In the traditional view of the central dogma, RNA molecules are linear vectors of information, defined by distinct 5' caps and 3' poly-A tails. However, a massive class of non-coding transcripts upends this geometry: circular RNAs (circRNAs). Formed via a non-canonical splicing event known as "back-splicing"—where a downstream splice donor covalently links back to an upstream splice acceptor—these molecules form a continuous, closed loop.
This structural difference changes how these molecules behave inside a cell. Because they lack free 5' and 3' ends, circRNAs are entirely immune to exonucleases, the cellular enzymes responsible for chewing up and recycling linear RNA. This structural resilience gives circRNAs an exceptional half-life, allowing them to persist and accumulate as stable regulatory hubs in cellular environments where linear transcripts rapidly degrade.
The Sponge Mechanism and ceRNA Networks
The most prevalent way circRNAs exert control is by serving as competitive endogenous RNA (ceRNA) networks, a mechanism colloquially known as "miRNA sponging."
To understand why this works, consider the standard lifecycle of a microRNA (miRNA). MiRNAs are short, linear non-coding RNAs that bind to complementary sequences on target messenger RNAs (mRNAs), marking them for translational repression or degradation. Under normal conditions, this keeps specific proteins within a strict homeostatic range.
[Normal State]
miRNA ──> Binds & Represses ──> Target mRNA ──> Low Protein Expression
[Sponged State]
circRNA (Multiple Binding Sites) ──> Sequesters miRNA
│
└─> Target mRNA Released ──> High Protein Expression
A circRNA can evolve to possess multiple, highly conserved binding sites for a specific miRNA. Because the circRNA loop is so stable, it acts as a high-capacity molecular sponge, sequestering free miRNAs away from their intended targets. When the miRNAs are trapped on the circular loop, the corresponding linear mRNAs are freed from repression, causing a sharp uptick in downstream protein expression.
This single mechanism can drive complex pathologies like cancer progression by selectively turning off tumor suppressors or turning up oncogenes:
- Osteosarcoma: Xin Jin et al. demonstrated in Circular RNA hsa-circ-0016347 promotes proliferation, invasion and metastasis of osteosarcoma cells that hsa-circ-0016347 sponges miR-214, which directly releases the inhibition on caspase-1, accelerating tumor proliferation and metastasis.
- Triple-Negative Breast Cancer: Lei Xing et al. showed in The circRNA circIFI30 promotes progression of triple-negative breast cancer and correlates with prognosis that circIFI30 acts as a sponge for miR-520b-3p, leading to an overproduction of the oncogene CD44.
- Gastric Cancer: The modularity of this system allows a single circRNA to influence multiple pathways simultaneously. As detailed by Guifang Xu et al. in Circular RNA CCDC66 promotes gastric cancer progression by regulating c-Myc and TGF-β signaling pathways, circCCDC66 drives gastric cancer by concurrently activating the c-Myc and TGF-β signaling cascades.
This sponging behavior is not unique to oncology; it serves as a fundamental dial for cellular stress. For example, Xiang Li et al. mapped these networks across different stages of pressure overload-induced heart failure in Differential mRNA Expression and Circular RNA-Based Competitive Endogenous RNA Networks in the Three Stages of Heart Failure in Transverse Aortic Constriction Mice, showing how cells reconfigure their circular RNA profiles to manage metabolic and mitochondrial strain over time.
Beyond Sponging: Protein Scaffolding and Epitranscriptomics
While miRNA sponging relies on RNA-to-RNA interactions, circRNAs also interact directly with proteins and undergo chemical modifications that alter their functional behavior.
Instead of acting as a sponge, a circRNA can act as a physical shield or scaffold for a protein. In Circular RNA FEACR inhibits ferroptosis and alleviates myocardial ischemia/reperfusion injury by interacting with NAMPT, Jie Ju et al. discovered that the circRNA FEACR directly binds to and stabilizes the enzyme NAMPT. By preventing the degradation of NAMPT, FEACR triggers a protective signaling cascade that suppresses ferroptosis—an iron-dependent form of cell death—during acute cardiac events.
[Protein Stabilization Mechanism]
circRNA + Target Protein ──> Covalent/Structural Shielding ──> Prevents Degradation ──> Enhanced Pathway Activity
Furthermore, circRNAs are subject to the same epitranscriptomic modifications as linear RNAs, which fundamentally alters how they interact with other cellular machinery. Juanyi Shi et al. explored this in m 6 A‐Modified circRAPGEF1 Interaction with IGF2BP3 Promotes Hepatocellular Carcinoma Progression via Reprogramming Aspartate Metabolism. They found that an $N^6$-methyladenosine ($m^6A$) modification on circRAPGEF1 dictates its ability to bind the RNA-binding protein IGF2BP3. This modification-dependent interaction destabilizes specific metabolic mRNAs, effectively reprogramming aspartate metabolism to sustain liver cancer stem cells.
The Macro-Scale: Extrachromosomal Circular DNA
The biological utility of circular architecture is not restricted to non-coding transcripts; it also manifests at the genomic level as extrachromosomal circular DNA (ecDNA).
In Extrachromosomal circular DNA orchestrates genome heterogeneity in urothelial bladder carcinoma, Wei Lv et al. mapped how large, circular DNA loops detach from chromosomes to drive genetic diversity. Because these circular DNA elements lack centromeres, they do not segregate equally during cell division. Instead, they partition randomly into daughter cells, leading to rapid, asymmetric amplification of oncogenes and massive genomic heterogeneity within a single tumor.
Resolving the Functional Uncertainty
Despite the clear biophysical advantages of circularity, mapping these networks presents a significant challenge. Because circRNAs are spliced from the exact same genomic loci as their linear counterparts, they share identical sequence segments. This sequence overlap creates a "functional uncertainty," making it incredibly difficult to distinguish between linear and circular transcripts using standard sequencing workflows.
To resolve this, researchers are developing specialized computational and cataloging tools:
- Machine Learning Identification: Xin Wang et al. developed StackCirRNAPred: computational classification of long circRNA from other lncRNA based on stacking strategy, an ensemble machine learning framework designed to accurately differentiate long circRNAs from linear long non-coding RNAs based on sequence features across species.
- Systemic Repositories: To transition from isolated case studies to systemic biology, Ayesha Nisar et al. built CircAge: A Comprehensive Resource for Aging-associated Circular RNAs Across Species and Tissues, consolidating expression profiles and predicted functional interactions to map how these circular networks change across lifespans.
By combining structural stability, high-capacity sponging, and direct protein regulation, circular genetic elements function as highly resilient control hubs. Resolving the remaining gaps in these networks will be critical for designing targeted therapies that can disrupt these stable loops within diseased cells.
Key Foundational Work
CircHIPK3/miR-876-5p/PIK3R1 axis regulates regulation proliferation, migration, invasion, and glutaminolysis in gastric cancer cells (2020) — contributed 3 concept elements to this direction.
Novel evidence for oncogenic piRNA‐823 as a promising prognostic biomarker and a potential therapeutic target in colorectal cancer (2020) — contributed 2 concept elements to this direction.
CircR2Cancer: a manually curated database of associations between circRNAs and cancers (2020) — contributed 2 concept elements to this direction.