Exploring the Clinical Transformation of circRNA as a Biomarker in Breast Cancer

Abstract

This narrative review systematically examines the potential of circular RNAs (circRNAs) as clinical biomarkers in breast cancer, focusing on the translational gap between basic discoveries and clinical applications. circRNAs are covalently closed non-coding RNA molecules characterized by high stability and tissue-specific expression, making them promising candidates for liquid biopsy. In breast cancer, a highly heterogeneous disease, circRNA expression profiles exhibit subtype-specific patterns and are associated with tumorigenesis, progression, and therapeutic resistance. Mechanistically, circRNAs function as microRNA sponges, interact with RNA-binding proteins, and can even encode functional polypeptides, thereby influencing cancer stem cell properties, immune escape, and organ-specific metastasis. Tumor-derived extracellular vesicles (EVs) carry circRNA signatures that offer a novel source for non-invasive biomarker detection. However, clinical translation is hindered by challenges in detection standardization, functional validation, and a lack of prospective, large-cohort studies. This review critically evaluates the biological functions and clinical relevance of circRNAs, discusses the role of EV-derived circRNAs in metastasis monitoring, and outlines a roadmap to overcome current translational bottlenecks, ultimately assessing their potential as diagnostic, prognostic, and predictive tools in breast cancer management.

1. Introduction

Breast cancer is one of the most common malignant tumors among women worldwide and poses a severe threat to women’s health. According to the estimated new breast cancer cases extracted from the Global Cancer Observatory (GLOBOCAN), the number of new breast cancer cases globally has reached 2.3 million, surpassing lung cancer to become the “most commonly diagnosed cancer worldwide”, and its mortality rate remains high, with the burden of breast cancer expected to continue to rise in the future, especially in transitioning countries. ¹ The high heterogeneity of breast cancer is reflected in its complex molecular subtypes—including luminal A, luminal B, HER2-positive, and triple-negative breast cancer (basal-like and claudin-low)—as well as its diverse clinical behaviors and markedly different treatment responses. ² Despite remarkable progress having been made in the diagnosis and treatment of early-stage breast cancer, strategies for preventing or treating metastatic disease remain inadequate, and recurrence, metastasis and therapeutic resistance are still core clinical challenges. ³ Therefore, the discovery of novel biomarkers for precise diagnosis, prognosis assessment and treatment monitoring is an important direction in breast cancer research.

In recent years, the role of non-coding RNA (ncRNA) in the occurrence and development of cancer has attracted increasing attention. As a class of functional transcripts that do not encode proteins, non-coding RNAs include microRNAs, long non-coding RNAs, circular RNAs and others. They regulate gene expression through diverse mechanisms and play critical roles in tumorigenesis and tumor progression. ⁴ Among them, circRNA has evolved from a transcriptional byproduct to a key molecular regulator due to its unique covalently closed circular structure, high stability, tissue specificity and evolutionary conservation, showing great potential in the research on cancer liquid biopsy biomarkers. ⁵

circRNA was first discovered in plant viroids in 1976 and subsequently confirmed to exist in eukaryotic cells in 1979, but it was long regarded as a byproduct of RNA splicing errors or the splicing process. ⁶ With the development of high-throughput sequencing technology and bioinformatics, thousands of circRNAs have been identified in eukaryotic cells, and their roles in gene regulation and disease pathogenesis have been gradually revealed. ⁷ circRNA is a class of covalently closed circular non-coding RNAs formed by back-splicing. With a covalently closed circular structure and no free 5′ or 3′ ends, circRNA exhibits stronger resistance to RNases and significantly higher stability than linear RNA. It can stably exist in cells for a long time, which provides a solid foundation for its biological functions. ⁸ This characteristic enables circRNA to stably exist in tissues and body fluids, making it an ideal biomarker molecule.

In the field of breast cancer, numerous studies have demonstrated that circRNA expression profiles differ significantly between tumor tissues and normal tissues, and are closely associated with clinicopathological features, molecular subtypes, metastatic potential and patient prognosis. ⁹ Notably, tumor cells actively release circRNAs into the circulatory system within extracellular vesicles (EVs), protecting them from degradation and enabling remote intercellular communication. This has opened new avenues for liquid biopsy, where EV-derived circRNAs hold promise for non-invasive, real-time monitoring of tumor dynamics and organ-specific metastasis.^10,11 Despite the wealth of accumulating data, the clinical translation of circRNA biomarkers remains nascent, facing significant hurdles in standardization, functional causality, and large-scale validation. ¹² This narrative review, guided by the Scale for the Assessment of Narrative Review Articles (SANRA) aims to provide a critical and integrated overview of the current landscape. ¹³ We begin by dissecting the biological foundations and functional mechanisms of circRNAs, emphasizing areas of consensus and controversy. Subsequently, we systematically evaluate the clinical evidence for circRNAs as diagnostic, prognostic, and predictive biomarkers, paying special attention to the emerging role of EV-derived circRNAs in metastasis. Finally, we critically analyze the key translational barriers and propose a strategic roadmap to facilitate the transition of circRNA research from bench to bedside. In addition, the role of circRNAs in regulating cancer stem cell properties, immune escape and metastasis renders them potential therapeutic targets. ¹⁴ The narrative review literature search was performed in PubMed, Web of Science, and Scopus databases up to December 2025, using keywords such as ‘circular RNA’, ‘breast cancer’, ‘biomarker’, ‘liquid biopsy’, ‘extracellular vesicles’, and ‘clinical translation’. Only English-language, peer-reviewed original articles and reviews were considered. Reference lists of identified articles were manually screened for additional relevant studies.

2. Biological Basis of Circular RNAs

2.1 Biogenesis Mechanism

CircRNAs are mainly generated from precursor mRNAs via back-splicing, a processing mechanism distinct from canonical linear splicing (Figure 1). In back-splicing, a downstream 5′ splice donor site is covalently linked to an upstream 3′ splice acceptor site, spanning one or more exons to form a closed circular structure. ¹⁵ This process requires the involvement of the spliceosome machinery, yet its regulatory mechanisms differ substantially from those of linear splicing.

Figure 1.

Biogenesis and functional mechanisms of CircRNA

Current studies have revealed several major mechanisms underlying circRNA biogenesis. The first is mediated by complementary sequences in flanking introns. These sequences (such as Alu repetitive elements) bring splice donor and acceptor sites into close proximity through base pairing, thereby promoting back-splicing. ¹⁶ The second is regulated by RNA-binding proteins (RBPs). Certain RBPs, such as Muscleblind (MBL), Quaking (QKI) and FUS, can bring splice sites into close proximity by binding to flanking intronic sequences, thereby facilitating the circularization process. ¹⁷ Third, exon lariat structures generated by exon skipping can also serve as circRNA precursors, forming exon-derived circRNAs through intramolecular splicing. ¹⁸

According to their origins, circRNAs can be classified into three categories: exonic circRNAs (ecircRNAs), which are mainly located in the cytoplasm and represent the most common type; circular intronic RNAs (ciRNAs), which are mostly retained in the nucleus and participate in transcriptional regulation; and exon–intron circRNAs (EIciRNAs), which contain both exon and intron sequences and are localized in the nucleus. ¹⁹ The biogenesis mechanisms and functions of different types of circRNAs vary, reflecting the complexity of the circRNA regulatory network (Figure 1).

2.2 Structural Features and Stability

This high stability is a key attribute for their biomarker potential, enabling detection in stable bodily fluids. ²⁰ Most circRNAs are composed of 2-5 exons and contain a unique back-splice junction (BSJ) sequence, which serves as a definitive molecular signature for their specific identification. ²¹

The sequence composition of circRNAs also exhibits certain characteristics. Most circRNAs are composed of exons, typically containing 2 to 5 exons, with lengths ranging from hundreds to thousands of nucleotides. ²² The back-splice junction of circRNAs is a characteristic sequence that distinguishes them from linear RNAs and can serve as a molecular marker for specific detection. Furthermore, circRNA sequences often contain miRNA response elements (MREs), laying the structural foundation for their function as miRNA sponges. ²³

2.3 Major Functional Mechanisms

2.3.1 miRNA Sponging

The most classic functional mechanism of circRNAs is to act as miRNA sponges, competitively binding to miRNAs and thereby relieving the inhibitory effect of miRNAs on their target genes. ²⁴ A classic and representative example is ciRS-7 (also known as CDR1as), which harbors more than 70 binding sites for miR-7, enabling it to efficiently sequester miR-7 and upregulate the expression of miR-7 target genes. ²⁵ However, the universality and physiological relevance of this mechanism have been increasingly challenged. First, stoichiometric concerns suggest that for a circRNA to effectively sponge a miRNA, it must be present in molar excess relative to the miRNA and its target mRNAs, a condition not always met in vivo ²⁶ ; Second, spatial expression data reveal that the archetypal sponge ciRS-7 is predominantly expressed in stromal cells rather than cancer cells, questioning its functional relevance in tumor cell-autonomous processes. ²⁷ These findings urge a cautious interpretation of the ceRNA hypothesis and emphasize the need for rigorous validation of sponge interactions in relevant cellular contexts.

2.3.2 Interaction With RNA-Binding Proteins

CircRNAs can also directly interact with RNA-binding proteins (RBPs) to regulate protein functions or participate in the assembly of protein complexes. MBL protein promotes the biogenesis of circMbl by binding to pre-mRNA, whereas circMbl contains multiple high-affinity binding sites for MBL and can further sequester MBL protein, forming a feedback loop that regulates MBL levels. ¹⁷ circFoxo3 forms a ternary complex with p53 and MDM2, promoting MDM2-induced p53 ubiquitination and degradation, while protecting Foxo3 from MDM2-mediated degradation, thereby upregulating PUMA expression and inducing tumor cell apoptosis. ²⁸ In breast cancer, circANKS1B directly binds to TRIM25 protein, enhances its E3 ubiquitin ligase activity, and promotes the ubiquitination and degradation of transcription factor STAT1. This in turn relieves the transcriptional repression of miR-148a-3p by STAT1, ultimately promoting invasion and metastasis of breast cancer by upregulating key EMT factors such as ZEB1/ZEB2. ²⁹

2.3.3 Transcriptional Regulatory Functions

Nuclear-localized circRNAs (such as ciRNAs and EIciRNAs) can participate in transcriptional regulation. circPAIP2 is recruited to the promoter region of its parental gene PAIP2 via specific binding with U1 snRNP, and interacts with RNA polymerase II, thereby cis-activating the transcription of the PAIP2 gene. ³⁰ circSEP3 can bind strongly to its homologous DNA locus to form an RNA:DNA hybrid (R-loop), leading to transcriptional pausing and facilitating the recruitment of splicing factors, thereby regulating alternative splicing of its cognate mRNA and increasing the abundance of exon-skipping isoforms. ³¹ These findings expand the functional spectrum of circRNAs and reveal their multilayered regulatory roles at both transcriptional and post-transcriptional levels.

2.3.4 Translation Into Polypeptides

Recent groundbreaking studies have shown that some circRNAs, despite lacking a 5′ cap, can be translated through cap-independent mechanisms driven by internal ribosome entry sites (IRES) or N⁶-methyladenosine (m⁶A) modifications.^31,32 This adds a new layer to their functionality. In breast cancer, circCAPG encodes the 171-amino-acid peptide CAPG-171aa, which promotes TNBC migration by activating the MEKK2-MEK1/2-ERK1/2 pathway. ¹⁰ Similarly, circHER2 encodes the HER2-103 peptide, which potentiates signaling via EGFR/HER3 heterodimers, driving proliferation and metastasis in TNBC. ³³ While this finding is paradigm-shifting, the overall translation efficiency of circRNAs is generally low, and the expression levels of the encoded peptides are often orders of magnitude lower than their linear counterparts. The in vivo significance and functional relevance of the majority of these peptides remain to be established. The biological contexts (e.g., stress conditions, specific cell types) under which cap-independent translation is favored are still being elucidated, representing a critical area for future investigation. ³⁴ These findings reveal a novel dimension of circRNA functions and provide new insights for targeted therapy.

2.4 A Note on Functionality vs. Passenger Events

It is crucial to acknowledge that the vast majority of the thousands of circRNAs identified in mammalian transcriptomes are likely non-functional byproducts of splicing. ³⁵ Genome-wide analyses suggest that most circRNAs lack the necessary sequence features to exert potent regulatory functions. Distinguishing the rare, functional circRNAs from the abundant, noise-derived passenger molecules remains a major challenge. High-throughput genetic screening approaches, such as CRISPR-Cas13-based functional screens, are essential tools to systematically identify and validate those circRNAs with genuine phenotypic consequences. ³⁶

3. Expression Profiles and Clinical Relevance of circRNAs in Breast Cancer

3.1 Expression Characteristics of circRNAs in Breast Cancer Tissues

Advances in high-throughput sequencing technologies have enabled the systematic identification of thousands of circRNAs in breast cancer tissues. Compared with normal breast tissues, the circRNA expression profile in breast cancer tissues is significantly reprogrammed, forming tumor-specific circRNA signatures that can be used to distinguish breast cancer. ³² These differentially expressed circRNAs are not only abundant in quantity but also exhibit tissue specificity and subtype specificity. Their expression alterations are involved in multiple tumor-associated biological processes, including cell proliferation, metabolic reprogramming, apoptosis, angiogenesis, epithelial-mesenchymal transition and metastasis. ³³

At the genomic level, breast cancer-associated circRNAs are mostly derived from exonic regions of protein-coding genes, and are particularly enriched in genes of cancer-related signaling pathways, such as key molecules in the TGF-β, Wnt/β-catenin, VEGF, and PI3K/AKT pathways. ³⁴ The expression levels of circRNAs are not always positively correlated with those of their parental linear mRNAs. CircRNAs and mRNAs follow distinct and largely independent regulatory logics, suggesting that circRNA biogenesis is governed by regulatory mechanisms separate from those governing linear RNA splicing. ³⁵ For example, the RNA-binding protein QKI is induced during epithelial-mesenchymal transition. It can promote the biogenesis of numerous circRNAs by binding to QKI response elements in flanking introns, thereby playing a key regulatory role in breast cancer metastasis. ³⁶

3.2 Subtype-Specific Expression Patterns

Molecular classification of breast cancer (luminal subtype, HER2-positive subtype, triple-negative subtype) displays distinct clinical behaviors and molecular characteristics. CircRNA expression profiles also exhibit remarkable differences across these subtypes, making them potential biomarkers for the molecular subtyping diagnosis of breast cancer. ⁹

3.2.1 Luminal Breast Cancer

This subtype is characterized by estrogen receptor (ER) positivity, and circRNA expression is closely correlated with the ER signaling pathway. circESR1 (circRNA_0025202) is formed by the circularization of exons from the ERα gene (ESR1). It is highly expressed in luminal breast cancer and can be upregulated by estrogen induction. ³⁷ circESR1 interacts with HNRNPAB protein to form a positive feedback loop: HNRNPAB, activated by the ER/SP1 signaling pathway, promotes the biogenesis of circESR1, while the circESR1 transcript enhances the stability and expression of HNRNPAB. The two synergistically regulate the expression of cell cycle–related genes such as CDK1/CDK6, augment the activity of the ER signaling pathway, and contribute to endocrine therapy resistance. ³⁸ circGREB1 is also induced by the estrogen signaling pathway and shows high expression in ER-positive breast cancer, with its expression level closely associated with the sensitivity to endocrine therapy. ³⁹ In addition, hsa_circ_0086735 is significantly upregulated in luminal breast cancer tissues and cell lines. Its high expression level is closely associated with malignant tumor progression and tamoxifen resistance, and is significantly correlated with poor overall survival and distant metastasis-free survival. ⁴⁰

3.2.2 HER2-Positive Breast Cancer

This subtype is characterized by HER2 gene amplification and overexpression. Currently, anti-HER2 therapy serves as a crucial therapeutic strategy for HER2-positive breast cancer. However, primary or acquired resistance occurs in some patients, leading to treatment failure and compromised efficacy. ⁴¹ Therefore, it is of great clinical significance to identify specific diagnostic biomarkers and predictive indicators of therapeutic response for HER2-positive breast cancer. circ-ERBB2 is significantly highly expressed in HER2-positive breast cancer tissues, and its expression level is closely associated with lymph node metastasis and HER2 status in patients. Mechanistically, circ-ERBB2 can act as a sponge for miR-136-5p and miR-198, relieving the inhibitory effect on the downstream target gene TFAP2C, thereby promoting proliferation, migration and invasion of HER2-positive breast cancer cells, inhibiting apoptosis, and driving tumor progression and metastasis. ⁴² In addition, circCDYL2 is highly expressed in trastuzumab-resistant HER2-positive breast cancer. Its upregulation is closely associated with shorter disease-free survival and overall survival in patients. By stabilizing the GRB7 protein, circCDYL2 continuously activates the downstream AKT and ERK1/2 signaling pathways, thereby reducing the sensitivity of cancer cells to trastuzumab and mediating drug resistance. ⁴³

3.2.3 Triple-Negative Breast Cancer

TNBC is the most aggressive subtype of breast cancer with the poorest prognosis and lacks effective targeted therapies. Multiple studies have identified TNBC-specific circRNA profiles using high-throughput sequencing. circRAD18 is significantly highly expressed in TNBC tissues, and its expression level is positively correlated with T stage, clinical stage and histological grade. Mechanistically, it acts as a molecular sponge for miR-208a and miR-3164, and promotes the proliferation, migration and tumor growth of TNBC by upregulating the expression of IGF1 and FGF2. ²⁶ circANKS1B is significantly upregulated in TNBC. It upregulates the expression of transcription factor USF1 by sponging miR-148a-3p and miR-152-3p, which in turn transcriptionally activates TGF-β1, initiates the TGF-β1/Smad signaling pathway, and induces EMT, thereby promoting invasion and metastasis of breast cancer. ²⁹ circCFL1 is remarkably highly expressed in TNBC. Acting as a scaffold molecule, it enhances the interaction between HDAC1 and c-Myc, promotes the stability of the c-Myc protein by inhibiting deacetylation-mediated K48 ubiquitination. The stably expressed c-Myc further binds to the TP53 promoter to facilitate the transcription of mutant p53, thereby maintaining cancer stem cell properties and promoting immune escape. ⁴⁴ In addition, hsa_circ_0006220 is significantly downregulated in TNBC tissues, and its low expression is associated with lymph node metastasis and poor patient prognosis. Overexpression of this circular RNA suppresses the proliferation, migration and invasion of TNBC cells by sponging miR-197-5p to regulate CDH19 expression, suggesting its potential role as a tumor suppressor. ⁴⁵

3.3 Correlation With Clinicopathological Characteristics

The expression levels of circRNAs are extensively and closely correlated with the clinicopathological features of breast cancer patients. These correlations are not only statistically significant but also reveal the functional implications of circRNAs in tumor progression. This correlation is summarized in Table 1.

Table 1.

Clinical Significance of Dysregulated circRNAs in Breast Cancer

circRNA	Sample Type	Expression in BC	Clinical relevance	Functional mechanism
Diagnostic
hsa_circ_0001785	Plasma	Up	Distinguishes BC from controls (AUC=0.784)	N/A
circHIPK3	Plasma/Tissue	Up	Distinguishes BC from controls (AUC-0. 809)	N/A
circ-FAF1	Serum	Down	Distinguishes BC from controls (AUC=0.885)	N/A
Prognostic (Poor)
circ_0001008	Tissue (TNBC)	Up	Independent risk factor for OS (HR not provided)	Promotes proliferation/migration
circANKS1B	Tissue	Up	Independent risk factor for OS and distant metastasis (HR=4.2)	Sponges miR-148a/152, activates TGF-β1
circCFL1	Tissue (TNBC)	Up	Independent poor prognostic factor for OS and DFS (HR>2)	Stabilizes c-Myc, promotes immune escape
circCDYL	Tissue	Up	Poor DFS and OS in ERα- patients (P<0.01)	Sponges miR-1275, promotes autophagy
Prognostic (Favorable)
hsa_circ_0006220	Tissue (TNBC)	Down	Low expression linked to poor outcome	Sponges miR-197-5p, upregulates CDH19
circFBXW7	Tissue (TNBC)	Down	Independent protective prognostic factor	Sponges miR-197-3p, encodes 185-aa peptide
Predictive
circPVT1	Tissue (ER+)	Up	Predicts resistance to endocrine therapy	Sponges miR-181a-2-3p, upregulates ESR1
circCDYL2	Tissue (HER2+)	Up	Predicts trastuzumab resistance and poor survival	Stabilizes GRB7, activates HER2 signaling
CircPLK1	Tissue (TNBC)	Up	Predicts resistance to anthracycline-based chemo	Promotes doxorubicin resistance

Abbreviations: BC: breast cancer; OS: overallsurvival; DFS: disease-free survival; HR: hazard ratio; N/A: not applicable

3.3.1 Tumor Size and Growth

circDENND4C is highly expressed in breast cancer tissues, and its expression level is positively correlated with primary tumor diameter. ⁴⁶ Mechanistic studies have shown that circDENND4C is transcriptionally regulated by HIF1α. Under hypoxic conditions, it regulates the expression of downstream target genes by adsorbing microRNAs such as miR-33b-5p, thereby promoting the proliferation and angiogenesis of breast cancer cells. circ_0001006 is significantly upregulated in TNBC tissues, and its expression level is significantly positively correlated with histological grade and Ki-67 proliferation index in patients. High expression is mostly found in high-grade tumors with greater proliferative activity, and can also predict poorer patient prognosis and a higher risk of disease progression. ⁴⁷

3.3.2 Lymph Node Metastasis

circANKS1B is significantly highly expressed in breast cancer tissues, and its high expression is significantly positively correlated with lymph node metastasis and advanced clinical stage. Multivariate Cox regression analysis shows that circANKS1B is an independent risk factor for overall survival in breast cancer patients, suggesting that it plays an important role in breast cancer metastasis and poor prognosis. ²⁹ hsa_circ_0061260 and hsa_circ_0060876 are significantly upregulated in lymph node-positive patients. Functional enrichment analysis reveals that these circRNAs are mainly involved in tumorigenesis and metastasis-related pathways, suggesting that their high expression is not only significantly associated with lymph node-positive status but may also play an important regulatory role in the process of lymph node metastasis. ⁴⁸ circCLASP1 is significantly highly expressed in breast cancer tissues, its expression level is positively correlated with lymph node metastasis and Ki67 index, it can induce EMT by stabilizing GLI1 protein and upregulating SNAIL expression, thereby promoting lymph node metastasis and distant lung metastasis. ⁴⁹

3.3.3 Distant Metastasis

circEPSTI1 is significantly highly expressed in TNBC tissues, its high expression is significantly associated with shortened overall survival and distant metastasis-free survival in patients, multivariate Cox regression analysis further confirms that its expression level is an independent predictor of postoperative distant metastasis in TNBC patients. ⁵⁰ hsa_circRNA_002178 is highly expressed in breast cancer tissues and is associated with poor prognosis in patients, it can relieve the inhibition of COL1A1 by competitively binding to miR-328-3p, upregulate COL1A1 expression to promote tumor cell proliferation and migration, and participate in the formation and remodeling of the pre-metastatic microenvironment. ⁵¹ In recent years, Bao et al. (2025) systematically reviewed the dual regulatory roles of circRNAs in distant metastasis of breast cancer, uncovered the complex networks through which circRNAs modulate metastasis via EMT, angiogenesis, immune escape and remodeling of organ-specific microenvironments, and discussed their translational prospects in metastasis prediction and targeted therapy. ⁵²

3.3.4 Hormone Receptor Status

circRNA-SFMBT2 is expressed at significantly higher levels in ER+ breast cancer cells than in ER− cells and is positively correlated with ERα protein levels, it can enhance protein stability by directly binding to ERα protein and regulating its ubiquitination and degradation, thereby promoting the transcriptional activation of ERα target genes and malignant tumor progression. ⁵³ circCDYL expression level in breast cancer tissues is 3.2-fold higher than that in adjacent tissues, and its expression level is negatively correlated with estrogen receptor α expression, suggesting that circCDYL has higher expression activity in ER-negative subtypes. ⁵⁴ circPGR derived from the PGR gene is specifically highly expressed in ER-positive breast cancer and is associated with hormone receptor-positive status and poor prognosis, it can regulate cell cycle genes to participate in proliferation control by sponging miR-301a-5p and may be a potential molecular marker for the PR signaling pathway. ⁵⁵

3.3.5 HER2 Status

The aforementioned experiments confirmed through clinical sample analysis that circ-ERBB2 is significantly associated with HER2-positive status, but has no significant correlation with other clinicopathological characteristics such as age, tumor size, lymph node metastasis and TNM stage. ⁴² circ-PPID is significantly downregulated in trastuzumab-resistant cells and tissues of HER2-positive breast cancer, as a negative regulator, its low expression can relieve the inhibition of the HER2 signaling pathway, thereby promoting abnormal activation of this pathway. ⁵⁶

3.3.6 Histological Grade

RNA sequencing and qRT-PCR validation based on TNBC tissues and cells show that circIFI30 is significantly upregulated in TNBC, its high expression level is significantly positively correlated with histological grade in patients, the expression level in histological grade III tumors is significantly higher than that in histological grade II tumors (P<0.05). ⁵⁷ hsa_circ_0043278 is significantly lowly expressed in breast cancer tissues, it is negatively correlated with histological grade and TNM stage, its low expression indicates shortened patient survival and may serve as a tumor suppressor to inhibit malignant progression of tumors. ⁵⁸

3.3.7 Molecular Subtyping

Bioinformatics analysis based on circRNA expression characteristics can screen molecular axes related to breast cancer molecular subtyping, the integrally constructed regulatory network can distinguish subtypes, especially has differential value for the basal-like subtype, and can be used as an auxiliary tool for molecular subtyping to supplement the traditional classification system. ⁵⁹ Circular RNAs exhibit specific expression in various molecular subtypes of breast cancer and can assist in molecular subtyping, among which circCSPP1 is significantly highly expressed in TNBC and multiple receptor-negative subgroups with the most obvious difference in TNBC, circNRIP1 is also highly expressed in ER-negative and TNBC subgroups, ⁶⁰ these findings suggest that expression profiling analysis of subtype-specific circRNAs is expected to serve as a supplementary tool for the traditional PAM50 molecular subtyping and provide new molecular basis for precise classification and individualized diagnosis and treatment of breast cancer.

3.4 Prognostic Value

circRNA expression is closely related to the survival prognosis of breast cancer patients, numerous studies have confirmed that specific circRNAs can serve as independent prognostic markers, and their predictive ability is not affected by traditional clinicopathological factors such as age, tumor size, lymph node status and grade.

3.4.1 Poor Prognosis-Related circRNAs

circ_0001006:In TNBC patients, overall survival was significantly shorter in those with high circ_0001006 expression. Kaplan–Meier survival analysis showed a negative correlation between its expression level and overall survival, and multivariate Cox regression analysis confirmed it as an independent prognostic indicator. Functional experiments demonstrated that knockdown of this molecule could inhibit the proliferation, migration and invasion of TNBC cells, suggesting that its high expression promotes malignant progression of tumors. ⁴⁷

circANKS1B: Breast cancer patients with high circANKS1B expression exhibit a significantly elevated risk of distant metastasis and recurrence. Kaplan–Meier survival analysis revealed a significant association with poor prognosis (log-rank test, p=0.002), and multivariate analysis confirmed it as an independent risk factor. This circRNA activates the TGF-β1/Smad pathway and induces EMT by sponging miR-148a-3p and miR-152-3p, thereby promoting invasion and distant metastasis of breast cancer cells. ²⁹

circCFL1: In TNBC patients, high expression of circCFL1 is significantly associated with poor prognosis. Univariate and multivariate analyses confirmed that circCFL1 is an independent poor prognostic factor affecting overall survival and disease-free survival. ⁴⁴ circCFL1 can promote stemness maintenance and immune escape in TNBC by stabilizing c-Myc, upregulating mutp53 and activating related pathways. It can also upregulate PD-L1 to suppress CD8+ T-cell function. Its high expression is closely associated with tumor grade, metastasis and recurrence, and plays a key regulatory role in the malignant progression of TNBC.

circCDYL: Breast cancer patients with high circCDYL expression exhibit poor disease-free survival and overall survival, and its expression level is higher in ERα-negative breast cancer patients. ⁵⁴ Mechanistically, circCDYL upregulates ATG7 and ULK1 expression by sponging miR-1275, enhances autophagy, thereby promoting breast cancer progression. circCDYL can serve as a potential prognostic and predictive molecule for breast cancer.

circPVT1: circPVT1 is highly expressed in ERα-positive breast cancer. Acting as a competing endogenous RNA, it sponges miR-181a-2-3p, upregulates the expression of ESR1 and its downstream ERα target genes, and promotes the tumorigenesis and endocrine therapy resistance of ERα-positive breast cancer. These findings suggest that circPVT1 can serve as a potential diagnostic biomarker and therapeutic target for ERα-positive breast cancer. ⁶¹

3.4.2 Favorable Prognosis-Related circRNAs

circTADA2A-E6: circTADA2A-E6 is significantly downregulated in TNBC. Patients with its high expression exhibit significantly prolonged disease-free survival, whereas low expression is significantly associated with higher histological grade and lymph node metastasis. Overexpression of circTADA2A-E6 can inhibit cell proliferation, migration, invasion and colony formation, exerting a tumor-suppressive function, and can serve as a protective prognostic biomarker for TNBC. ⁶²

hsa_circ_0006220: hsa_circ_0006220 is specifically downregulated in TNBC. Its overexpression can significantly inhibit cell proliferation, migration and invasion. Acting as a ceRNA, it exerts tumor-suppressive effects through the miR-197-5p/CDH19 axis, and is expected to become a potential diagnostic biomarker and therapeutic target for TNBC. ⁴⁵

circFBXW7: circFBXW7 is downregulated in TNBC and is significantly associated with poor clinical outcomes, larger tumor size and lymph node metastasis, acting as an independent prognostic factor. It can not only serve as a sponge for miR-197-3p to upregulate FBXW7, but also encode a polypeptide to promote c-Myc degradation, thereby inhibiting the proliferation and migration of TNBC cells as well as tumor growth, and exerting tumor-suppressive functions. ⁶³

circFAM53B: circFAM53B can encode cryptic antigenic peptides through non-canonical translation, which efficiently activate the immune responses of CD4⁺ and CD8⁺ T cells. Its expression is significantly correlated with the infiltration of antigen-specific CD8⁺ T cells, and patients with high expression have longer survival. This peptide exhibits strong binding affinity to HLA-I/II, enhances antigen presentation and activates adaptive immunity. These findings indicate that circRNA-encoded peptides can negatively regulate tumor progression, and their high expression can suppress tumor growth and improve prognosis. ⁶⁴

3.4.3 Multi-circRNA Prognostic Models

The prognostic value of a single circRNA is limited, while combinations of multiple circRNAs can significantly improve the accuracy of prognostic prediction. Prognostic risk score models constructed based on circRNA-associated ceRNA regulatory networks can effectively divide breast cancer patients into high-risk and low-risk groups. Studies have shown that the overall survival rate of patients in the high-risk group is significantly lower than that in the low-risk group. Multivariate Cox regression analysis confirms that this risk score is an independent risk factor for overall survival, suggesting that the model can serve as an effective tool for prognostic evaluation of breast cancer. ⁶⁵ One study constructed a ceRNA network and established a TNBC prognostic model based on 9 mRNA genes including SH3BGRL2 and CA12, this model could effectively distinguish high-risk and low-risk groups, the median overall survival of the high-risk group was significantly shortened with an AUC of 0.90, the study also identified a variety of circRNAs that regulate key genes in the MAPK pathway, providing new ideas for the prognostic evaluation of TNBC. ⁶⁶

3.5 Diagnostic Value

The application value of circRNA in the diagnosis of breast cancer mainly derives from its specific expression in tumor tissues and detectability in body fluids, and numerous studies have evaluated the efficacy of circRNA used alone or in combination for the diagnosis of breast cancer. The expression of hsa_circ_0068033 in breast cancer tissues is significantly lower than that in non-tumor control tissues, and ROC curve analysis shows that the area under the curve (AUC) for distinguishing the two is as high as 0.8480. ⁶⁷ CircHIPK3 is significantly upregulated in both breast cancer tissues and plasma samples, and the AUC for distinguishing breast cancer patients from healthy controls is 0.8087 (95%CI: 0.7309–0.8866, P<0.0001). ⁶⁸

The diagnostic efficacy of a single circRNA is usually limited, and combinations of multiple circRNAs can significantly improve diagnostic accuracy. A meta-analysis found that the overall diagnostic performance of multiple circRNA combinations was superior to that of a single circRNA, with an AUC of 0.82 (95% CI 0.82–0.88) for distinguishing breast cancer patients from healthy controls, and the combined sensitivity and specificity were 85% and 86%, respectively, suggesting that the diagnostic strategy combining multiple circRNAs can effectively improve the diagnostic accuracy of breast cancer. ⁶⁹ Another meta-analysis including 16 studies showed that the pooled AUC of circRNAs (including circAGFG1, circANKS1B, etc.) for distinguishing breast cancer patients from healthy controls was 0.66, with a sensitivity of 65% and a specificity of 68%, suggesting that circRNAs can serve as auxiliary tools for breast cancer diagnosis. ⁷⁰ These studies have laid a foundation for the development of circRNA-based liquid biopsy products. Notably, innovative detection approaches based on CRISPR-Cas technology are offering novel technical routes for the clinical detection of circRNAs. Tan et al. (2024) established a one-step dual fluorescence detection system relying on CRISPR-Cas13a/Cas12a, which enables simultaneous detection of circROBO1 and BRCA1 with limits of detection as low as 0.013 pM and 0.26 pM, respectively. This system provides a highly sensitive technical platform for multi-target combined diagnosis based on circRNAs. ⁷¹

4. Role of Extracellular Vesicle Circular RNAs in Breast Cancer Metastasis

4.1 Basic Characteristics of Extracellular Vesicles

EVs are vesicles released by cells with a lipid bilayer membrane structure, and according to their size, EVs can be divided into three subgroups: exosomes (30–150 nm in diameter), microvesicles (100–1000 nm in diameter), and apoptotic bodies (1–5 μm in diameter). ⁷² EVs carry a variety of biological macromolecules such as proteins, mRNA, miRNA, lncRNA and circRNA, and mediate intercellular communication by delivering them to recipient cells, playing a key role in tumor microenvironment regulation and the formation of pre-metastatic niches at distant sites. ⁷³

Studies have found that exosome-derived circRNAs can resist degradation by RNases and stably exist in the circulatory system due to the protection of the vesicle membrane. ⁷⁴ The composition of exosomal circRNAs is closely related to donor cells, and their differences originate from the selective packaging mechanism of exosomes, which can reflect the molecular characteristics of tumors and make them ideal biomarker sources for liquid biopsy, and some RNA-binding proteins can drive the selective packaging of circRNAs into exosomes by recognizing specific sequence motifs, thereby regulating intercellular communication and affecting tumor progression. ¹²

4.2 EVs circRNA in Organ-Specific Metastasis

Breast cancer metastasis shows a distinct organ-selective tendency, with common metastatic sites including bone, lung, liver, and brain. ⁷⁵ Recent studies have revealed that tumor-derived EVs play a key role in organotropic metastasis. By carrying specific molecular cargoes, they home to targeted organs, remodel the local microenvironment, and form a pre-metastatic niche that facilitates the colonization of tumor cells. ⁷⁶ As depicted in Figure 2, circRNAs within EVs participate in the regulation of this process.

Figure 2.

The role of EV circRNA in organ specific metastasis of breast cancer

4.2.1 EVs circRNA Associated With Lung Metastasis

TNBC-derived EVs can induce lung pre-metastatic niche formation by activating α-SMA+ myofibroblasts and remodeling the lung endothelium. ⁶⁴ Mechanistically, reduced LSD1 in breast cancer cells leads to decreased circDOCK1 expression and increased miR-1270 packaging into exosomes. After uptake by lung endothelial cells, miR-1270 downregulates ZO-1, increasing vascular permeability and facilitating lung metastasis. ⁶⁵

4.2.2 EVs circRNA Associated With Bone Metastasis

Bone is a common metastatic site for luminal breast cancer. In breast cancer cells with bone metastatic propensity, the expression level of a novel circular RNA, circPEX13, is significantly upregulated and packaged into a special type of EVs known as Large Oncosomes (LOs). ⁷⁷ CircIKBKB is significantly upregulated in bone-metastatic breast cancer tissues. It promotes IKKβ-mediated IκBα phosphorylation by activating the NF-κB pathway, upregulates the expression of various bone remodeling factors, thereby facilitating osteoclast differentiation, forming an osteolytic pre-metastatic niche, and driving bone metastasis of breast cancer. ⁷⁸

4.2.3 EVs circRNA Associated With Liver Metastasis

Liver metastasis also represents a common metastatic pattern of breast cancer. Overexpressed circPDE7B in breast cancer cell-derived exosomes can remodel the hepatic microenvironment and facilitate liver metastasis. ⁷⁹ Meanwhile, EVs secreted by triple-negative breast cancer cells continuously reshape the liver pre-metastatic niche and upregulate the expression of genes such as Cx3cr1, thereby constructing a favorable microenvironment for the colonization and proliferation of tumor cells. ⁸⁰

4.2.4 EVs circRNA Associated With Brain Metastasis

Brain metastasis is one of the most fatal complications of breast cancer. Current studies worldwide have revealed the correlation between EVs and brain metastasis. ⁸¹ Meanwhile, accumulating evidence indicates that circRNAs can be efficiently packaged into extracellular vesicles and regulate recipient cells via intercellular communication, exerting pivotal regulatory effects on tumor progression and metastasis. ⁸² A latest study based on the EV-circRNA omics atlas of multiple human body fluids has demonstrated that cerebrospinal fluid EVs are enriched with brain-specific circRNAs and cancer-related circRNAs. The expression profiles of EV-circRNAs are correlated with clinical outcomes, suggesting that they can serve as novel molecular targets for liquid biopsy of central nervous system tumors and provide a theoretical basis for non-invasive dynamic monitoring of breast cancer patients with brain metastasis. ⁸³

4.3 EVs circRNA as Metastatic Biomarkers

Given the vital roles of EVs circRNAs in organ-specific metastasis and their high stability in body fluids, they are promising liquid biopsy biomarkers for monitoring metastasis occurrence and evaluating metastatic burden. CircEGFR is markedly upregulated in plasma exosomes of breast cancer patients. Its expression level is positively correlated with malignant features such as lymph node metastasis, and it can promote autophagy, malignant progression and metastasis of TNBC cells. ⁸⁴ Exosomal circRNAs stably exist in the body fluids of breast cancer patients. Their aberrantly elevated expression is closely correlated with tumor metastasis and poor patient prognosis, acting as potential non-invasive biomarkers for breast cancer diagnosis and prognostic evaluation. ⁸⁵ A clinical study enrolling 379 breast cancer patients receiving neoadjuvant therapy demonstrated that the baseline level of plasma exosomal circRNAs was significantly elevated in patients with postoperative distant metastasis. It serves as an independent risk factor for postoperative distant metastasis, and presents excellent performance in AUC, sensitivity and specificity for metastasis prediction. These findings indicate that plasma exosomal circRNAs can act as efficient and non-invasive novel biomarkers for prognostic evaluation of breast cancer. ⁸⁶

5. Clinical Translation of circRNAs as Biomarkers in Breast Cancer

5.1 Diagnostic Biomarkers

Given the high stability, tissue specificity and detectability in body fluids, circRNAs have been extensively investigated for their values as diagnostic biomarkers in breast cancer. The differential expression of circRNAs between tumor tissues and normal tissues enables them to be applied for auxiliary differential diagnosis. Circ-FAF1 is significantly downregulated in the serum of breast cancer patients, with a specificity of 77% and sensitivity of 74% in distinguishing breast cancer patients from healthy controls. Combined detection of Circ-FAF1 and Circ-ELP3 further improves the diagnostic efficiency, yielding an AUC of 0.891. ⁸⁷ Circ-UBAP2 (hsa_circ_0001846) is markedly upregulated in TNBC. Its high expression is significantly correlated with poor prognosis and malignant clinicopathological features, including larger primary tumor size, advanced TNM stage and lymph node metastasis, suggesting that circ-UBAP2 can serve as an independent prognostic biomarker and a potential therapeutic target for TNBC. ⁸⁸

Since the diagnostic efficacy of a single circRNA is limited, combined detection of multiple circRNAs can remarkably improve diagnostic accuracy. A combined diagnostic model constructed based on three tumor-derived plasma circRNAs, including hsa_circ_0000091, hsa_circ_0067772 and hsa_circ_0000512, exhibits excellent diagnostic performance for breast cancer. In the training set, the model achieves an AUC of 0.974 (95%CI: 0.952–0.996), with a sensitivity of 97.1% and a specificity of 90.2%. Its favorable efficacy is well maintained in the validation set and significantly superior to the combined detection of conventional serum tumor biomarkers (AUC=0.838). This panel is expected to become a non-invasive biomarker for the early diagnosis of breast cancer. ⁸⁹ Based on plasma extracellular vesicles, the diagnostic value of circRNAs in breast cancer was screened and validated in a total of 259 samples, consisting of 144 breast cancer patients and 115 benign controls. A 9-circRNA combined classifier named BCExoC was constructed using machine learning algorithms. The AUC value was 0.83 (95% CI: 0.77–0.88) in the training cohort and 0.80 (95% CI: 0.71–0.89) in the validation cohort. This panel exhibited better diagnostic efficacy than single circRNA biomarkers. Collectively, the combined detection of plasma EVs-derived circRNAs holds great promising prospects for the early diagnosis of breast cancer. ⁹⁰

The combination of circRNAs with other biomarkers can also improve diagnostic efficacy. The combined detection strategy based on plasma hsa_circ_0001785, CEA and CA15-3 showed that the area under the ROC curve was significantly increased from 0.629 for CA15-3 alone to 0.851 when the three biomarkers were combined. This finding confirms that the combined application of circRNA and traditional tumor biomarkers can effectively improve the diagnostic accuracy of breast cancer. ⁹¹ The combined evaluation of circHIPK3 with CA15-3 and CEA presents higher sensitivity and specificity, indicating that the combination of circRNAs and conventional tumor biomarkers can markedly enhance diagnostic efficacy. ⁶⁸ These studies lay a foundation for the clinical development of multi-biomarker combined detection kits.

5.2 Prognostic Biomarkers

The expression level of circRNAs is closely associated with the prognosis of breast cancer patients and can serve as an independent prognostic biomarker. Circ_0000520 is aberrantly upregulated in breast cancer tissues and cells. Survival analysis revealed that patients with high circ_0000520 expression had a significantly lower 5-year overall survival rate than those in the low-expression group (63.3% vs. 86.7%), indicating that circ_0000520 acts as a powerful prognostic predictor for breast cancer patients. ⁹² CircPVT1 is significantly overexpressed in breast cancer tissues and cell lines. Kaplan-Meier survival analysis showed that patients with high circPVT1 expression exhibited markedly worse DFS than those with low expression (χ²=7.174, p=0.007), and a similar trend was observed for OS (χ²=3.946, p=0.047). High expression of hsa_circ_0086735 is significantly correlated with worse overall survival (OS) and shorter distant metastasis-free survival (DMFS) in patients with luminal-type breast cancer (log-rank P<0.05). Multivariate Cox regression analysis further confirmed that hsa_circ_0086735 serves as an independent adverse prognostic factor (HR=11.889, P=0.007; HR=6.945, P=0.004), suggesting that its overexpression reflects an unfavorable clinical prognosis in patients. ⁴⁰ Circ_0001522, circ_0001278 and circ_0001801 are upregulated in TNBC and correlated with poor patient prognosis. Kaplan-Meier survival analysis demonstrated that high expression of the three circRNAs was significantly associated with inferior 10-year relapse-free survival in TNBC patients (log-rank test, P < 0.05). ⁹³ CircRNA prognostic biomarkers can be used to guide adjuvant treatment decision-making. Patients in the high-risk group may benefit from more aggressive adjuvant chemotherapy or combined therapy, while those in the low-risk group can avoid overtreatment. Prospective clinical studies are ongoing to evaluate the value of circRNA prognostic models in guiding adjuvant treatment decisions for TNBC.

5.3 Predictive Biomarkers

The expression of circRNAs is correlated with the response of breast cancer to specific therapies, which can serve as predictive biomarkers to guide individualized treatment selection. CircPVT1 is highly expressed in ERα-positive breast cancer. It upregulates ESR1 by sponging miR-181a-2-3p and inhibits the type I interferon signaling pathway, thereby facilitating tumorigenesis and inducing resistance to endocrine therapy and tamoxifen. Antisense oligonucleotides targeting circPVT1 can suppress tumor growth and reverse drug resistance, and its high expression can serve as a predictive biomarker for the efficacy of endocrine therapy. ⁶¹ In estrogen receptor-positive breast cancer patients receiving adjuvant tamoxifen therapy, patients with high circRNA-SFMBT2 expression exhibit a significantly increased risk of recurrence. Its expression level is closely correlated with larger tumor size and poor prognosis, indicating that this circRNA is a potential biomarker for predicting the recurrence risk of ER+ breast cancer patients after tamoxifen treatment. ⁵³ In HER2-positive breast cancer, the circular RNA circ-β-TrCP is upregulated, and its encoded protein isoform can drive trastuzumab resistance, patients with its high expression present a worse response to trastuzumab treatment and are correlated with poor prognosis, suggesting that it can act as a novel biomarker for predicting trastuzumab resistance. ⁹⁴

In terms of chemotherapy sensitivity prediction, CircPLK1 is significantly upregulated in non-pCR TNBC patients and is associated with resistance to anthracycline-based chemotherapy, its high expression can promote doxorubicin resistance in TNBC cells and reduce patient survival, suggesting that CircPLK1 can serve as a potential biomarker and therapeutic target for predicting the response to anthracycline-containing chemotherapy. ⁹⁵ Circ_0006528 is significantly upregulated in paclitaxel-resistant breast cancer tissues and cells, its high expression is an independent risk factor for patient prognosis (HR=2.22, 95%CI=1.14–4.51, P<0.05), and knockdown of this molecule can reduce the IC50 of paclitaxel and enhance drug sensitivity, which has important clinical value in predicting the efficacy of paclitaxel regimens for breast cancer. ⁹⁶ Based on these findings, circRNA detection can be incorporated into the treatment decision-making process for breast cancer to screen patients who may benefit from specific therapies.

5.4 Application in Liquid Biopsy

Liquid biopsy has become an important component of precision oncology due to its advantages such as minimal invasiveness and repeatable sampling. Circulating circRNAs possess unique advantages as liquid biopsy biomarkers: 1) high stability — owing to their covalently closed circular structure, circRNAs are inherently resistant to RNA exonucleases, maintain intracellular stability, and can be reliably detected in body fluids such as plasma and serum, 2) tissue specificity — the expression profiles of circRNAs exhibit tissue and temporal specificity, enabling tumor origin tracing and reflection of molecular characteristics, 3) dynamic alteration — circRNA expression fluctuates dynamically with tumor burden and therapeutic response, making them suitable for non-invasive real-time monitoring of treatment efficacy and tumor recurrence. ⁹⁷

5.4.1 Plasma/Serum circRNA

Multiple studies have evaluated the diagnostic value of plasma/serum circRNA in breast cancer. Plasma hsa_circ_0001785 is significantly upregulated in breast cancer patients, yielding a diagnostic AUC of 0.784 in a validation cohort of 57 samples. ⁹¹ Serum circ-FAF1 is significantly downregulated in breast cancer patients, and its AUC for distinguishing breast cancer patients from healthy controls is 0.885, with a diagnostic sensitivity of 81.67% and a specificity of 76.67%. ⁹⁸ Circ-FOXO3 is significantly downregulated in blood samples of TNBC patients, and its low expression is correlated with lymph node metastasis and poor prognosis, suggesting that circ-FOXO3 can serve as a potential prognostic biomarker and diagnostic target for TNBC. ⁹⁹ Plasma hsa_circ_0000091 expression in breast cancer patients is positively correlated with axillary lymph node metastasis, and its expression level is markedly higher in metastatic patients than in non-metastatic patients, the combined diagnosis of axillary lymph node metastasis using this indicator combined with ultrasound achieves an AUC of up to 0.80. ⁸⁹ Numerous additional breast cancer studies have also explored plasma/serum circRNAs as circulating tumor biomarkers, indicating that plasma exosomal circRNAs may serve as efficient and non-invasive novel biomarkers for the prognostic evaluation of breast cancer.

Dynamic changes of plasma circRNA can reflect therapeutic response. In the peripheral blood of breast cancer patients, the expression level of circRNA-000284 is significantly downregulated after chemotherapy, while the expression of circ-ITCH shows no statistically significant difference before and after chemotherapy, suggesting that different circRNAs exhibit heterogeneous response patterns to chemotherapy. ¹⁰⁰ hsa_circ_0001785 is significantly upregulated in the plasma of breast cancer patients, and the plasma level in postoperative patients (0.283±0.043) is significantly lower than that before surgery (0.109±0.037) (P<0.01). Continuous monitoring of the dynamic fluctuations in plasma circRNA levels is expected to be applied to postoperative surveillance. ⁹¹

5.4.2 Extracellular Vesicle circRNA

EVs circRNA is protected by the vesicle membrane, rendering it more stable in the circulatory system and achieving higher detection sensitivity. A study based on 190 clinical samples demonstrated that the expression level of serum exosomal hsa_circ_0000615 was significantly higher in breast cancer patients than in healthy controls (P<0.01). Its AUC for distinguishing breast cancer patients from healthy controls was 0.904 (95% CI: 0.863–0.944), with a sensitivity of 76.8% and a specificity of 88.4%. ¹⁰¹ CircZCCHC2 is significantly upregulated in plasma exosomes of TNBC patients. Its area under the ROC curve for differentiating TNBC from non-TNBC patients is 0.787 (p < 0.0001). At the optimal cut-off value of 0.471, the diagnostic sensitivity and specificity are 76.5% and 70.6%, respectively. ¹⁰² hsa_circ_0058514 expression in plasma extracellular vesicles is significantly correlated with the risk of recurrence and metastasis in breast cancer patients and can assist in evaluating the efficacy of neoadjuvant therapy, suggesting that dynamic changes of plasma circRNA can serve as an effective liquid biopsy tool for monitoring tumor recurrence and metastasis. ¹⁰³ In contrast, a meta-analysis including 21 studies revealed that the diagnostic AUC based on total plasma circRNAs was only 0.67. ¹⁰⁴ EVs as a liquid biopsy platform can effectively enrich circRNA molecules related to diagnosis, and their diagnostic performance is superior to the direct detection of total plasma circRNAs.

5.4.3 CircRNAs Derived From Other Body Fluids

In addition to blood, circRNAs derived from other body fluids are also under investigation. CircRNAs exhibit specific dynamic expression in the lactating mammary gland, and their dysregulation can promote the occurrence and development of early breast cancer, suggesting that circRNAs can serve as potential biomarkers and therapeutic targets for mammary gland diseases and lactation function. ¹⁰⁵ Existing reviews have noted that circRNAs can stably exist in urine, and their expression patterns differ significantly between cancer patients and healthy controls. Changes in urinary circRNA levels may reflect the systemic tumor burden status. ¹⁰⁶ In terms of saliva, exosomal circRNAs can be stably detected in saliva due to their stability and tissue specificity, and can serve as non-invasive diagnostic biomarkers for breast cancer. ¹⁰⁷

This chapter has systematically reviewed the clinical evidence for circRNAs as diagnostic, prognostic, and predictive biomarkers, as well as their application in liquid biopsy. To provide an integrated overview of the entire translational pathway from discovery to clinical implementation, Figure 3 summarizes the five developmental stages outlined above, while also highlighting the major bottlenecks and future technological directions that will be discussed in the following chapter (see Figure 3).

Figure 3.

Clinical translation roadmap for circRNA biomakers

6. Challenges and Prospects

6.1 Current Challenges

6.1.1 Standardization Issues in Testing

The standardization of circRNA detection is a prerequisite for clinical translation. Current commonly used detection methods in circRNA research include quantitative real-time PCR (qRT-PCR), microarray chips and RNA sequencing, yet each method has its own advantages and disadvantages, and the lack of standardized experimental procedures and normalization strategies limits the comparability of results across different platforms. ¹⁰⁸ In the qRT-PCR detection of circRNA, unified specifications are required for the selection of reference genes, primer design especially for specific divergent primers targeting back-splicing junction sites, and data normalization methods. ¹⁰⁹ International experts jointly published consensus guidelines in Nature Methods, calling for an urgent establishment of universal experimental standards for circRNA research and proposing full-process best practice specifications covering circRNA enrichment and purification, identification and validation, quantitative analysis, expression regulation, and data interpretation, aiming to promote the sound development of this field and improve the comparability and reproducibility of research findings. ¹¹⁰

6.1.2 Insufficient Depth of Functional Mechanism Research

Although a large number of differentially expressed circRNAs have been identified, the functional mechanisms of most circRNAs remain unclear. CircRNAs exert their functions through multiple mechanisms including acting as miRNA or protein inhibitors (“sponges”), regulating protein functions, and being translated themselves. A single circRNA may possess multiple functional patterns simultaneously, and such functional diversity increases the research complexity of mechanism interpretation. ¹¹¹ One important reason lies in the lack of efficient and universal circRNA knockout/knockdown methods, tools such as siRNA, DNAzyme and CRISPR-Cas system have been gradually developed, but existing methods still struggle to distinguish circRNAs from their linear homologous mRNAs and have limited application scope, therefore, it is urgent to develop novel research tools such as circRNA-specific knockout and overexpression systems as well as functional screening platforms to systematically dissect the functional landscape of circRNAs. ¹¹² Meanwhile, integrative analysis of circRNA regulatory networks—including circRNA-miRNA-mRNA networks and circRNA-protein interaction networks—contributes to a comprehensive understanding of their overall roles in maintaining cellular homeostasis and the occurrence and progression of diseases. ¹¹³

6.1.3 Demand For Large-Sample Verification

Current studies on circRNAs as biomarkers are mostly single-center and small-sample retrospective researches with limited evidence level. ¹¹⁴ Large-sample, multicenter and prospective clinical studies are required to verify the diagnostic and prognostic value of circRNA biomarkers. Study design shall comply with reporting guidelines such as the REMARK guidelines and clearly define the study population, sample size calculation, statistical analytical methods and validation strategies. ¹¹⁵ For prognostic biomarkers, it is necessary to validate their independent prognostic value in independent cohorts and compare them with existing clinicopathological factors and molecular subtypes.

6.1.4 Challenges of Delivery Systems

As therapeutic targets or tools, the in vivo delivery of circRNAs is a major challenge. As a member of nucleic acid macromolecules, circRNA possesses inherent physicochemical properties of such macromolecules, including large molecular weight, hydrophilicity and negative charge, which render it difficult to directly penetrate the phospholipid bilayer of cell membranes. ¹¹⁶ Although recombinant adeno-associated virus (rAAV) holds promising application prospects for the delivery of exogenous genes in gene therapy, it has potential risks of preferential intragenic integration accompanied by chromosomal deletions and the induction of loss-of-function insertional mutations. This suggests that when using rAAV to deliver circRNAs for gene therapy, it is necessary to carefully evaluate its risks of insertional mutations and carcinogenicity. ¹¹⁷ In recent years, non-viral nanocarrier delivery systems are being developed at an accelerated pace. Delivery platforms such as liposomes, polymer nanoparticles, and extracellular vesicles can stabilize nucleic acids, regulate biodistribution, and achieve organ- and cell-selective targeting, demonstrating significant potential in the clinical translation of RNA therapies, including circRNA therapy. ¹¹⁸ On the path of clinical application of circRNA technologies such as CRISPR-Cas13, in addition to addressing standardization issues, the development of safe and efficient in vivo delivery systems is one of the key challenges they are facing. ⁵

6.1.5 Limitations of This Review

As a narrative review, this work has several inherent limitations that should be acknowledged. First, although we performed a comprehensive literature search across multiple databases, we did not conduct a formal systematic review with predefined inclusion and exclusion criteria, nor did we perform a quality assessment of the included studies using standardized tools such as QUADAS-2 or the Newcastle-Ottawa Scale. ¹¹⁹ Consequently, the selection of cited studies may be subject to selection bias, and the conclusions drawn are inherently qualitative rather than quantitative. Second, the published literature on circRNA biomarkers is heavily skewed toward positive findings, with studies reporting high diagnostic AUC values or significant prognostic HRs being more likely to be published than those with null or negative results. ¹²⁰ This publication bias may inflate the perceived clinical utility of circRNA biomarkers and obscure the true distribution of their performance across different cohorts. Third, our review predominantly focuses on well-characterized circRNAs that have been repeatedly studied (e.g., ciRS-7, circANKS1B, circCFL1), potentially overlooking rare, low-abundance, or tissue-specific circRNAs that may have equally important but underappreciated roles in breast cancer biology. ¹²¹ The functional relevance of most differentially expressed circRNAs remains unvalidated, and many of the reported associations may represent passenger events rather than causal drivers, as discussed in Section 2.4. Fourth, the majority of studies summarized in this review originate from single-center, retrospective analyses with limited sample sizes, and the heterogeneity in detection methods (qRT-PCR vs. RNA-seq), normalization strategies, and sample processing protocols complicates cross-study comparisons and meta-analyses. ¹²² Finally, the rapid pace of technological advances in circRNA detection and functional screening means that some of the findings discussed here may become outdated as new high-throughput platforms and analytical tools emerge. Despite these limitations, we believe this review provides a balanced and critical overview of the current state of circRNA biomarker research in breast cancer, while clearly highlighting the translational gaps that must be addressed in future studies.

6.2 Future Research Directions

6.2.1 Technological Innovation

Single-cell RNA sequencing (scRNA-seq) has been extended to circRNA detection as sc-circRNA-seq, which enables the profiling of circRNA expression at the single-cell level and reveals tumor heterogeneity as well as circRNA regulatory networks specific to cell subpopulations. ¹²³ Spatial transcriptomics technology enables the profiling of in situ circRNA expression in tissues and provides spatial distribution information. ¹¹⁹ These technologies will deepen the understanding of the roles of circRNAs in the tumor microenvironment.

Li et al. constructed a high-throughput gRNA lentiviral library targeting back-splicing sites via the CRISPR-Cas13 system, which enables efficient genome-wide screening of functional circRNAs, successfully identifies circFAM120A (pro-proliferative) and circMan1a2 (regulating embryonic development), and accelerates functional circRNA research. ¹²⁰ Various machine learning algorithms built on graph neural networks and deep learning have been successfully developed for circRNA research. These methods can effectively integrate multi-omics biological data and construct heterogeneous molecular association networks, thereby realizing circRNA-disease association prediction, circRNA function inference, and accelerating the discovery of circRNAs as therapeutic targets or related drugs.^121,122

6.2.2 Multi-Omics Integration Analysis

The research on circRNAs is evolving from single-omics analysis to multi-omics integration analysis. Integrating multi-dimensional high-throughput data such as circRNA expression profiles and proteomes helps construct ceRNA regulatory networks (e.g., circRNA-miRNA-mRNA). Integration analysis based on transcriptomics and network biology can comprehensively clarify the regulatory mechanisms of circRNAs, and in-depth cross-fusion with cutting-edge technologies such as single-cell sequencing will enable systematic analysis of the mechanisms of circRNAs in breast cancer. ¹²⁴ CircRNA-protein interactomics (circRNA pull-down combined with mass spectrometry) can identify circRNA-binding proteins and reveal the functions of circRNAs at the protein level. ¹²⁵

Multi-omics integration analysis can also be used to discover new biomarker combinations. Studies have shown that in breast cancer, circRNAs can be combined with miRNAs, lncRNAs, and traditional protein markers (such as CEA and CA-153). By constructing multi-molecular regulatory networks (e.g., the circ_0065214/miR-188-3p/GPNMB axis or circRNA/lncRNA-miRNA-transcription factor-mRNA networks), they can significantly improve the diagnostic efficiency of breast cancer and effectively predict neoadjuvant therapy response and prognosis.^126,127 Integrating circRNA expression profiles with molecular typing to construct prognostic models can achieve precise individualized risk stratification for breast cancer. Subtype-specific m6A epitranscriptomic studies can screen out m6A circRNA methylation characteristics specific to TNBC and luminal breast cancer, and multiple circRNAs derived from host genes are significantly associated with patients’ survival time and have independent prognostic value, providing a new basis for improving traditional risk stratification and individualized precise prognosis assessment. ¹²⁸

6.2.3 Clinical Translation Research

The clinical translation of circRNAs requires a leap from basic research to clinical application. The five-stage biomarker validation framework of EDRN provides a systematic pathway for biomarker translation: retrospective exploratory studies identify candidate biomarkers; independent cohort validation of diagnostic efficacy; optimization of detection methods and development of kits; conduct of prospective studies to evaluate clinical screening value; and finally verification of its improvement effect on clinical decision-making and patient prognosis through clinical trials. ¹²⁹ The transformation path of prognostic biomarkers is similar, but it requires longer-term follow-up and large-scale validation.

The transformation path of therapeutic targets based on circRNAs includes: target validation (confirming its key role in tumor occurrence and development through functional research); targeted strategy development (designing small-molecule inhibitors, antisense oligonucleotides, RNA vaccines, etc.); preclinical research (evaluating drug efficacy, pharmacokinetics, and toxicological safety); and advancing to clinical trials to verify the safety and effectiveness of the treatment. ¹³⁰ As a new antigen, circRNA-encoded polypeptides can be used to develop personalized tumor vaccines. Preclinical studies have confirmed that they can activate anti-tumor immune responses in animal models, and these studies have entered the early clinical verification stage. ⁶⁴ Future efforts should prioritize prospective, multicenter registries with standardized pre-analytical and analytical protocols, as recommended by the ERIN (European Network for circRNA) initiatives.

6.2.4 Artificial Assistance in Clinical Applications

Artificial Intelligence (AI) technology has broad application prospects in the clinical translation of circRNAs. Integrated analytical frameworks based on deep learning, such as IntRNA, can accurately identify circRNAs from RNA-seq data, systematically evaluate their coding potential, and assist in predicting their biological functions. ¹³¹ AI-assisted liquid biopsy analysis platforms can integrate markers (including various types of RNAs such as circRNAs) in tumor-derived extracellular vesicles, enhancing the diagnostic accuracy and detection efficiency of early breast cancer. ¹³² Studies have shown that AI-assisted decision-making systems integrating circRNA expression profiles with machine learning technologies can effectively integrate multi-dimensional data such as circRNA expression profiles, providing a new strategy for non-invasive diagnosis and management of breast cancer, as well as decision support for the selection of individualized treatment regimens. ⁹⁰

6.3 Personalized Treatment Strategies

Based on circRNA, personalized treatment strategies are the future development direction. Based on the patient’s circRNA expression profile, circRNA-encoded polypeptide expression status, tumor molecular typing, and immune microenvironment status, the most appropriate treatment plan should be selected. For example, TNBC patients with high circCFL1 expression are expected to benefit from targeted intervention strategies such as ASOs targeting circCFL1 ⁴⁴ ; For TNBC patients positive for circ-HER2/HER2–103 expression, pertuzumab may be more suitable for their treatment because it can effectively inhibit the tumorigenicity of circ-HER2/HER2-103-positive TNBC ¹³³ ; The circRNA molecular profile carried by EVs is closely related to the organ-specific metastatic tendency of breast cancer. The molecular heterogeneity of different subtypes of EVs provides new clues for identifying potential metastasis-predictive biomarkers, which is expected to be used for stratified management of high-risk patients and provides a translational direction for future precise diagnosis and treatment strategies. ⁸¹

Combination therapy strategies are the key to improving treatment efficacy. CircRNA-targeted therapy can be combined with chemotherapy, targeted therapy and immunotherapy to exert synergistic effects. CircRNA-targeted therapy can enhance the sensitivity of conventional treatments by reducing the proportion of cancer stem cells, reversing epithelial-mesenchymal transition (EMT), and inhibiting immune escape. ¹³⁴ The combined application of immunotherapy and circRNA vaccines can significantly enhance antigen-specific immune responses, produce synergistically enhanced anti-tumor immune effects, and thereby induce a more potent anti-tumor immune response. ¹³⁵

7. Conclusion

CircRNAs represent a burgeoning class of non-coding RNAs with substantial potential to advance the precision oncology landscape of breast cancer. Their unique structural stability, tissue-specific expression patterns, and detectability in diverse body fluids position them as ideal candidates for non-invasive biomarker development. Accumulating evidence has established robust associations between dysregulated circRNA expression and critical clinicopathological features, including molecular subtypes, metastatic propensity, and therapeutic resistance, providing a strong rationale for their further evaluation as diagnostic, prognostic, and predictive tools.

Beyond their biomarker potential, circRNAs are increasingly recognized as active participants in breast cancer pathogenesis through diverse mechanisms, ranging from miRNA sponging and protein scaffolding to transcriptional regulation and micropeptide translation. The discovery of immunogenic circRNA-encoded peptides has further expanded their therapeutic relevance. While extracellular vesicle-derived circRNAs offer exciting opportunities for real-time monitoring of organ-specific metastasis and treatment response, their clinical adoption remains contingent upon overcoming substantial barriers, including detection standardization, rigorous functional validation, and large-scale prospective studies. Through coordinated efforts in technological innovation, multi-omics integration, and collaborative clinical research, circRNAs are poised to become integral components of next-generation diagnostic and therapeutic strategies for breast cancer.

ORCID iD

Xin Ma https://orcid.org/0009-0006-1599-883X

Ethical Considerations

This is a review article; ethics approval and consent to participate are not applicable to this work.

Consent to Participate

This is a review article; ethics approval and consent to participate are not applicable to this work.

Data Availability Statement

This is a review article; the data sources are the references cited in the text.*