and urine [3,4,5]. The identification of a substantial number of circRNAs in EVs represents a significant advancement in circRNA biology and offers new avenues for research and potential applications. Nevertheless, a comprehensive understanding of the characteristics and applications of circRNAs in EVs remains to be unveiled.

EVs, which are lipid bilayer-enclosed nanosized vesicles derived from various cell types, serve as crucial mediators of intercellular communication [6]. They are abundant in different biological fluids and hold promise as biomarkers in liquid biopsy, offering a minimally invasive approach to disease diagnosis and monitoring [7, 8]. EVs function through their transmembrane molecules and diverse internal cargos, including nucleic acids, proteins, and lipids [9]. Notably, circRNAs exhibit selective packaging into EVs and can be transferred to neighboring and distant recipient cells [10]. Recent evidence highlights the role of circRNAs in EVs in regulating disease development, progression, and serving as disease biomarkers [11,12,13]. However, existing studies primarily focus on identifying individual circRNAs in EVs from specific cancer types or small cohorts.

In this study, we conducted a comprehensive analysis of circRNAs derived from EVs in various human body fluids, referred to as EV-circRNAs, with a focus on their characterization and potential clinical applications. We identified a total of 136,327 EV-circRNAs from 1082 EV samples derived from plasma, urine, cerebrospinal fluid (CSF), and bile. EVs exhibited a considerable abundance of circRNAs characterized by a high back-splicing ratio. Moreover, we identified a significant number of cancer-associated circRNAs within EVs that were highly expressed and associated with clinical outcomes. Importantly, our results revealed the involvement of RNA-binding proteins (RBPs) in the sorting of circRNAs into EVs.

Methods

EV collection and RNA-sequencing

Extracellular vesicles (EVs) were isolated from a comprehensive collection of human biofluids and supernatants from five cell lines. Detailed information on the biofluids is provided in Additional file 1: Table S1. In summary, the dataset consists of four types of human biofluids: plasma samples (n = 1031), urine samples (n = 28), bile samples (n = 19), and cerebrospinal fluid samples (n = 4). Specifically, for plasma samples, the dataset includes 707 samples from 11 disease types: breast cancer (n = 141), colon cancer (n = 74), hepatocellular carcinoma (n = 92), pancreatic adenocarcinoma (n = 150), ovarian cancer (n = 51), gastric cancer (n = 9), kidney cancer (n = 15), malignant lymphoma (n = 42), non-small cell lung cancer (n = 55), small cell lung cancer (n = 67), and coronary heart disease (n = 11). Additionally, 172 samples were collected from patients with benign conditions (including benign breast (n = 19), ovarian (n = 28), pancreatic (n = 49), colorectal (n = 43), and hepatic (n = 33) conditions), and 152 samples from healthy individuals. The samples were collected following standardized protocols and were approved by the Institutional Review Board of Fudan University Shanghai Cancer Center.

The isolation process commenced with the initial centrifugation of the biofluids at 3000 g for 10 min at room temperature (25 °C), followed by a further purification step at 13,000 g for 15 min at 4 °C. Subsequently, the supernatants were treated with a binding buffer and transferred onto an exoRNeasy affinity membrane centrifuge column (Qiagen, Hilden, Germany). After the EVs were bound to the membrane, they were cleaved and purified with QIAzol (Qiagen, Hilden, Germany). Finally, all EV-RNAs were separated and purified with the miRNeasy (RNeasy MinElute spin column).

After the extraction of EV RNAs, any adulterated DNAs were removed by deoxyribonuclease I (DNase I, NEB, Ipswich, Massachusetts, USA). Then, the strand-specific RNA-seq library was prepared using SMARTer Stranded Total RNA-Seq Kit—Pico Input Mammalian (Clontech, USA). The purified double-stranded DNA was amplified by 13–16 cycles of PCR. Finally, the quality of the library was assessed using Qubit (Thermo Fisher Scientific, USA) and Qsep100 (BiOptic Inc.), and sequencing was performed on the Illumina sequencing platform (San Diego, California, USA) using 150-bp paired-end sequencing.

Detailed methodologies for the isolation of EV RNAs, characterization of the isolated EVs, and preparation of the EV RNA-seq library were provided in the supplementary materials (Additional file 2) and can also be referenced in our previously published work [14].

Identification of EV-circRNAs in EVs

After the raw data underwent quantity control, EV-circRNAs were identified using two software, CIRI2 (V2.0.6) [15] and ASJA [16]. For the CIRI2, the sequences were aligned to hg38 using BWA-MEM (V0.7.17) with default settings, and then EV-circRNAs were identified using CIRI2.pl with the GENCODE v29 annotation file. Information regarding the expression count, host gene, circRNA types, and genome origin was obtained. ASJA also performed with default setting. Briefly, the chimeric genome was aligned using STAR (V2.5.3a) [17] in a two-pass mapping approach to GRCh38. The alignment was performed with the parameters "–chimOutType WithinBAM" and "–chimSegmentMin 20" to detect chimeric alignments and generate a Chimeric.out.junction file. To identify back-splicing junctions, we utilized the ASJA-All.pl script. Additionally, ASJA-All.pl allowed us to retrieve exon length and number information.

To ensure the reliability of the identified EV-circRNAs, candidate EV-circRNAs needed to be detected by both CIRI2 and ASJA identification methods. Furthermore, for inclusion in our analysis, EV-circRNAs had to be present in at least three plasma samples or detected at least once in other human biofluids, including urine, CSF, or bile. The expression levels of these EV-circRNAs were normalized to counts per million (CPM) [5], using the counts obtained from CIRI2. To ensure consistent levels of EV-circRNA expression across samples within each cohort, we introduced a filtering step to remove samples exhibiting a Pearson correlation coefficient lower than 0.4 among their EV-circRNA expression profiles in each cohort. In total, we identified 136,327 EV-circRNAs from 1082 EV samples.

Back-splicing junction ratio of EV-circRNAs

The back-splicing junction ratio of EV-circRNA can be calculated using the formula (2 * #junction_reads / (2 * #junction_reads + #non_junction_reads)), as described in the CIRI2 manual. To compare with cellular circRNA in tissue, we downloaded the raw data from five studies [18,19,20,21] (GSE144269, GSE77661, GSE142279, GSE172032, GSE221240), encompassing a total of 196 samples. The tissue cohort includes normal tissue, cancer tissue, and matched adjacent tissue (Additional file 1: Table S3). Only the circRNAs detected by both CIRI2 and ASJA methods, and expressed in at least two samples were considered reliable. Their back-splicing junctions were also calculated using the same method as described above. To compare the circRNA back-splicing ratio between different groups, we calculated the median value of the circRNA back-splicing ratio in each group as a representative measure (related to Figs. 2, 3, 4, and 5).

Gene ontology and KEGG enrichment

The function of host gene of EV-circRNA was enriched by DAVID [22], including biological process (BP), cellular component (CC), and Kyoto Encyclopedia of Genes and Genomes (KEGG).

Differential expression EV-circRNAs

A strategy akin to that employed in prior studies was utilized to estimate the specific expression of EV-circRNAs. Specifically, EV-circRNA with frequency greater than 30% in only one cancer type were defined as cancer-specific, while those with frequency greater than 30% in more than two cancer types were defined as pan-cancer specific. Note, that the frequency of these EV-circRNAs in plasma healthy samples was less than 10%.

To identify reliable differential expression of EV-circRNA between cancer patients and healthy samples, we performed a Wilcoxon rank-sum test and fold change analysis using EV-circRNA with an average expression greater than 0.1. EV-circRNA with a fold change of 1.5-fold (FC ≥ 1.5 or FC ≤ 0.67), a P-value less than 0.01, and a frequency greater than 5% were defined as differential EV-circRNA.

For the specific analysis of biomarkers related to NSCLC immunotherapy response, clinical response data were obtained from our recent study [23]. Out of the 55 samples in the NSCLC cohort, 36 samples had available clinical information on immunotherapy response. These 36 samples were divided into two groups: 16 classified as responders and 20 as non-responders. This subset of samples, referred to as the "NSCLC immunotherapy group," was used for the biomarker analysis. EV-circRNAs with an average expression greater than 0.1 in the immunotherapy group were selected for downstream analysis. Differentially expressed EV-circRNAs were identified using the Wilcoxon test and fold change (median). For PFS and OS survival analysis, the optimal cutoff was determined using the surv_cutpoint function, and the Kaplan–Meier analysis was performed using the survfit and surv functions with the log-rank test. We identified 20 candidate EV-circRNAs as biomarkers for NSCLC immunotherapy based on the following criteria: First, EV-circRNAs were filtered by differential expression analysis with a p-value less than 0.05, higher expression in the response group, and an AUC (calculated by pROC package) greater than 0.65, which is higher than the AUC of the clinical biomarker NLR (0.622). Among these filtered EV-circRNAs, we then selected those meeting at least one of the following conditions: (1) OS analysis with a p-value less than 0.05 and expression in more than 10 responders (10/16); (2) PFS analysis with a p-value less than 0.05 and expression in more than 10 responders (10/16); (3) Exclusive expression in the response group, with expression in more than 6 samples. For training the Lasso model, we included all 36 samples from the NSCLC immunotherapy group. We first obtained all pairwise combinations of the 20 candidate EV-circRNAs identified above. For each pair of EV-circRNAs, we used the cv.glmnet function (nfolds = 10, family = "binomial") to determine the best lambda for the Lasso model. The glmnet function, with “binomial” and best lambda parameters, was then used to establish the model. AUC was used to assess predictive performance. For internal validation of the model, we randomly selected 60% of the samples (22/36) from the NSCLC immunotherapy group 50 times and calculated the AUC for each iteration. Additionally, we collected another cohort for external validation, consisting of 17 responders and 19 non-responders, to validate the model’s performance using AUC. qPCR was performed on 57 samples (32 non-responders and 25 responders) to validate these findings. Single EV-circRNA and Lasso models were also established based on the qPCR results of EV-circRNAs.

EV-circRNAs as prognosis biomarkers

For the prognosis biomarker analysis, we retained EV-circRNA with an average expression greater than 0.1 and frequency greater than 10%. Patients were divided into two groups based on median expression, and the survfit and surv functions were used to compare survival analysis. The coxph function was used to obtain the risk and protective factors. Potential biomarkers were defined as EV-circRNAs that expressed and had a p-value of less than 0.05 for at least ten patients in the high-expression group.

Motif of EV-circRNAs

The top 1% (1363) most highly frequency expressed EV-circRNAs were analyzed by retrieving their genome sequences using samtools faidx. We incorporated 119 motifs from previous studies by Ray et al. [24] and Fabbiano et al. [25], which were then converted to the strand input format of MEME using iupac2meme. The resulting sequences were analyzed using online MEME SEA (online version 5.5.3) [26] with default settings.

RBP regulated EV-circRNAs into EVs

To further validate the binding between RBPs and EV-circRNA, we obtained the bed file for RBP CLIP-seq data from ENCODE (K562) [27] and GEO (GSE133620 [28] and GSE39682 [29]). Additionally, we downloaded a list of proteins expressed in EVs from the study by Kugeratski et al. [30]. We utilized RBP CLIP-seq data only when its corresponding protein was expressed in EVs. The binding site was defined as the overlap between EV-circRNA and RBP CLIP-seq data using bedtools intersect (V2.30.0) [31].

To compare with tissue-derived circRNA, we randomly selected 1363 cellular circRNAs in a tissue cohort with similar exon length and number distribution to EV-circRNA, and this step was repeated 100 times. Additionally, we conducted an analysis comparing the top 1% of EV-circRNAs with a subset of highly expressed tissue circRNAs. This subset consists of the top 1000 circRNAs, ranked by their expression levels (CPM) and occurrence frequency. The number of binding sites between EV-circRNAs and tissue cellular circRNAs was calculated using the paired Wilcoxon test.

To annotate the position of RBP binding on EV-circRNA, we normalized the binding site positions from 0 to 1, based on the genome position of the corresponding EV-circRNA.

Brain-derived circRNA enriched in plasma EVs

To investigate the origin of EV-circRNAs in healthy plasma EVs, we downloaded the human normal tissue gene expression matrix from GTEx [32] (GTEx_Analysis_2017-06-05_v8_RNASeQCv1.1.9_gene_tpm.gct.gz) to obtain tissue-specific genes. We calculated the row mean of each tissue and filtered out those with values less than 0.1 for downstream analysis. Using the same method as our previous study [14], we calculated the Ti-score and retained only the tissue-specific genes with a Ti-score greater than 1 and a relative frequency greater than twice that of the second-most frequent tissue. We removed whole blood and testis, and merged brain, nerve, and pituitary as one category “Brain”. The EV-circRNA was classified as originating from the tissue if its host gene is a tissue-specific gene. To obtain more accurate tissue-specific EV-circRNAs, we queried each EV-circRNA in CircAtlas3.0 [33] to confirm its tissue specificity. The final results were determined by selecting circRNAs that maintained consistent tissue specificity according to CircAtlas3.0. To calculate the Brain-derived circRNAs score, we first obtained the average expression of brain-derived circRNAs excluding zero values. Subsequently, we transformed this average expression to the log2 scale.

To compare plasma EV-circRNAs with other cellular components in blood, we obtained cell-type markers from the TISCH database [34]. The EV-circRNA is classified as originating from the blood cell type if its host gene is a cell type-specific gene.

Cell culture and treatments

HuH7 cells were obtained from the Japanese Collection of Research Bioresources (JCRB, Tokyo, Japan). The HepG2, Hela, SK-Hep-1, and HEK-293 T cell lines were purchased from American Type Culture Collection (ATCC, Manassas, Virginia, USA) and cultured in DMEM supplemented with 10% FBS and Penicillin–Streptomycin at 37℃ with 5% CO2.

RT-qPCR

RNA was extracted using Trizol reagent (Life Technologies, Carlsbad, CA, USA), and the cDNA was synthesized using Evo M-MLV RT Master Mix (Accurate Biology, Hunan, China). SYBR Premix Ex Taq (Accurate Biology, Hunan, China) was used to measure the expression level of gene expression. The 2^ − ΔΔCt method was used to calculate each RNA’s relative expression level. The primers are provided in Additional file 1: Table S11.

RNA immunoprecipitation (RIP) assay

The cells were lysed with RIP lysis buffer (150 mM NaCl, 25 mM Tris–HCl (pH 7.4), 0.1% NP-40, 1 mM EDTA, 5% glycerol) at 4 °C for 1 h. After incubating the anti-FLAG antibody (Sigma-Aldrich, St. Louis, MO, USA) and immunoglobulin G antibody (Sigma-Aldrich, St. Louis, MO, USA) with protein G beads (Life Technologies, Carlsbad, CA, USA) for 1 h at room temperature. The supernatants were collected after centrifugation at 12,000 × g for 20 min. The antibody-bound beads were added to the supernatants and incubated overnight at 4 °C after being washed three times with NT2 buffer (150 mM NaCl, 1 mM MgCl2, 50 mM Tris–HCl (pH 7.4), 0.05% NP-40). The complexes were then lysed with Trizol reagent after the beads had been washed six times with NT2 buffer. Reverse transcribed FLAG-interacting RNAs were isolated. RNA binding to YBX1 was detected by qPCR. The normalization process uses a 1% input sample as a reference. Specifically, the calculated value is derived from 2^ − ΔCT, with ΔCT values determined using the formula: ΔCT = Ct(Sample) − Ct(1% input).

Plasmid construction

The coding sequence of YBX1 was amplified and then cloned into the laboratory modification plasmid pCDH-3*Flag. In order to confirm YBX1 regulation circRNA into EVs, we used CRISPR/Cas9 technology to knockdown YBX1. sgRNA of YBX1 (Additional file 1: Table S11) was cloned into lentiGuide-Puro vector (Addgene #52,963).

Lentiviral production and transduction

HEK-293 T cells were transfected with pCDH-3*Flag-YBX1, lentiGuide-puro-gRNA, and the packaging and envelope plasmids psPAX2 and pMD2.G. Virus particles were collected 48 h after transfection. SK-Hep-1 cells were infected with lentiGuide-puro-gRNA lentivirus and polybrene (Sigma-Aldrich, St.Louis, MO, USA).

Statistical analyses

All experimental data were expressed as mean ± standard deviation (SD). P-values were obtained by using two-tailed Student’s t-test as indicated in corresponding figure legends. A p-value less than 0.05 was considered statistically significant and was noted by asterisks (*, p < 0.05; **, p < 0.01; ***, p < 0.001, ****, p < 0.0001).

Results

Characterization of circRNA atlas from 1082 EV samples by exLR-seq

To explore the circRNAs in EVs, we performed an optimized EVs long RNA-seq (exLR-seq) analysis [14] of four human body biofluids including plasma, urine, bile, and cerebrospinal fluid (CSF) (Additional file 1: Table S1, Fig. 1a). In addition, we investigated the circRNA profile in five cell lines (HepG2, Hela, SK-Hep-1, HuH-7, and MHCC-97L) and their secreted EVs. Using CIRI2 and ASJA programs, a total of 136,327 EV-circRNAs were identified with a stringent cutoff (Additional file 1: Table S2). These circRNAs included 134,197 in plasma, 66,694 in urine, 13,289 in bile, and 3363 in CSF (Fig. 1b). While the number of EV-circRNAs can vary in different physiological or pathological conditions, our findings indicate that, overall, cancer patients tend to exhibit a slightly higher number of EV-circRNAs compared to healthy donors (Additional file 2: Fig.S1a-b). We further compared the EV-circRNAs with the circRNAs in circBase [35] and CircAtlas3.0 [33], which encompass the circRNAs identified in multiple tissue types. Only 18% of these EV-circRNAs were reported in circBase (Fig. 1c). But the majority of EV-circRNAs (89.8%, 122,502/136,327) were found in CircAtlas3.0 (tissue-derived circRNAs) (Fig. 1d). This result revealed the reliability of EV-circRNAs identified in this study.

Fig. 1
figure 1

Characterization of circRNA atlas from 1082 EV samples by exLR-seq. a The figure represents the EV samples analyzed in our study, with the number in parentheses indicating the number of samples in each cohort. Extracellular vesicles, EVs; Breast cancer, BRCA; Colon cancer, CRC; Hepatocellular Carcinoma, HCC; Pancreatic Adenocarcinoma, PAAD; Ovarian cancer, OV; Coronary heart disease, CHD; Gastric cancer, GC; Kidney cancer, KIRC; Malignant lymphoma, ML; Non-small cell lung cancer, NSCLC; Small cell lung cancer, SCLC; Cerebrospinal fluid, CSF; Diabetes, D; Diabetic Nephropathy, DN. b Venn diagram showing the overlap of circRNAs detected in EVs from plasma, CSF, bile, and urine. c Overlap of circRNAs between EVs and circBase. d Overlap of circRNAs between EVs and tissue-derived circRNAs (circAltas 3.0). e Genomic origin of EV-circRNAs. f Distribution of circRNA exon numbers in cells (5 cell lines) and secreted EVs. g Exon length distribution of circRNAs in cells (5 cell lines) and secreted EVs.

When analyzing the distribution of circRNAs across the genome and their host genes in cell lines and their secreted EVs, we observed that 78.1% of the EV-circRNAs originated from exonic regions (Fig. 1e). More than 60.7% of the exonic EV-circRNAs in cellular EVs contained less than 3 exons, among which the proportion of exons with a length of 200–400 nt was the highest (mean 840 nt) (Fig. 1f,g). In contrast, the average length of intracellular circRNA was 972 nt, with a higher proportion containing more than 10 exons (Fig. 1f,g). Compared to tissue-derived circRNAs, EV-circRNAs also have fewer exons and shorter lengths (Additional file 2: Fig. S1c-d).

High back-splicing ratio of circRNAs in EVs

CircRNAs are predominantly generated through the back-splicing of pre-mRNAs, with their expression levels in cells typically being considerably lower than those of their linear RNA counterparts [36]. The back-splicing ratio can reflect the abundance of circRNAs relative to their linear host genes. Our analysis revealed that EV-circRNAs exhibited a higher back-splicing ratio, with more than 0.5 of the back-splicing ratio accounting for 30.9% (42,134/136,327) of the total EV-circRNAs. In contrast, this proportion was only 3.55% (2,149/60,417) in tissues (Fig. 2a). Similarly, when comparing cell lines and their secreted circRNAs, the results also indicated that the proportion of EV-circRNAs with a high back-splicing ratio was significantly higher than that in cells (Fig. 2b). Interestingly, 3437 EV-circRNAs were observed with a back-splicing ratio of 1 and high expression levels (median CPM > 50), suggesting that these circRNAs were exclusively present in EVs. Among these EV-specific circRNAs, more than 82.3% of the host genes belonged to protein-coding genes, while 13.1% were derived from long non-coding RNAs (Fig. 2c). By enrichment analysis, we found that the host genes of these EV-circRNAs were significantly enriched in processes related to ion transport, endocytosis, and exocytosis, with many of these genes being located on the cell membrane (Fig. 2d). These findings suggest that circRNAs are highly enriched in EVs compared to linear RNAs, and display selective sorting and function enrichment.

Fig. 2
figure 2

High back-splicing ratio of circRNAs in EVs. a Density plot showing the distribution of the back-splicing ratio between EVs and tissues cohort, the bar plot displays the percentage of circRNAs in different back-splicing ratio intervals. The black dashed line represents the median back-splicing ratio in tissue, while the red line represents the median back-splicing ratio in EVs. b Bar plot showing the percentage of circRNAs with high back-splicing ratio in 5 cell lines and their secreted EVs. c Distribution of host genes for EV-circRNAs with a back-splicing ratio of 1 and median CPM >50. d Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis for the host genes in Figure 2c

Prevalence of circSMARCA5 with high back-splicing ratio in EVs

To explore the potential of EV-circRNAs with a high back-splicing ratio as markers for EVs, we analyzed their expression levels and frequency within EVs. Our results demonstrated that EV-circRNAs with high back-splicing ratio were predominantly present in the group with high frequency (Fig. 3a), and these circRNAs were also highly expressed (Fig. 3b). Among the most highly expressed EV-circRNAs, circSMARCA5 was expressed in almost all samples (99.26%) and had significantly higher expression levels in EVs compared to their host gene (Fig. 3b,c). To further validate the differential expression of circSMARCA5 and its host gene in cell lines and their secreted EVs, we analyzed changes in back-splicing ratio based on exLR-seq (Additional file 2: Fig.S2, Fig. 3d) and performed RT-qPCR analysis (Fig. 3e). Our results confirmed that the expression of circSMARCA5 was significantly higher in EVs, whereas linear RNAs were highly expressed in cells (Fig. 3e). Furthermore, the expression levels of circSMARCA5 were significantly higher than its host gene in EVs from both plasma and urine samples, as determined by RT-qPCR (Fig. 3f). Collectively, the high back-splicing ratio of circSMARCA5 was prevalent in EVs and represents a potential marker for distinguishing EVs from other tissue/cell components.

Fig. 3
figure 3

Identification of circSMARCA5 as EV marker. a Percentage of EV-circRNAs in different frequency ranges. High: back-splicing ratio>0.8. b The expression (median CPM) and frequency of EV-circRNAs with a high back-splicing ratio. c IGV plots showing the reads coverage of circSMARCA5. d Boxplot showing the back-splicing ratio of circSMARCA5 between cells and their secreted EVs. Each point represents a cell line. Paired T-test, two-sided, ***p < 0.001. e Relative expression of circSMARCA5 and SMARCA5 in cells and their secreted EVs. Statistical analysis was performed using unpaired Student’s t-tests. *p < 0.05; **p < 0.01; ***p < 0.001, ****p<0.0001. f Relative expression of circSMARCA5 and SMARCA5 in plasma EVs (left) and urine (right) EVs. Each point represents a sample. Statistical analysis was performed using paired Student’s t-tests. **p < 0.01; ***p < 0.001

Enrichment of brain-specific circRNAs in plasma EVs

Multiple studies have proposed that extracellular RNA originates from diverse tissue cells, indicating that EV-circRNA encompasses a broad range of tissue sources [5, 10, 37,38,39]. To explore the tissue origin of EV-circRNAs, we analyzed the tissue-specific circRNAs using gene expression profiles from the Genotype-Tissue Expression Project (GTEx). We classified circRNAs as tissue-specific if their host genes exhibited tissue-specific in GTEx. Furthermore, we retained circRNAs that exhibited consistent tissue specificity with CircAtlas3.0 [33]. We identified a total of 230 tissue-specific EV-circRNAs from 14 different tissues. Notably, 84 brain-specific circRNAs are enriched in plasma EVs, with the highest number of host genes (43 genes) (Additional file 1: Table S4), suggesting a potentially high probability of circRNA secretion into EVs from brain tissue (Fig. 4a). Moreover, among these tissue-specific EV-circRNAs, brain-derived circRNAs showed a higher rate of back splicing (mean = 0.87) compared to other tissue sources (mean = 0.76), significantly higher than the overall healthy plasma EV-circRNAs (mean = 0.50) (Fig. 4b). Our analysis of the tissue specificity scores for these brain-derived circRNAs (84 EV-circRNAs) in plasma and CSF, calculated as the average of nonzero expression values transformed to log2 scale, revealed their higher expression burden in CSF compared to plasma (Fig. 4c). As EVs are gaining recognition for their role in the pathogenesis and progression of neurodegenerative diseases [40], we found age-related differences in these brain-specific EV-circRNAs, indicating an increased enrichment in EVs with age (Fig. 4d). Besides, our analysis showed that the majority of circRNAs (72%) identified in peripheral blood mononuclear cells (PBMCs) were also detected in plasma EVs, but with lower back-splicing ratios in PBMCs when contrasted with those found in EVs (Additional file 2: Fig.S3a-b). Moreover, by categorizing circRNAs according to their source blood cell types, we discovered that cell type-specific circRNAs from a range of blood cell types were present in EVs, notably with a greater representation from B cells and CD4 T cells (Additional file 2: Fig.S3c). These results highlight the complex and diverse landscape of circRNA profiles in EVs, providing a new approach for early detection and monitoring of brain-related diseases while also being able to capture a wide range of physiological and pathological states.

Fig. 4
figure 4

Enrichment of brain-specific circRNAs in plasma EVs. a Circular plot showing the distribution of tissue-specific circRNA in healthy plasma EVs, green presents the number of circRNA, grey presents the number of host genes of circRNA. b The plot illustrates the back-splicing ratio between brain-specific circRNAs, other tissue-specific circRNAs in healthy plasma EVs, and the overall healthy plasma EV-circRNAs. The error bars represent the mean ± standard error of the mean (SEM). Unpaired Wilcoxon. c The brain-derived circRNAs score in healthy plasma and CSF EVs. Unpaired Wilcoxon. d Brain-derived circRNAs score in healthy plasma EVs across different age ranges, excluding cases where the score was zero. Unpaired Wilcoxon

Distinctive expression patterns of EV-circRNAs in different types of cancers

We further explored the EV-circRNA expression in cancer patients by evaluating 9 different cancer types with a total of 687 samples. A substantial number of EV-circRNAs (median = 4877) expressed at a frequency higher than 30% in different tumor types (Fig. 5a). Of these EV-circRNAs, we identified 414 cancer-associated EV-circRNAs among the 9 tumor types, including 325 exhibiting cancer-specific expression patterns and 89 recurrently expressed in at least two tumor types (Additional file 1: Table S5, Fig. 5b). Moreover, the expression profiles of certain cancer-associated EV-circRNAs, such as BRCA, CRC, HCC, KIRC, SCLC, and OV, exhibited significant differences compared to other tumor types in EVs (Fig. 5c). We also conducted a differential expression analysis of EV-circRNAs by comparing them with plasma EVs from healthy samples. This analysis resulted in the identification of numerous differentially expressed EV-circRNAs across various tumor types (Fig. 5d). Among these, we identified 3921 EV-circRNAs that displayed overexpression across at least two different tumor types (Additional file 1: Table S6, Fig. 5e). Notably, seven EV-circRNAs exhibited high expression levels in at least seven types of tumors (Fig. 5f). These findings highlight the potential of these EV-circRNAs as valuable biomarkers for tumor diagnosis.

Fig. 5
figure 5

Distinctive expression patterns of EV-circRNAs in different types of tumors. a Overview of EV-circRNAs in 9 cancer types, showing the number of samples (top), EV-circRNAs (middle), and the number of EV-circRNAs with a frequency greater than 30% (bottom). b Heatmap displaying cancer-specific and pan-cancer specific EV-circRNAs. The red color represents EV-circRNAs which were detected in more than 30% of samples in one cohort. c Violin plot illustrating the expression (average CPM) of cancer-specific EV-circRNAs across different cohorts of cancer types in EVs, the numbers in parentheses represent the number of cancer-specific EV-circRNAs. d Up- and downregulated EV-circRNAs compared with healthy samples. e Overlap of upregulated EV-circRNAs in 9 cancer types, top 30 order by “degree” of R package UpSetR. f Expression fold change of 7 EV-circRNAs in 9 cancer types compared to healthy samples

EV-circRNAs as biomarkers for clinical outcomes

EV-circRNAs may serve as promising liquid biopsy biomarkers for cancer [37]. By survival analysis, we identified 5340 EV-circRNAs as risk factors, 5010 as protective factors, and 1882 as significantly correlated with overall survival (OS) in patients (Additional file 1: Table S7, Fig. 6a). Although many EV-circRNAs related to OS were shared among different cancer types, none were common to all cancer types (Fig. 6b). For example, circLTBP1, encoded by latent transforming growth factor-beta binding protein-1, is associated with poor survival in patients with BRCA, CRC, and SCLC who exhibit high circLTBP1 expression (Fig. 6c–e). Additionally, we explored the potential of EV-circRNAs for assessing immunotherapy effectiveness in a subset of the NSCLC cohort treated with immunotherapy, referred to as the “NSCLC immunotherapy group” (n = 36). Interestingly, we observed a higher back-splicing ratio of EV-circRNAs in the non-response group compared to the response group (Fig. 6f), suggesting that the generation of circRNAs or selective secretion of EVs may be related to the efficacy of immunotherapy and may have better performance than the neutrophil-to-lymphocyte ratio (NLR, AUC = 0.622) (Additional file 2, Fig.S4a), a known clinical indicator [41]. Therefore, we conducted differential EV-circRNA expression analysis, progression-free survival, and overall survival analysis, as well as prediction performance (AUC) for EV-circRNAs in the NSCLC immunotherapy group. We identified 20 EV-circRNAs with AUCs greater than 0.65 that were significantly related to survival (OS or PFS, p-value < 0.05) (Additional file 1: Table S8, Fig. 6g). Although the prediction performance of these EV-circRNAs was higher than that of the NLR, the performance of EV-circRNAs can still be improved. Therefore, we established a Lasso model using all pairwise combinations of the 20 EV-circRNAs and found that the model with circNFATC2 and circTBC1D22A achieved an AUC of 0.812 in the NSCLC immunotherapy group (Fig. 6h). The median AUC of the model reached 0.816 across 50 random iterations, each using 60% of the samples (22/36) from the NSCLC immunotherapy group for internal validation (Additional file 2, Fig.S4b). Furthermore, we collected another cohort for external validation cohort with 17 responders and 19 non-responders, the AUC reached 0.802 (Fig. 6h). Additionally, we performed qPCR on circNFATC2 and circTBC1D22A in 57 NSCLC immunotherapy samples (32 non-responders and 25 responders). The qPCR results also predicted immunotherapy response with accuracies of 0.716 and 0.699 for the two EV-circRNAs, respectively (Additional file 2, Fig.S4c). The model established using qPCR data for circNFATC2 and circTBC1D22A showed a significant increase in AUC, reaching 0.812 (Fig. 6i). Our findings indicate that EV-circRNAs have the potential to serve as liquid biopsy biomarkers for predicting patient outcomes and evaluating the efficacy of immunotherapy.

Fig. 6
figure 6

EV-circRNAs as biomarkers for clinical outcomes. a Circular plot displaying the number of EV-circRNAs as risk, protection, and prognosis factors. b Venn plot showing the overlap of EV-circRNAs related to OS in different cancer types (p <= 0.05). c–e The expression of circLTBP1 in EVs was related to OS in BRCA, CRC, and SCLC. f A comparison between the non-response and response groups in back-splicing ratios of EV-circRNA. Unpaired Wilcoxon, ***p<0.001. g Heatmap showing the 20 EV-circRNAs that can predict NSCLC immunotherapy response. h AUC of the Lasso model using RNA-seq expression of circTBC1D22A and circNFATC2 in EVs to predict response in the NSCLC immunotherapy group and validation cohort. i AUC of the Lasso model using qPCR relative expression of circTBC1D22A and circNFATC2 in EVs to predict NSCLC immunotherapy response

RNA binding proteins (RBPs) as regulators of enrichment of EV-circRNAs

To investigate the potential regulators of circRNAs sorting into EVs, we analyzed the interaction of circRNAs and RBPs, which have been reported as essential factors in sorting miRNAs into EVs [25, 42]. We observed that the sequences of highly expressed EV-circRNAs had a higher AT content compared to the GC content (Fig. 7a). By using MEME SEA [26], we identified many RBP-related motifs enriched across circRNA sequences (Additional file 1: Table S9, Additional file 2: Fig.S5, Fig. 7b). To further demonstrate whether EV-circRNAs are more likely to bind with RBPs than cellular circRNAs, we analyzed CLIP-seq data for 47 RBPs expressed in EVs. To reduce the potential influence of differences in the length and number of exons between EV and cellular circRNAs, we randomly selected cellular circRNAs (tissue cohort) with similar length and exon distribution to EV-circRNAs 100 times. Additionally, we also compared the top 1% of EV-circRNAs against a subset of highly expressed tissue circRNAs (Fig. 7c). The results showed that there were many RBP-binding sites on EV-circRNAs (median = 925) (Fig. 7d), and the binding sites for single-exon EV-circRNAs were enriched in any region of the exon, while those for multi-exon EV-circRNAs were mostly enriched in splicing sites (Fig. 7e). Importantly, compared to cellular circRNAs, RBP-binding sites were more enriched on EV-circRNAs (~ two fold) (Fig. 7f, Additional file 2: Fig.S6a). We further validated the interaction of EV-circRNA and YBX1, one of the top RBPs identified in this study. We observed that EV-circRNAs contain motifs that bind with YBX1, and the CLIP-seq results showed that there were more YBX1-binding sites in EVs (~ two fold) (Additional file 1: Table S10, Fig. 7g). By using the RNA immunoprecipitation (RIP) assay, we confirmed the interaction between YBX1 and several circRNAs, including circSMARCA5 (Fig. 7h). Furthermore, expression of these EV-circRNAs was significantly decreased in the EVs from YBX1 knockdown cells, contrasting with the unchanged levels of linear RNAs (Fig. 7i). Comparing the expression of linear and circRNAs in cells, we demonstrated the knockout of YBX1 primarily affected the abundance of circRNAs within cells, subsequently influencing circRNA levels in EVs. Notably, most linear RNAs (except for MYO9B) corresponding to these circRNAs exhibited negligible changes after knockdown of YBX1 expression (Additional file 2: Fig.S6b). These results revealed a significant association between YBX1 and the packaging of circRNAs into EVs, indicating that RBPs can regulate the sorting of circRNAs into EVs.

Fig. 7
figure 7

RBPs as regulators of enrichment of EV-circRNAs. a Percentage of GC and AT in exon of EV circRNA of Top 1% (ranked by frequency). Unpaired Wilcoxon, ***p<0.001. b The top 3 motifs, ranked by p-value, identified through MEME SEA enrichment. c Workflow of identification of RBP binding sites for circRNAs in tissues and EVs. d The network showing the RBP binding with EV-circRNAs by Gephi. e Position distribution of RBP binding sites in EV-circRNAs. The horizontal axis was the length of the normalized EV-circRNA. f The number of binding sites for circRNA and RBP in EVs and tissue. Each point represents the RBP. Paired Wilcoxon test, ***p < 0.001. g The density plot represents the number of binding sites of YBX1 on circRNAs in tissue across 100 random iterations, using two different strategies. The dashed line represents the number of binding sites between YBX1 and the top 1% of EV-circRNAs, as well as highly expressed circRNAs in tissue circRNAs (ranked by CPM and frequency). h The association of the YBX1 with EV-circRNAs was tested by RIP analysis in SK-Hep-1 cells infected with pCDH-3*Flag-YBX1 lentivirus. YBX1 as positive control and FAM99B as negative control. The normalization process uses a 1% input sample as a reference (Normalized to 1%Input). The calculated value is derived from 2^−ΔCT, with ΔCT values determined using the formula: ΔCT = Ct(Sample) − Ct(1% input). i RT‐qPCR analysis of circRNAs and their host linear RNAs in YBX1-knockdown SK-Hep1 cell-derived EVs

Discussion

In this study, we explored the intricate landscape of EV-circRNAs by analyzing 1082 EV samples from diverse human body fluids, including plasma, urine, CSF, and bile. Our investigation led to the identification of 136,327 high-confidence EV-circRNAs. Notably, we observed a pronounced enrichment of EV-circRNAs exhibiting a high back-splicing junction ratio within EVs. Among these, circSMARCA5 emerged as a prominent potential EV biomarker, distinguished by its elevated expression, frequency, and back-splicing junction ratio. Furthermore, our research shed light on the presence and enrichment of brain-specific EV-circRNAs within plasma EVs, suggesting their potential as novel biomarkers for neurological conditions. In the realm of oncology, our study delved into the expression patterns of EV-circRNAs in several cancer patients. This led to the identification of diverse expression profiles and clinically relevant EV-circRNA responses, opening new avenues for their application in clinical settings. Additionally, we explored the interaction between RBPs and circRNAs in EVs. Our findings revealed a significant association between YBX1 and the packaging of circRNAs into EVs, offering insights into the underlying mechanisms governing circRNA enrichment in EVs. Overall, our study provides a comprehensive analysis of EV-circRNAs, revealing their characteristics, the mechanisms behind their enrichment, and their potential clinical applications (Fig. 8). This research not only enriches our understanding of circRNA biology but also paves the way for future explorations into their clinical applications.

Fig. 8
figure 8

Overview of the characterization and potential clinical applications of EV-circRNAs. EV-circRNAs were identified in a variety of human body fluids, and their main characteristics and potential clinical applications were summarized

EVs are complex structures containing proteins, lipids, RNA, and DNA, necessitating the selection of appropriate methods based on research needs. In our study, we used the exoEasy kit to collect EVs. Despite the lower number of EV proteins, the RNA profiles of the isolated EVs were similar to others and showed comparable or even higher RNA yields [43]. Additionally, the kit demonstrated enrichment for common EV-enriched tetraspanins by flow cytometry and exhibited a classical EV-like RNA pattern [44]. However, careful consideration of EV purity is essential. We speculate that intact circRNAs are present within EVs; however, we cannot rule out the possibility that some fragmented RNAs, encapsulated by proteins, may be present outside EVs. This is because the exoEasy kit is capable of capturing protein particles in plasma alongside EVs.

CircRNAs are typically expressed in a tissue-specific and even a cell type-specific manner [2]. Numerous circRNAs exhibit differential expression in tumors compared to adjacent non-malignant tissues, and they are associated with clinical outcomes. Such findings highlight circRNAs as promising diagnostic and prognostic biomarkers. Particularly, these circRNAs could be sorted into EVs and detected in biofluids with high stability. In this study, we found many clinically relevant EV-circRNAs with cancer-specific expression patterns. On the other hand, PD-L1 expression, tumor mutational burden (TMB) [45], and microsatellite instability (MSI)/mismatch repair deficiency (dMMR) are being considered predictive biomarkers of lung cancer immune therapy response; however, NSCLC patients with low PD-L1 expression or TMB can still benefit from PD-1 inhibitors combined with chemotherapy. Therefore, there is a need to explore novel biological markers for effective immunotherapy strategies. Interestingly, through analyzing the EVs of NSCLC patients treated with PD-1 inhibitors, we found that non-responding patients had higher levels of back-splicing, suggesting changes in the proportion of circular and linear RNAs in their EVs. This phenomenon suggests that these abnormally excreted circRNAs may play roles in receptor cells to counteract the effects of drugs or as potential molecular markers for drug resistance. We also detected brain-specific circRNAs in healthy plasma EVs. Acquiring brain tissue can be a difficult task, and numerous substances that are specific to the brain cannot cross the blood–brain barrier directly, resulting in challenges for both prevention and diagnosis. However, EVs can carry cargo that freely shuttles between brain tissue and other tissues [46]. Studies have shown that some neuron-specific proteins, such as tau andα-synuclein [47, 48], which were highly correlated with Parkinson’s disease and Alzheimer’s disease, can be detected in plasma EVs, providing a new insight for non-invasive diagnosis of brain or neurological diseases. In our study, we found that brain-specific circRNAs can be detected and relatively enriched in plasma EVs, such as CDR1as circRNA (chrX:140,783,380–140,784,563). Compared to linear genes, circRNAs are more stable, suggesting that circRNAs in EVs have broad application prospects in the prevention and diagnosis of brain diseases.

Most circRNAs are derived from protein-coding genes and are produced by back-splicing events carried out by the canonical splicing machinery. The efficiency of back-splicing is much lower than that of canonical splicing [36], resulting in a generally low abundance of circRNAs in most cells and tissues. However, we observed that circRNAs are enriched in EVs with high back-splicing ratio, indicating the regulation of circRNAs sorting into EVs. Both endosomal and plasma membrane-derived EVs are capable of selectively capturing cargoes during their biogenesis. The selection of cargoes by transmembrane receptors through ubiquitination and ESCRT recognition is well-established in the biogenesis of EVs. RNA-binding proteins (RBPs) and their partners can direct RNA to the location of EV production while protecting it from destruction. Increasing evidence suggests an active RNA packaging process involving the sorting of coding and non-coding RNAs by RBPs [25]. For instance, HNRNP2AB1 regulates the miRNAs into endothelial cell-secreted EVs [49], while ALYREF and FUS are involved in the entry of miRNAs with the “CGGGAG” motif into exosomes [42]. Most recently, a study demonstrated that IGF2BP1 (RBP) that binds directly to circular RNAs recruits Ran-GTP and exportin-2 to export circRNAs [50]. Our study found that EV-circRNAs had more RBP-binding sites compared to cellular circRNAs by motif and CLIP data analysis of RBPs. We further validated that one of the RBPs, YBX1, could interact with circRNAs and affect their sorting into EVs. These findings suggest that RBPs are involved in the sorting of circRNAs into EVs. Although we have identified RBPs as crucial factors influencing the differential cargo content between EVs and cells, it is important to note that the cargo composition of EVs varies based on the originating cell type. While we have employed bioinformatics strategies to analyze the source of circRNAs in EVs, it is advantageous to obtain single EVs, as this facilitates a better understanding of EVs’ function and applications. For instance, He et al. employed the SEVtras algorithm to identify sEV-containing droplets from scRNA-seq, and the sEV secretion activity (ESAI) proved to be a robust indicator of tumor progression, particularly in the early stages [51]. Thus, the mechanism of selectively capturing circRNAs during EV biogenesis by RBPs from different cell types remains to be explored.

Conclusions

In summary, our study provided a comprehensive landscape of EV-circRNAs, characterizing their back-splicing ratio, which can serve as a factor for identifying EV markers such as circSMARC5. We found that circRNAs derived from the brain were enriched in healthy plasma EVs. We also demonstrated that EV-circRNAs have the potential to serve as biomarkers for evaluating the immunotherapy efficacy in NSCLC. Moreover, our research demonstrated the involvement of YBX1 in the sorting and entry of circRNAs into EVs. These findings emphasized the potential of EV-circRNA as diagnostic targets and enhanced our understanding of the mechanism behind the selective packaging of circRNAs into EVs.

Data Availability

Raw data of the EV-RNA sequencing data is available in the National Center for Biotechnology Information (NCBI) under accession PRJNA1012336 [52] and PRJNA1010887 [53]. The count, CPM (counts per million), and back-splicing ratio data for the expression profile of EV-derived circRNAs can be accessed through the figshare [54].

Raw data of the EV-RNA sequencing data can be accessed from NCBI with accession numbers PRJNA1012336 and PRJNA1010887. The count, CPM (Counts Per Million), and back-splicing ratio data for the expression profile of EV-derived circRNAs can be accessed through the figshare (https://doi.org/10.6084/m9.figshare.25559121.v1).

Abbreviations

EVs:

Extracellular vesicles

EV-circRNA:

CircRNA in EVs

BRCA:

Breast cancer

CRC:

Colon cancer

HCC:

Hepatocellular carcinoma

PAAD:

Pancreatic adenocarcinoma

OV:

Ovarian cancer

CHD:

Coronary heart disease

GC:

Gastric cancer

KIRC:

Kidney cancer

ML:

Malignant lymphoma

NSCLC:

Non-small cell lung cancer

SCLC:

Small cell lung cancer

CSF:

Cerebrospinal fluid

D:

Diabetes

DN:

Diabetic nephropathy

RBPs:

RNA-binding proteins

References

  1. Liu CX, Chen LL. Circular RNAs: Characterization, cellular roles, and applications. Cell. 2022;185(12):2016–34. https://doi.org/10.1016/j.cell.2022.04.021.

    Article  CAS  PubMed  Google Scholar 

  2. Kristensen LS, Jakobsen T, Hager H, Kjems J. The emerging roles of circRNAs in cancer and oncology. Nat Rev Clin Oncol. 2022;19(3):188–206. https://doi.org/10.1038/s41571-021-00585-y.

    Article  CAS  PubMed  Google Scholar 

  3. Wang Y, Liu J, Ma J, Sun T, Zhou Q, Wang W, et al. Exosomal circRNAs: biogenesis, effect and application in human diseases. Mol Cancer. 2019;18(1):116. https://doi.org/10.1186/s12943-019-1041-z.

    Article  CAS  PubMed  Google Scholar 

  4. Li Y, Zheng Q, Bao C, Li S, Guo W, Zhao J, et al. Circular RNA is enriched and stable in exosomes: a promising biomarker for cancer diagnosis. Cell Res. 2015;25(8):981–4. https://doi.org/10.1038/cr.2015.82.

    Article  CAS  PubMed  Google Scholar 

  5. . Lai H, Li Y, Zhang H, Hu J, Liao J, Su Y, et al. exoRBase 2.0: an atlas of mRNA, lncRNA and circRNA in extracellular vesicles from human biofluids. Nucleic Acids Res. 2022;50(D1):D118-D28. https://doi.org/10.1093/nar/gkab1085

  6. Raposo G, Stahl PD. Extracellular vesicles - on the cusp of a new language in the biological sciences. Extracell Vesicles Circ Nucl Acids. 2023;4(2):240–54. https://doi.org/10.20517/evcna.2023.18.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Yu W, Hurley J, Roberts D, Chakrabortty SK, Enderle D, Noerholm M, et al. Exosome-based liquid biopsies in cancer: opportunities and challenges. Ann Oncol. 2021;32(4):466–77. https://doi.org/10.1016/j.annonc.2021.01.074.

    Article  CAS  PubMed  Google Scholar 

  8. Asao T, Tobias GC, Lucotti S, Jones DR, Matei I, Lyden D. Extracellular vesicles and particles as mediators of long-range communication in cancer: connecting biological function to clinical applications. Extracell Vesicles Circ Nucl Acids. 2023;4(3):461–85. https://doi.org/10.20517/evcna.2023.37.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  9. Maas SLN, Breakefield XO, Weaver AM. Extracellular Vesicles: Unique Intercellular Delivery Vehicles. Trends Cell Biol. 2017;27(3):172–88. https://doi.org/10.1016/j.tcb.2016.11.003.

    Article  CAS  PubMed  Google Scholar 

  10. Li Z, Zhu X, Huang S. Extracellular vesicle long non-coding RNAs and circular RNAs: Biology, functions and applications in cancer. Cancer Lett. 2020;489:111–20. https://doi.org/10.1016/j.canlet.2020.06.006.

    Article  CAS  PubMed  Google Scholar 

  11. Yang SJ, Wang DD, Zhong SL, Chen WQ, Wang FL, Zhang J, et al. Tumor-derived exosomal circPSMA1 facilitates the tumorigenesis, metastasis, and migration in triple-negative breast cancer (TNBC) through miR-637/Akt1/beta-catenin (cyclin D1) axis. Cell Death Dis. 2021;12(5):420. https://doi.org/10.1038/s41419-021-03680-1.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Li Y, Hu J, Wang M, Yuan Y, Zhou F, Zhao H, et al. Exosomal circPABPC1 promotes colorectal cancer liver metastases by regulating HMGA2 in the nucleus and BMP4/ADAM19 in the cytoplasm. Cell Death Discov. 2022;8(1):335. https://doi.org/10.1038/s41420-022-01124-z.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Zhang F, Jiang J, Qian H, Yan Y, Xu W. Exosomal circRNA: emerging insights into cancer progression and clinical application potential. J Hematol Oncol. 2023;16(1):67. https://doi.org/10.1186/s13045-023-01452-2.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Li Y, Zhao J, Yu S, Wang Z, He X, Su Y, et al. Extracellular Vesicles Long RNA Sequencing Reveals Abundant mRNA, circRNA, and lncRNA in Human Blood as Potential Biomarkers for Cancer Diagnosis. Clin Chem. 2019;65(6):798–808. https://doi.org/10.1373/clinchem.2018.301291.

    Article  CAS  PubMed  Google Scholar 

  15. Gao Y, Zhang J, Zhao F. Circular RNA identification based on multiple seed matching. Brief Bioinform. 2018;19(5):803–10. https://doi.org/10.1093/bib/bbx014.

    Article  CAS  PubMed  Google Scholar 

  16. Zhao J, Li Q, Li Y, He X, Zheng Q, Huang S. ASJA: A Program for Assembling Splice Junctions Analysis. Comput Struct Biotechnol J. 2019;17:1143–50. https://doi.org/10.1016/j.csbj.2019.08.001.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  17. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29(1):15–21. https://doi.org/10.1093/bioinformatics/bts635.

    Article  CAS  PubMed  Google Scholar 

  18. Candia J, Bayarsaikhan E, Tandon M, Budhu A, Forgues M, Tovuu LO, et al. The genomic landscape of Mongolian hepatocellular carcinoma. Nat Commun. 2020;11(1):4383. https://doi.org/10.1038/s41467-020-18186-1.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Zheng Q, Bao C, Guo W, Li S, Chen J, Chen B, et al. Circular RNA profiling reveals an abundant circHIPK3 that regulates cell growth by sponging multiple miRNAs. Nat Commun. 2016;7:11215. https://doi.org/10.1038/ncomms11215.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Wang Q, Chen C, Xu X, Shu C, Cao C, Wang Z, et al. APAF1-Binding Long Noncoding RNA Promotes Tumor Growth and Multidrug Resistance in Gastric Cancer by Blocking Apoptosome Assembly. Adv Sci (Weinh). 2022;9(28): e2201889. https://doi.org/10.1002/advs.202201889.

    Article  CAS  PubMed  Google Scholar 

  21. . Zhao S, Ly A, Mudd JL, Rozycki EB, Webster J, Coonrod E, et al. Characterization of cell-type specific circular RNAs associated with colorectal cancer metastasis. NAR Cancer. 2023;5(2):zcad021. https://doi.org/10.1093/narcan/zcad021

  22. Sherman BT, Hao M, Qiu J, Jiao X, Baseler MW, Lane HC, et al. DAVID: a web server for functional enrichment analysis and functional annotation of gene lists (2021 update). Nucleic Acids Res. 2022;50(W1):W216–21. https://doi.org/10.1093/nar/gkac194.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Liao J, Lai H, Liu C, Zhang X, Ou Q, Li Q, et al. Plasma extracellular vesicle transcriptomics identifies CD160 for predicting immunochemotherapy efficacy in lung cancer. Cancer Sci. 2023;114(7):2774–86. https://doi.org/10.1111/cas.15804.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Ray D, Kazan H, Cook KB, Weirauch MT, Najafabadi HS, Li X, et al. A compendium of RNA-binding motifs for decoding gene regulation. Nature. 2013;499(7457):172–7. https://doi.org/10.1038/nature12311.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  25. Fabbiano F, Corsi J, Gurrieri E, Trevisan C, Notarangelo M, D’Agostino VG. RNA packaging into extracellular vesicles: An orchestra of RNA-binding proteins? J Extracell Vesicles. 2020;10(2): e12043. https://doi.org/10.1002/jev2.12043.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  26. Bailey TL, Grant CESEA. Simple Enrichment Analysis of motifs bioRxiv. 2021. https://doi.org/10.1101/2021.08.23.457422.

    Article  Google Scholar 

  27. Van Nostrand EL, Freese P, Pratt GA, Wang X, Wei X, Xiao R, et al. A large-scale binding and functional map of human RNA-binding proteins. Nature. 2020;583(7818):711–9. https://doi.org/10.1038/s41586-020-2077-3.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Chen X, Li A, Sun BF, Yang Y, Han YN, Yuan X, et al. 5-methylcytosine promotes pathogenesis of bladder cancer through stabilizing mRNAs. Nat Cell Biol. 2019;21(8):978–90. https://doi.org/10.1038/s41556-019-0361-y.

    Article  CAS  PubMed  Google Scholar 

  29. Ascano M Jr, Mukherjee N, Bandaru P, Miller JB, Nusbaum JD, Corcoran DL, et al. FMRP targets distinct mRNA sequence elements to regulate protein expression. Nature. 2012;492(7429):382–6. https://doi.org/10.1038/nature11737.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  30. Kugeratski FG, Hodge K, Lilla S, McAndrews KM, Zhou X, Hwang RF, et al. Quantitative proteomics identifies the core proteome of exosomes with syntenin-1 as the highest abundant protein and a putative universal biomarker. Nat Cell Biol. 2021;23(6):631–41. https://doi.org/10.1038/s41556-021-00693-y.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics. 2010;26(6):841–2. https://doi.org/10.1093/bioinformatics/btq033.

    Article  CAS  PubMed  Google Scholar 

  32. Consortium GT. The Genotype-Tissue Expression (GTEx) project. Nat Genet. 2013;45(6):580–5. https://doi.org/10.1038/ng.2653.

    Article  CAS  Google Scholar 

  33. Wu W, Ji P, Zhao F. CircAtlas: an integrated resource of one million highly accurate circular RNAs from 1070 vertebrate transcriptomes. Genome Biol. 2020;21(1):101. https://doi.org/10.1186/s13059-020-02018-y.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Sun D, Wang J, Han Y, Dong X, Ge J, Zheng R, et al. TISCH: a comprehensive web resource enabling interactive single-cell transcriptome visualization of tumor microenvironment. Nucleic Acids Res. 2021;49(D1):D1420–30. https://doi.org/10.1093/nar/gkaa1020.

    Article  CAS  PubMed  Google Scholar 

  35. Glazar P, Papavasileiou P, Rajewsky N. circBase: a database for circular RNAs. RNA. 2014;20(11):1666–70. https://doi.org/10.1261/rna.043687.113.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  36. Zhang Y, Xue W, Li X, Zhang J, Chen S, Zhang JL, et al. The Biogenesis of Nascent Circular RNAs. Cell Rep. 2016;15(3):611–24. https://doi.org/10.1016/j.celrep.2016.03.058.

    Article  CAS  PubMed  Google Scholar 

  37. Wang S, Zhang K, Tan S, Xin J, Yuan Q, Xu H, et al. Circular RNAs in body fluids as cancer biomarkers: the new frontier of liquid biopsies. Mol Cancer. 2021;20(1):13. https://doi.org/10.1186/s12943-020-01298-z.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Wen G, Zhou T, Gu W. The potential of using blood circular RNA as liquid biopsy biomarker for human diseases. Protein Cell. 2021;12(12):911–46. https://doi.org/10.1007/s13238-020-00799-3.

    Article  CAS  PubMed  Google Scholar 

  39. Li Y, He X, Li Q, Lai H, Zhang H, Hu Z, et al. EV-origin: Enumerating the tissue-cellular origin of circulating extracellular vesicles using exLR profile. Comput Struct Biotechnol J. 2020;18:2851–9. https://doi.org/10.1016/j.csbj.2020.10.002.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Xia X, Wang Y, Zheng JC. Extracellular vesicles, from the pathogenesis to the therapy of neurodegenerative diseases. Transl Neurodegener. 2022;11(1):53. https://doi.org/10.1186/s40035-022-00330-0.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Mandaliya H, Jones M, Oldmeadow C, Nordman II. Prognostic biomarkers in stage IV non-small cell lung cancer (NSCLC): neutrophil to lymphocyte ratio (NLR), lymphocyte to monocyte ratio (LMR), platelet to lymphocyte ratio (PLR) and advanced lung cancer inflammation index (ALI). Transl Lung Cancer Res. 2019;8(6):886–94. https://doi.org/10.21037/tlcr.2019.11.16.

    Article  PubMed  PubMed Central  Google Scholar 

  42. Garcia-Martin R, Wang G, Brandao BB, Zanotto TM, Shah S, Kumar Patel S, et al. MicroRNA sequence codes for small extracellular vesicle release and cellular retention. Nature. 2022;601(7893):446–51. https://doi.org/10.1038/s41586-021-04234-3.

    Article  CAS  PubMed  Google Scholar 

  43. Stranska R, Gysbrechts L, Wouters J, Vermeersch P, Bloch K, Dierickx D, et al. Comparison of membrane affinity-based method with size-exclusion chromatography for isolation of exosome-like vesicles from human plasma. J Transl Med. 2018;16(1):1. https://doi.org/10.1186/s12967-017-1374-6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Veerman RE, Teeuwen L, Czarnewski P, Gucluler Akpinar G, Sandberg A, Cao X, et al. Molecular evaluation of five different isolation methods for extracellular vesicles reveals different clinical applicability and subcellular origin. J Extracell Vesicles. 2021;10(9): e12128. https://doi.org/10.1002/jev2.12128.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Muquith M, Espinoza M, Elliott A, Xiu J, Seeber A, El-Deiry W, et al. Tissue-specific thresholds of mutation burden associated with anti-PD-1/L1 therapy benefit and prognosis in microsatellite-stable cancers. Nat Cancer. 2024. https://doi.org/10.1038/s43018-024-00752-x.

    Article  PubMed  Google Scholar 

  46. . Shetgaonkar GG, Marques SM, CEM DC, Vibhavari RJA, Kumar L, Shirodkar RK. Exosomes as cell-derivative carriers in the diagnosis and treatment of central nervous system diseases. Drug Deliv Transl Res. 2022;12(5):1047–79. https://doi.org/10.1007/s13346-021-01026-0

  47. Younas N, Fernandez Flores LC, Hopfner F, Hoglinger GU, Zerr I. A new paradigm for diagnosis of neurodegenerative diseases: peripheral exosomes of brain origin. Transl Neurodegener. 2022;11(1):28. https://doi.org/10.1186/s40035-022-00301-5.

    Article  PubMed  PubMed Central  Google Scholar 

  48. Hornung S, Dutta S, Bitan G. CNS-Derived Blood Exosomes as a Promising Source of Biomarkers: Opportunities and Challenges. Front Mol Neurosci. 2020;13:38. https://doi.org/10.3389/fnmol.2020.00038.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  49. Perez-Boza J, Boeckx A, Lion M, Dequiedt F, Struman I. hnRNPA2B1 inhibits the exosomal export of miR-503 in endothelial cells. Cell Mol Life Sci. 2020;77(21):4413–28. https://doi.org/10.1007/s00018-019-03425-6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Ngo LH, Bert AG, Dredge BK, Williams T, Murphy V, Li W, et al. Nuclear export of circular RNA. Nature. 2024;627(8002):212–20. https://doi.org/10.1038/s41586-024-07060-5.

    Article  CAS  PubMed  Google Scholar 

  51. He R, Zhu J, Ji P, Zhao F. SEVtras delineates small extracellular vesicles at droplet resolution from single-cell transcriptomes. Nat Methods. 2024;21(2):259–66. https://doi.org/10.1038/s41592-023-02117-1.

    Article  CAS  PubMed  Google Scholar 

  52. . Huang SL. RNA-seq analysis of extracellular vesicles from human body biofluids including plasma and urine. SRA. 2023. https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA1012336

  53. . Huang SL. RNA-seq analysis of extracellular vesicles from human body biofluids including plasma, urine, bile, and cerebrospinal fluid (CSF). SRA. 2023. https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA1010887

  54. . Zhao, JJ. Expression profile of circRNAs within extracellular vesicles (EVs) and their corresponding back-splicing ratio profile.. figshare. Dataset. 2024. https://doi.org/10.6084/m9.figshare.25559121.v1

Download references

Acknowledgements

Not applicable

Funding

This work was supported by grants from National Key Research and Development Project of China (2021YFA1300500), National Natural Science Foundation of China (82272625, 82072694), Joint Funds for the innovation of science and Technology, Fujian province (2023Y9320), Fujian Province Natural Science Fund Project (2024J011105), and Financial scheme for young talents training program of Fujian Health industry (No.2017-ZQN-1).

Author information

Authors and Affiliations

Authors

Contributions

Shenglin Huang and Zhuting Fang designed the study; Yan Li acquired data; Jingjing Zhao, Hena Zhang, Jia Hu, and Hongyan Lai performed the bioinformatics analysis; Qiaojuan Li conducted the experiments; Jingjing Zhao, Shenglin Huang, Qiaojuan Li, Hongwu Yu and Youmin Shen wrote and revised the manuscript. All authors have read and agreed to the published version of the manuscript. All authors read and approved the final manuscript.

Corresponding authors

Correspondence to Zhuting Fang or Shenglin Huang.

Ethics declarations

Ethics approval and consent to participate

The NSCLC immunotherapy clinical study was approved by the Institutional Review Board of the Fudan University Shanghai Cancer Center (Approval No. 2004216–20-2005).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Zhao, J., Li, Q., Hu, J. et al. Circular RNA landscape in extracellular vesicles from human biofluids. Genome Med 16, 126 (2024). https://doi.org/10.1186/s13073-024-01400-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s13073-024-01400-w

Keywords