SciencePediaIn the complex world of gene expression, RNA molecules have long been seen as transient messengers, carrying genetic information from DNA to the protein-making factories of the cell. However, a fascinating and once-overlooked class of RNA has challenged this linear view: circular RNAs (circRNAs). These molecules are formed when the ends of a linear RNA are joined together, creating a covalently closed loop. For decades, they were dismissed as rare byproducts of splicing errors, but we now understand they are abundant, exceptionally stable, and play critical roles in cellular regulation. This article addresses the knowledge gap between their initial dismissal and their current status as key players in biology, exploring how their unique structure dictates their remarkable lifecycle and function.
This article will guide you through the intricate world of circRNAs in two parts. First, under "Principles and Mechanisms," we will delve into the molecular biology of these circles, exploring how they are made, why they are so stable, and the various ways they are regulated and eventually degraded. Following that, in "Applications and Interdisciplinary Connections," we will examine their diverse functions, the innovative methods developed to study them, and their profound implications in fields ranging from neuroscience to medicine.
Imagine you write a secret message on a long ribbon of paper. To keep it safe, you might worry about the ends fraying or someone tearing off a piece from the beginning or the end. What’s a simple, elegant solution? You could tape the two ends of the ribbon together to form a continuous loop. Now there are no ends to fray, no easy place to start tearing. The message is secure, stable, and long-lasting. It turns out that our cells, in their quiet and ceaseless wisdom, discovered this very trick billions of years ago. This is the essence of a circular RNA, or circRNA.
The cytoplasm of a cell is a bustling, chaotic place, teeming with enzymes whose job it is to clean house. Among the most diligent of these are exonucleases, molecules that act like relentless little Pac-Men, chomping away at RNA strands starting from their free ends—the so-called and termini. For a typical linear RNA molecule, this means its lifespan is often fleeting, measured in minutes or hours before it's degraded.
But circRNAs are different. By having their ends covalently linked into a closed loop, they lack the very "handles" that exonucleases need to grab onto. They are, in a fundamental structural sense, immune to this primary mode of RNA degradation. This simple topological feature is their superpower. It grants them extraordinary stability, allowing them to persist for days, far outlasting their linear counterparts.
Of course, this doesn't mean they are immortal. Nothing in the cell is. While exonucleases are thwarted, another class of enzymes called endonucleases can still attack them. Endonucleases are like molecular scissors that can cut the RNA strand at an internal site, independent of its ends. This initial "nick" linearizes the circRNA, instantly making it vulnerable to the waiting exonucleases. So, the story of a circRNA's life is a story of resisting the constant threat of exonucleases while evading the more specialized attacks of endonucleases. We'll see later just how this happens. But first, how does the cell even tie this molecular knot?
The formation of most circRNAs is a fascinating subversion of one of the cell's most fundamental processes: RNA splicing. When a gene is transcribed into a pre-messenger RNA (pre-mRNA), it's a long strand containing coding regions (exons) interrupted by non-coding regions (introns). The cell's splicing machinery, a massive complex called the spliceosome, acts like a film editor, snipping out the introns and stitching the exons together in a precise, linear order. For example, it joins the end of exon 1 to the start of exon 2, then the end of exon 2 to the start of exon 3, and so on. This is canonical splicing.
Every once in a while, however, the spliceosome makes a surprising and creative "mistake." Instead of joining an exon to the one immediately downstream, it can be guided to reach backwards and join the end of a downstream exon (say, the end of exon 3) to the beginning of an upstream exon (say, the start of exon 2). This event is called back-splicing. The result is that exons 2 and 3, along with any introns between them that are also spliced out, are released not as a linear piece of RNA, but as a covalently closed circle. The spliceosome, in a moment of non-canonical flair, has tied the knot.
Now, this isn't just a random event. For back-splicing to happen, the two splice sites—the one at the start of the upstream exon and the one at the end of the downstream exon—have to be brought close together in space. The cell has evolved two main strategies to accomplish this feat.
The first is a cis-driven model, where the instructions are written directly into the RNA sequence itself. Often, the introns flanking the exons that will be circularized contain long, complementary sequences, like two sides of a zipper. These intronic complementary sequences (ICS) can base-pair with each other, forming a stable double-stranded RNA stem that folds the pre-mRNA back on itself, bringing the required splice sites into perfect proximity for the spliceosome to act. We can test this idea in the lab. For instance, the enzyme ADAR1, which recognizes and edits double-stranded RNA, can disrupt this pairing and inhibit circRNA formation. Conversely, depleting a helicase like DHX9, which normally unwinds these RNA stems, can dramatically increase the production of these circRNAs.
The second strategy is a trans-driven model, which relies on protein intermediaries. Instead of the introns themselves pairing up, specific RNA-binding proteins (RBPs) act as matchmakers. These proteins recognize and bind to short sequence motifs in both flanking introns. If the RBP has a natural tendency to dimerize (pair up with another copy of itself), the two bound proteins can act as a bridge, physically linking the two intron regions and looping the RNA to present the back-splicing sites to the spliceosome. The protein Quaking (QKI) is a famous example of such a "bridging" factor.
While exonic circRNAs made by back-splicing are the most studied, they are not the only members of the circular RNA family. The cell, it seems, has multiple ways to make a loop.
One fascinating class is the circular intronic RNA (ciRNA). These are not products of back-splicing but are essentially "escaped" introns. During normal splicing, an intron is removed as a lariat—a loop with a short tail, closed by a peculiar phosphodiester bond instead of the usual bond that forms the RNA backbone. Normally, a debranching enzyme quickly snips this unique bond to linearize the lariat for degradation. However, some introns contain specific sequence motifs (typically a GU-rich sequence near the end and a C-rich sequence near the branchpoint) that seem to protect this lariat structure from the debranching enzyme. The lariat's tail is trimmed, and what remains is a stable circular intronic RNA, held together by its signature linkage.
Yet another pathway exists for introns found within transfer RNAs (tRNAs). These are excised by a completely different enzymatic system, involving the TSEN endonuclease and the RTCB ligase. This machinery cuts out the intron and then directly pastes its ends together, forming a tRNA intronic circular RNA (tricRNA). Because this pathway is totally independent of the main spliceosome, producing these tricRNAs is unaffected by drugs that shut down pre-mRNA splicing, a key clue that revealed their separate origin.
The stability of circRNAs doesn't just come from resisting exonucleases. They also carry a kind of "invisibility cloak" that allows them to evade one of the cell's most important quality control systems: Nonsense-Mediated Decay (NMD).
NMD is a surveillance pathway that destroys messenger RNAs that appear to have a "premature" stop codon, which could otherwise lead to the production of truncated, non-functional, and potentially harmful proteins. For NMD to be triggered, a number of things have to happen. The process usually requires a special cap structure on the mRNA to initiate a "pioneer round" of translation. As the ribosome moves along the RNA, it looks for stop signals. If it finds a stop codon and "sees" a specific marker called an Exon Junction Complex (EJC) a certain distance downstream, it sounds the alarm, and the mRNA is rapidly destroyed.
CircRNAs neatly sidestep this entire process for two main reasons. First, because they are closed loops, they lack the cap needed for the pioneer round of translation where NMD operates. Second, the very nature of back-splicing means that any EJCs deposited are not in the standard linear configuration that the NMD machinery is trained to recognize. Without the right entry ticket (the cap) and the right spatial landmark (the downstream EJC), the NMD police simply don't see the circRNA as a target. This NMD-resistance is a major reason why exonic circRNAs can have such long half-lives and accumulate in the cytoplasm.
So if exonucleases and NMD can't get them, how do circRNAs ever die? As we hinted earlier, their fate lies in the hands of endonucleases—the enzymes that can cut from within. Several such pathways have been discovered.
One prominent route involves microRNAs (miRNAs). While many circRNAs act as "sponges" by weakly binding to miRNAs, some circRNAs contain a site that is almost perfectly complementary to a specific miRNA. This high-affinity binding can guide the Argonaute 2 (AGO2) protein, the catalytic heart of the miRNA machinery, to act as an endonuclease. It makes a single, precise cut in the circRNA, linearizing it for rapid degradation. The well-studied cleavage of the circRNA CDR1as by miR-671 is a textbook example of this mechanism.
Another pathway is part of our innate immune system. When a cell detects viral double-stranded RNA, it can activate an enzyme called RNase L. This potent endonuclease patrols the cell, chopping up RNAs at specific single-stranded sites to shut down cellular- and viral-protein production. Because circRNAs are also on its hit list, this pathway represents a major route for circRNA decay during an immune response.
Finally, even the chemical modifications on the RNA itself can seal its fate. The common -methyladenosine (mA) modification can be "read" by specific proteins that recruit endonucleolytic machinery, marking the circRNA for destruction. The life of a circle, while long, eventually comes to an end, initiated by a single, fateful internal cut.
A final piece of the puzzle is where these molecules perform their functions. Are they confined to the nucleus where they are born, or do they travel out into the cytoplasm? The answer, it seems, is both, and it depends on their specific makeup.
Many exon-intron circular RNAs (EIciRNAs)—circles that retain one or more of their introns—are found to be predominantly nuclear. A leading hypothesis is that the retained intronic sequences often contain binding sites for the U1 snRNP, a core component of the spliceosome. This binding appears to tether the EIciRNA within the nucleus, where it can even interact with its parent gene's transcription machinery to regulate expression.
In contrast, most pure exonic circRNAs (ecircRNAs) are efficiently exported to the cytoplasm. The mechanisms are still being unraveled, but export appears to be size-dependent and can be facilitated by specific RNA helicases. Furthermore, the same modifications that can trigger decay can also be recognized by a nuclear "reader" protein, YTHDC1, which promotes the export of the marked circRNA into the cytoplasm.
From their elegant circular structure to the diverse mechanisms of their biogenesis and decay, circRNAs represent a rich and complex layer of gene regulation. They are not simply molecular curiosities but are stable, functional entities whose very existence is a testament to the cell's remarkable capacity for molecular innovation.
In the preceding chapter, we unraveled the clever molecular trick—back-splicing—that nature uses to create circular RNAs (circRNAs). We saw that these are not mere oddities but a distinct class of molecules born from the same genetic blueprints as their linear cousins. Now, we embark on a journey to discover what these circles are for and where they fit into the grand tapestry of life. We will see that understanding circRNAs is not an isolated pursuit; it forces us to refine our most powerful tools, rethink fundamental dogmas, and forge connections between genetics, neuroscience, evolution, and medicine.
Before you can study something, you must first be convinced it's real. Imagine the challenge: you are looking for a circle in a world you assumed was made of lines. How do you prove its existence? The first application of circRNA biology was, in a sense, the development of a new kind of "molecular vision."
The key was to find a feature unique to the circle, its "ghost in the machine"—the back-splice junction (BSJ). This is the covalent bond that closes the loop, a link that simply cannot exist in any linear RNA. The first brilliant idea was to use a specific enzyme, Ribonuclease R (RNase R), as a filter. This enzyme acts like a molecular shredder, but it can only start from an open end. It tears linear RNAs to pieces but is stymied by the continuous loop of a circRNA. If you treat an RNA sample with RNase R and a particular molecule survives, you have your first piece of evidence that it might be a circle.
To get a positive confirmation, scientists designed a clever trap using the polymerase chain reaction (PCR). In a standard PCR, primers are "convergent"—they face each other on a linear template. To catch a circle, they designed "divergent" primers that face away from each other on the linear gene map. On a linear template, these primers amplify nothing. But on a circular template, where the gene's end has been stitched to its beginning, these divergent primers suddenly find themselves pointing right at each other across the BSJ, happily amplifying a product. The appearance of this specific product is a smoking gun for circularity.
Nature, it seems, has a sense of humor. The very circularity that makes these molecules fascinating also creates a beautiful puzzle in our experiments. When we convert RNA back into DNA for analysis, the enzyme (reverse transcriptase) can get "stuck on repeat." Once it starts on the circular template, it can just keep going, round and round, like a toy train on a circular track. This "rolling-circle reverse transcription" produces long DNA molecules containing many tandem copies of the circRNA sequence. When we then amplify this DNA, we don't just get one band on a gel; we get a whole ladder of them, with sizes corresponding to one, two, or three times the circRNA's length. What appears to be a messy result is, in fact, a profound confirmation of the template's circular topology.
Once we could reliably see these circles, the next question was obvious: how many are there? This is where circRNA biology collides with the world of bioinformatics and statistics. The go-to tool for counting RNAs is high-throughput sequencing (RNA-seq), which shatters all the RNA in a cell into millions of tiny reads and then counts them.
Immediately, a problem arose. The workhorse method for preparing RNA-seq libraries, poly(A) selection, is designed to capture messenger RNAs (mRNAs) by grabbing their polyadenylated tails. But circRNAs, by their nature, have no tails. Using this method is like trying to catch fish with a magnet; you'll miss the entire population you're after. The solution was to switch tactics: instead of selecting for what we want, we eliminate what we don't want—the massively abundant ribosomal RNA (rRNA). This rRNA depletion strategy leaves behind a rich soup of all other RNA types, circles included, finally allowing for their discovery on a massive, genome-wide scale.
But even with the right data, counting circles is tricky. A popular method for quantifying linear mRNAs, called Reads Per Kilobase per Million (RPKM), "normalizes" for the length of the transcript. The logic is that a longer transcript will, by chance, produce more sequencing reads than a shorter one. This assumption, however, completely breaks down for circRNAs when we count them using their unique back-splice junction reads. Think about it: the number of reads that happen to cross that specific junction depends only on the read length, not the total circumference of the circle. Using RPKM is like assuming a longer necklace will have more clasps; every necklace has just one clasp. Applying this flawed logic would make long circRNAs appear artificially rare, introducing a severe bias.
The solution requires a more principled approach. To compare circRNA levels between samples, especially when one sample has been treated with RNase R and has a completely different composition, we can't rely on normalizing to the library's total content. Instead, we must use an "internal standard." This can be a known quantity of synthetic circular spike-ins added to each sample, or a carefully vetted set of "housekeeping" circRNAs whose levels are known to be stable. By normalizing to these constant references, we can correct for technical variations and make meaningful biological comparisons. This highlights a deep principle of measurement: to measure something accurately, you must compare it to a reliable ruler. And sometimes, you have to build the ruler yourself. The final layer of rigor comes from statistics, where we use decoy datasets and false discovery rates to ensure that the circles we count are real signals and not just computational noise.
So, we can find them and we can count them. Now, what do they do? It turns out that circRNAs are not just bystanders; they are active participants in the life of the cell, often in surprising ways.
Perhaps most stunningly, some circRNAs can be translated into proteins. This flies in the face of the textbook rule that eukaryotic ribosomes initiate translation by recognizing a cap on the mRNA. Since circRNAs are capless, how do they do it? They employ a strategy used by many viruses: they contain an Internal Ribosome Entry Site (IRES). An IRES is a complex, folded RNA structure that acts as a secret landing pad, recruiting the 40S ribosomal subunit directly from the cytoplasm, bypassing the need for a cap entirely. This discovery opens up a whole new proteome, encoded by what was once thought to be non-coding RNA.
More commonly, circRNAs function as powerful regulators of gene expression. Many act as "sponges" or "decoys" for other molecules. Imagine a microRNA (miRNA), a tiny RNA whose job is to find and silence specific mRNAs. A circRNA studded with binding sites for that miRNA can effectively soak it up, preventing it from acting on its linear targets. This single circRNA can thus de-repress a whole network of genes. Similarly, circRNAs can act as scaffolds or decoys for RNA-binding proteins (RBPs). An RBP might have a crucial job, but if it gets "stuck" binding to an abundant circRNA, its activity is sequestered.
Proving such a function requires exquisite experimental rigor. It's not enough to show that a circRNA and an RBP can bind. You must show they bind in vivo and that their binding matters. The modern toolkit for this is breathtaking. Scientists use methods like eCLIP to take a snapshot of the RBP-circRNA interaction in living cells, identifying the precise binding footprint down to the nucleotide. Then, to prove function, they deploy molecular scalpels in the form of steric-blocking antisense oligonucleotides (ASOs). These are designer molecules that bind to the RBP's docking site on the circRNA, preventing the protein from landing. Critically, these ASOs don't destroy the circRNA itself. If blocking the interaction recapitulates a cellular phenotype, and that phenotype can be rescued by introducing a mutant circRNA that the ASO can't bind but the RBP still can, then you have established true causality.
Finally, the life of a circRNA is not static. While they are renowned for their stability, their destruction can be actively regulated. A striking example comes from our own innate immune system. When a cell detects a viral infection, it activates a powerful endoribonuclease called RNase L. This enzyme's job is to chop up RNA—any single-stranded RNA—to halt viral replication and reset the cell. While circRNAs are safe from exonucleases, they are vulnerable to RNase L's internal cuts. Thus, in the heat of an immune response, the stable population of circRNAs is rapidly cleared out, demonstrating that these molecules are integrated into the cell's most dynamic and critical pathways.
The discovery of circRNAs is not just a molecular story; it has profound implications for organismal biology. Nowhere is this more apparent than in the brain. Neurons are practically overflowing with circRNAs, far more so than any other cell type. Why? It seems to be a "perfect storm" of factors. First, many neuronal genes are exceptionally long, with massive introns. These long introns provide ample opportunity for inverted repetitive sequences (like Alu elements) to flank an exon, allowing the introns to fold back and pair up, bringing the splice sites for back-splicing into perfect proximity. Second, neurons express a unique cast of RBPs that further promote this looping. Third, transcription in neurons can be slow, giving the more complex back-splicing reaction more time to occur. And finally, perhaps most importantly, neurons are post-mitotic; they don't divide. In a dividing cell, even a stable molecule's concentration is diluted with each division. In a neuron, a stable circRNA that is produced can persist for an incredibly long time, accumulating to high levels. This high abundance suggests critical roles in neuronal function, from synaptic plasticity to the regulation of local protein synthesis.
This interplay between linear and circular splicing also opens a new window into genetic disease. The choice between linear and circular splicing is a competitive process. If a genetic variant arises that strengthens a cryptic splice site for a linear transcript, the spliceosome may favor that pathway, producing less of the circRNA from that gene. If that circRNA has a vital function, its depletion could lead to disease. This provides a whole new class of mechanisms for how mutations, even those outside of protein-coding regions, can have pathological consequences.
Looking even broader, we can see the hand of evolution at play. Where do the intronic repeats that promote circularization come from? Often, they are the remnants of "jumping genes" or transposable elements. The primate genome, for instance, is littered with millions of copies of the Alu element. Over evolutionary time, these elements have inserted themselves throughout our genes. By sheer chance, some have landed in opposing orientations in the introns flanking an exon. This has created a primate-specific "scaffold" for back-splicing, giving rise to a repertoire of circRNAs found in humans but not, say, in mice. What was once dismissed as "junk DNA" is revealed to be a crucible of evolutionary innovation, generating new molecular players from old parts.
This theme—of a simple circle enabling complex biology—is not unique to animals. In the plant world, there exist viroids: tiny, naked, infectious circles of RNA. Unlike circRNAs, they are true parasites, hijacking the host cell's machinery to replicate themselves, but they are a testament to the power of the circular form. They are minimal, stable, and highly effective. From the simplest plant pathogen to the complexity of the human brain, it seems that life has repeatedly discovered the elegance and utility of thinking in circles.