The appeal of circular RNA begins with topology. Conventional linear messenger RNA, the platform behind the COVID-19 vaccines, is vulnerable to exoribonucleases—ubiquitous enzymes that chew RNA from its free 5-prime and 3-prime ends, degrading the molecule within hours. Circular RNA eliminates those ends entirely. The closed loop resists exonucleases, extending cytoplasmic residence from hours to days, and its topology can promote additional hairpin structures and long-range pseudoknots that form a protective ring around the molecule. The review notes that genome-wide analyses of cell-free human saliva identified more than 400 distinct circular RNA species, a striking demonstration of their durability even in nuclease-rich environments. That stability translates directly into therapeutic advantage: a single administration can sustain protein expression for a week or more, reducing dosing frequency, toxicity concerns, and manufacturing burden.

Stability, however, is only half the equation. Circular RNA lacks the 5-prime cap and 3-prime poly(A) tail that give linear mRNA its extraordinary translational efficiency, so it must rely on internal ribosome entry sites, or IRESs, to recruit ribosomes in a cap-independent manner. The review traces how viral-derived IRESs from picornaviruses such as encephalomyocarditis virus and foot-and-mouth disease virus drive strong translation but carry penalties: their 450 to 500 nucleotide length increases precursor complexity, impairs circularization efficiency, and elevates levels of linear RNA impurities. Cellular-derived IRESs, including short 40 to 60 nucleotide stem-loop structures, simplify design and improve circularization yields, but their activity is often modest and many require strict sequence preservation, such as the full-length 216 nucleotide Hsp70 IRES. The field, the authors argue, still lacks a universal IRES that balances high activity, short length, and manufacturing compatibility—a trade-off that currently forces designers to choose between translatability and scalable production.

Emerging cap-mimicking strategies offer a way around the IRES bottleneck. Covalent ligation approaches connect an m7G cap or its analogs to circular RNA through branched structures or chemical synthesis, and reported results are dramatic: m7G-capped circular RNA achieved over 100-fold higher NanoLuciferase expression in HeLa cells compared to uncapped counterparts, with 6-nucleotide branched strands showing optimal efficiency. Related work with internal branched caps gave 115-fold higher expression, and LNAm G/LNA/2-prime-OMe QRNA designs gave 144-fold higher expression. Non-covalent alternatives use cap-modified complementary oligonucleotides that hybridize to the circular RNA without disrupting its closed loop, enhancing translation efficiency by over 50 times even without an IRES. Yet both approaches carry manufacturing costs: covalent ligation requires complex chemical synthesis or enzymatic reactions, while non-covalent complexes are prone to dissociation in vivo, limiting the durability of the translational enhancement.

Nucleoside modifications add another layer of engineering control. The review distinguishes carefully between evidence from endogenous circular RNAs and direct engineering experiments on synthetic molecules. In colorectal cancer, the methyltransferase METTL1 catalyzes m7G modification of circKDM1A, preventing its degradation; in hepatocellular carcinoma, the same modification strengthens circIPP2A2’s resistance to RNase R. In lung cancer cells, NSUN4-mediated m5C modification promotes nuclear export of circERI3. These endogenous mechanisms suggest candidate modifications for synthetic platforms, but the authors caution that they do not by themselves establish validated stability strategies for therapeutic circular RNA. More direct evidence comes from synthetic systems: incorporation of 5 percent m6A maintained protein translation and increased resistance to nuclease degradation in fetal bovine serum assays, and recent experiments showed that 5-methylcytosine-modified circular RNA maintained rolling-circle translation while significantly reducing pro-inflammatory cytokines in cultured cells and in mice.

Sequence engineering complements chemical modification. Codon optimization leverages the degeneracy of the genetic code to improve stability and protein expression without altering the encoded amino acid sequence. Aligning codon usage with host cell tRNA abundance smooths translation and reduces ribosome stalling, while a higher codon adaptation index minimizes translation defects that would otherwise trigger nonsense-mediated decay, no-go decay, or non-stop decay. The circDesign algorithm, developed by Xu and colleagues, balances codon adaptation index, minimum free energy, and IRES structure to reduce base-pairing interference between the coding sequence and the IRES. When applied to a circular RNA vaccine targeting rabies virus glycoprotein, the optimized construct showed 1.5-fold higher translation efficiency, and combined with a lipopolyplex delivery system, it induced a stronger immune response. The review also highlights how the ribonuclease DIS3 preferentially degrades circular RNAs containing U-rich motifs, and how computational tools like mRNAdesigner can reduce multi-branch loops and exposed single-stranded regions—though designs optimized for linear RNA limit efficacy in circular constructs.

Delivery remains a critical determinant of therapeutic performance. Lipid nanoparticles optimized for linear mRNA are suboptimal for circular RNA because of weaker electrostatic affinity with cationic lipids and distinct physicochemical properties of the circular molecule. Recent work by Alshehry and colleagues adopted stepwise component screening of ionizable lipids, helper lipids, and sterol components, achieving significantly enhanced transfection efficiency in immune cells and stronger antigen-specific immune responses after vaccination in mice. Swingle and colleagues went further, screening novel high-efficiency ionizable lipids and iteratively optimizing formulation parameters through design-of-experiments methodology, realizing efficient circular RNA transfection in immune cells and stronger specific immunity than mRNA-LNP. Polymeric nanoparticles such as the reducible and degradable pABOL have also shown promise: under identical formulation conditions, circular RNA delivered via pABOL produced markedly higher expression duration and levels than linear mRNA, with a 5 microgram dose enhancing expression by 96.2-fold compared to only 11.7-fold and 1.2-fold increments for two linear mRNA constructs. Exosomes, naturally enriched in circular RNA, offer a third delivery route, though loading efficiency and large-scale production remain bottlenecks.

The clinical pipeline is advancing rapidly. Qu and colleagues developed a circular RNA vaccine expressing the trimeric receptor binding domain of the SARS-CoV-2 spike protein that effectively protected mice and rhesus monkeys from the virus and its variants. In cancer immunotherapy, personalized neoantigen circular RNA vaccines encode tumor-specific antigens identified by comparing tumor with matched normal tissue; a recent example induced strong CD8-positive T cell responses, significantly inhibiting tumor growth and prolonging survival in treated mice, with more durable antigen expression and stronger immune activation than conventional mRNA vaccines. In protein replacement, a single dose of VEGF-A circular RNA encapsulated in the ionizable lipid U-105 sustained expression for more than one week, leading to almost complete wound healing by day 12 in diabetic foot ulcer models—significantly superior to recombinant human VEGF protein injection and linear VEGF-A mRNA. For osteoarthritis, in vitro transcribed circular RNA encoding Musashi2 or SOX5 delivered via SM-102 lipid nanoparticles significantly delayed disease progression, with protein expression lasting more than seven days and no obvious immunogenicity.

In vivo CAR-T therapy represents another frontier. Traditional in vitro CAR-T requires isolating T cells, activating them, performing viral vector-mediated gene modification, expanding the modified cells, and infusing them back—a process that is time-consuming, costly, and reliant on professional production facilities. Circular RNA encoding chimeric antigen receptors delivered through targeted lipid nanoparticles can achieve sustained CAR expression and prolonged anti-tumor activity. Wang and colleagues intravenously injected SORT-circRNACAR into mice bearing colon cancer tumors, significantly inhibiting tumor growth; at 30 days after injection, mean tumor size in treated mice was only about half that of the control group. Intratumoral injection of LNP-circRNACAR in 4T1 tumor-bearing mice also significantly inhibited growth and even achieved tumor regression. The review notes that HM2002, developed by CirCode Biologics for ischemic heart disease, obtained Investigational New Drug approvals from both China and the United States in 2025, while RXRG001 from RiboX Therapeutics, encoding human aquaporin 1 for radiation-induced xerostomia, received FDA IND approval in 2024 and completed first patient administration in early 2025.

Artificial intelligence is reshaping the design process itself. The circDesign algorithm integrates codon adaptation index, minimum free energy, and IRES structure to screen circular RNA with higher stability and translatability, verifying that AI-guided maintenance of folding independence is a key strategy for designing high-expression constructs. trRosettaRNA uses a transformer network to predict RNA 3D structure, exposing structural distortions such as distorted circular structures caused by suboptimal junction placement. The BCMCMI system employs the BERT algorithm to predict circular RNA-miRNA interactions, while circSLNN uses neural networks to localize protein-binding sites along full-length sequences. Beyond protein coding, engineered non-coding circular RNAs such as circSCMH1 for ischemic stroke and circPOLR2A for inflammatory skin disorders and Alzheimer’s disease are broadening therapeutic boundaries. The authors emphasize that future priorities include integrating circular topology into algorithm training, expanding datasets to incorporate clinically relevant parameters such as manufacturing yield and in vivo immunogenicity, and validating predictions in translational models. As the field navigates the balance between stability, translational efficiency, immunogenicity, and manufacturing feasibility, the transition from promising platform to clinically validated medicine will depend on integrated engineering that treats the circular RNA molecule, its delivery carrier, and its production process as a single, functionally coherent therapeutic entity.

Subject of Research: Enhancing the stability and translatability of synthetic circular RNA: design strategies and emerging challenges

Article Title: Enhancing the stability and translatability of synthetic circular RNA: design strategies and emerging challenges

Article References: Ye, Z., Tan, X., Xu, C., Qi, S., Chen, R., Zhang, Y., Pi, F., Chou, H.-L., Gao, P., & Xu, C. (2026). Enhancing the stability and translatability of synthetic circular RNA: design strategies and emerging challenges. Frontiers in Biomaterials Science, 5, Article 1981461. https://doi.org/10.3389/fbiom.2026.1981461

Image Credits: AI Generated

DOI: 10.3389/fbiom.2026.1981461

Keywords: Enhancing, stability, translatability, synthetic, circular, design, strategies, emerging, challenges, scientific research