RNA comes in many shapes and sizes. Over the past few decades, researchers have characterized at least two dozen different RNA varieties beyond the textbook classics. But a type of RNA that long flew under the radar due to its designation as a molecular mishap is now taking center stage.

Circular RNAs (circRNAs), or simply “circles” to many researchers, are just what they sound like: nucleotides of RNA arranged in a closed loop. Much about the function of these molecules remains a mystery, but for some time, at least one thing seemed clear: unlike linear messenger RNA (mRNA), circles were not translated into proteins in living organisms. “When you have any type of RNA, you wonder whether it’s translated,” says Sebastian Kadener, a molecular biologist who has spent the last few years researching circRNA at the Hebrew University of Jerusalem. Despite reporting the presence of one...

A little over a year ago, however, Kadener and his colleagues detected something that would upend that assumption: an average-size (37 kilodaltons) protein encoded by a naturally occurring circRNA in Drosophila. Along with collaborators in Germany, Kadener’s group used a method known as ribosomal footprinting to detect RNAs being actively translated in extracts from fly heads. Not only did the researchers discover more than 100 different circRNAs—ranging from around 300 to more than 2,000 nucleotides in length—apparently associating with ribosomes in the cells, they also identified a protein that, based on its sequence, could only have been translated from one of these circles, not from a standard linear transcript. “We could see the protein by Western blot,” Kadener says. “It was being expressed in the synapses of flies.”

Kadener’s work was published earlier this year,1 back-to-back in Molecular Cell with another group’s study—on human and mouse cells—that had simultaneously come to the same conclusion: translation of circRNAs can and does occur in living cells.2 For now, neither group has any hint of the function of these proteins, or of how common circRNA translation really is, but “you can imagine that it has some biological importance,” Kadener notes. RNA researcher William Jeck, currently a fellow at Harvard Medical School, agrees. Many scientists had “written off translation,” he says. “This is extremely exciting evidence that other circles may produce peptides that may be biologically relevant. . . . It’s really changed the paradigm.”

First observed in electron micro­graphs of eukaryotic cells taken in the 1970s, circRNAs were for decades considered posttranscriptional mistakes.

When it comes to circRNAs, though, such paradigm shifts are par for the course. First observed in electron micrographs of eukaryotic cells taken in the 1970s, circRNAs were for decades considered posttranscriptional mistakes expressed at low levels in the cell—perhaps the results of splicing gone wrong, generated when an exon’s two ends are covalently joined together instead of to adjacent exons. But all that changed when Julia Salzman and colleagues at Stanford University set out to identify all types of RNA in human cells using an unbiased approach—one that diverged from standard methods by including RNAs that lacked polyA tails.3 In 2012, the team discovered thousands of circles using this method. What’s more, “we reported that there were hundreds of circular RNAs that were more abundant than their cognate linear transcripts,” Salzman tells The Scientist. “I think people were in a bit of disbelief.”

See “RNA Chases Its Tail

Around the same time, other labs were finding additional evidence to contradict the view of circular RNA as merely cellular “noise.” Within the year, Jeck, then at the University of North Carolina School of Medicine, and colleagues reported that at least one out of every eight genes expressed in human fibroblasts gave rise to circRNAs.4 “We were frankly shocked finding even one circular RNA,” Jeck recalls. “We thought it was a fascinating novelty.” The group also found that many circRNA sequences were highly conserved between humans and mice. And shortly afterward, two more groups published further evidence of circRNAs’ abundance in humans and mice, and additionally in nematode worms.

Now, research on circRNAs is exploding, and the molecules’ biogenesis is gradually becoming clearer. At least two proteins, Muscleblind and Quaking, have been linked to circle formation, which generally occurs when the cell’s splicing machinery connects a downstream splice donor to an upstream splice acceptor, such as joining an exon’s 5´ end to its own 3´ end or an upstream exon’s 3´ end, in a process known as backsplicing. Recently, several additional mechanisms have been proposed (see “Making the Rounds” here), and some circRNAs contain introns, either instead of or in addition to exons. Regardless of their genetic makeup, the lack of ends makes circles less vulnerable to exonuclease enzymes, allowing them to persist in cells for days, unlike their linear counterparts, whose life spans are measured in hours or minutes.

Despite a growing appreciation for the abundance—and now translation—of circRNAs in eukaryotes, there’s still very little understanding of what exactly circRNAs do. “We don’t even know how much of it is functional,” says Jeremy Wilusz, an RNA researcher at the University of Pennsylvania Perelman School of Medicine. “What’s the point of these circles? Why are they made?”

Salzman agrees that the topic is still wide open. “You can speculate to your heart’s desire. There is currently no consensus about what they do.”

In search of a function

Whether or not circRNAs are translated, it’s possible that the vast majority of circles do nothing at all. “It’s crazy to assume they’re doing something just because they’re there,” says Nikolaus Rajewsky, an RNA researcher at the Max Delbrück Center for Molecular Medicine in Berlin who collaborated on both Molecular Cell papers. “The null hypothesis is that they’re not doing anything.” He adds that although thousands of circRNAs are expressed in various tissues, few are expressed at levels that are likely to be particularly biologically relevant. “It’s not like there are thousands or millions of circles everywhere and they’re all important,” he says.

But there are certainly reasons to believe that at least some circRNAs are more than just molecular accidents. In addition to the fact that many circRNAs are conserved across species, research suggests that circularization is regulated. If circles were merely by-products of normal splicing, their levels might correlate with levels of linear transcripts expressed from the same gene, Salzman says. But in 2013, her group found that different cells showed different ratios of circular to linear transcripts from the same gene—although how the relative stability of each RNA molecule contributes to the overall balance remains to be determined.5

A couple of years later, Rajewsky’s team published hints that circRNAs play a role in the nervous system, showing that many circles in humans and mice are highly expressed in neural tissue, upregulated during neuronal differentiation, and enriched at synapses. “We looked at exactly what circles are expressed,” he says. “Our data indicate that we’re talking about a few hundred really interesting candidates in the brain.” Many of these candidates are tissue-specific—with some circles enriched in  the cerebellum, for example, and others in the cortex—and are expressed only at certain stages of neuronal development.6

Collectively, the studies hint at the functionality of circRNAs, but the exact nature of their roles has largely eluded researchers—though there have been a few tantalizing clues. In 2013, researchers discovered that some circRNAs act as molecular “sponges,” soaking up large quantities of specific microRNAs—tiny, noncoding molecules about 20–25 nucleotides in length. That year, two studies—one by Thomas Hansen of Aarhus University in Denmark and colleagues and one by Rajewsky’s group—simultaneously reported that a circRNA  transcribed from the antisense strand of the human CDR1 gene and highly expressed in the brain, called CDR1as by Hansen and ciRS-7 by Rajewsky, has dozens of binding sites for a microRNA known as miR-7.7,8 Hansen’s group also showed that another circRNA, transcribed from the sex-determining region Y (Sry) gene and expressed in mouse testes, could bind microRNA miR-138.  

See “Circular RNA Surprise

Because microRNAs are involved in regulating translation—by binding to specific mRNAs, they trigger degradation of transcripts through a process known as RNA interference (RNAi)—Hansen and his colleagues speculated that the findings might indicate a general role for circles in regulating gene expression. “At that time, we were searching for other circular RNAs, but pipelines for detection weren’t really established,” Hansen tells The Scientist. When they found similar roles for two circRNAs, “we didn’t know, but of course we hoped that it could be a general thing, that circular RNAs would emerge as [regulators] of these micro-RNAs—it made a lot of sense.”

But researchers now believe that most circles are unlikely to act as microRNA sponges. As the number of known circRNAs has climbed into the thousands, only a handful of sponges have been identified. And a 2014 study using computational methods to predict sequences likely to make good sponges identified only a few other candidates.9 The authors of that study “made a strong case that it wasn’t a general function,” says Salzman. Instead of sponging, circRNAs may be engaging in other types of microRNA interactions, Hansen notes. “I think [circles] could have more profound effects in terms of stabilizing [microRNA], or directing it to certain parts of the cell—although that’s of course hypothetical at the moment.”

CircRNAs also appear to associate with proteins, suggesting another suite of potential regulatory functions. For example, researchers recently showed that a circRNA produced by the Foxo3 gene (called circ-Foxo3) interacts with proteins involved in cell proliferation, including a key cyclin-dependent kinase and one of its inhibitors, suggesting a role in the cell cycle. And while most exon-containing circles accumulate in the cytoplasm, those that retain introns are often found in the nucleus, where they encounter proteins involved in transcription. In 2015, scientists in China showed that a group of exon-intron circRNAs promoted transcription of their parent genes via interaction with RNA polymerase II.10 Other studies have shown circles interacting with different RNA-binding proteins as well, including proteins now linked to circRNA biogenesis, such as Muscleblind and Quaking, and Argonaute proteins, well-known for their participation in RNAi-based gene regulation.

There’s the possibility that the regulation of circRNA biogenesis itself constitutes a function, too. Because each RNA transcript can be either linear or circular, but not both, upregulating circularization could act as a mechanism to reduce the proportion of linear mRNA generated from a particular gene. A recent study by Rajewsky and Kadener showed that strong competition between circularization and linear splicing can occur, most likely due to overlapping dependence on the same splicing machinery—although the extent to which it constitutes a function per se is still unclear.11

With the recent description of in vivo translation of circRNAs comes an entirely new dimension of possible functions—one that researchers are only beginning to explore. “I’m sure that people are now going to be looking to see when these proteins are produced, where these proteins are produced, et cetera,” says Kadener, adding that his team plans to further investigate the role of translated circRNAs in Drosophila brain function. “You can imagine so many hypotheses of what this translation might mean. . . . The protein made by the circle could modulate other proteins, for example. It opens a lot of possibilities.”

Like all of the speculation about circRNA function, though, hypotheses about translation will have to be pursued with a healthy dose of skepticism, notes Wilusz. “It’s certainly a very attractive idea,” he says. “It would make sense in some way, that if you’re making [a circRNA] from a protein-coding gene, you should make a protein. But there’s a lot more work that needs to be done to prove that the proteins are being produced at high levels—or even do anything.”

Putting circles to use

MAKING THE ROUNDS: Circular RNA biogenesis occurs when RNA fragments are bent into a closed loop of one or more exons and/or introns. This often occurs as the pre-mRNA molecule is processed into its final transcript via splicing, in which introns are removed and exons are linked together.
See full infographic: WEB | PDF
As RNA researchers continue to explore circles’ possible functions, multiple labs have discovered that circRNA expression levels vary substantially with disease, leading to growing interest in how these molecules might be harnessed for diagnosis and treatment. Certain circRNAs are up- or downregulated in cancers of the skin, liver, bladder, larynx, and stomach, to name a few. And it’s not just cancer; abnormal expression of several circRNAs has also been linked to cardiovascular disease and to neurological disorders such as Alzheimer’s and Parkinson’s.

CDR1as, for example—one of the original microRNA sponges and the best-studied circle to date—is linked to a number of diseases, in several cases via its sequestration of miR-7. A well-characterized tumor suppressor, miR-7 inhibits cell growth, and its loss is associated with poor prognosis. “High expression of CDR1as is not very good in terms of cancer,” Hansen explains, “because it inhibits the microRNA that would normally protect from cell proliferation.” Looking beyond cancer, researchers in China reported in 2015 that overexpression of miR-7 in pancreatic islet cells led to impaired insulin production and diabetes in mice—an outcome the team suggested was normally kept in check by the sponging activity of CDR1as.12 And reduced expression of CDR1as in the hippocampus has been associated with Alzheimer’s disease.

Another disease-linked circular RNA, circTCF25, also appears to act as a microRNA sponge. Expressed at high levels, circTCF25 downregulates two microRNAs, leading to cancer cell proliferation in vitro and in vivo in humans—mechanisms that could explain the link between high circTCF25 levels and bladder cancer. And earlier this year, researchers described a complex pathway in which peptide-binding circ-Foxo3—downregulated in several cancers—regulates proteins involved in cancer cell death. The team showed that through interactions with several peptides, circ-Foxo3 increases levels of its parent gene’s protein, Foxo3, which can trigger apoptosis in tumor cells.

These glimpses into circRNA’s role in disease have sparked interest in exploiting the molecules as potential therapeutic targets. A study published earlier this year noted that silencing CDR1as using specially-designed short hairpin RNAs (shRNAs) inhibited proliferation and invasiveness of colorectal cancer cells in culture. And a team at Mount Sinai School of Medicine in New York used similar methods to target ciRS-E2, a circle consisting of a single exon that is highly expressed in cancers such as leukemia and melanoma. The group reported that shRNA treatment dampened ciRS-E2 expression by more than 80 percent in cultured cancer cells, and resulted in significantly reduced proliferation.

For now, though, while functions for the vast majority of circRNAs remain unclear, many labs are focused on exploring the more immediate goal of using circles to classify and monitor diseases with which they are associated. For example, “we’re all very interested in trying to find ways to divvy up tumors into different categories of risk and potential response to therapy,” says Jeck. “CircRNAs do have one really nice feature, and that’s that they are stable. That means if you can get a good sample of cells, you have a really good shot at identifying [them].”

Indeed, researchers recently found that circRNAs are present in circulating extracellular vesicles such as exosomes, and could in some cases provide more information about gene expression in healthy and unhealthy cells than their linear counterparts in easily accessible human fluids. In a 2015 study of blood-borne circRNAs, Rajewsky’s lab discovered that detecting the circular transcripts served as a more faithful proxy for the expression of hundreds of genes than classical mRNA-specific assays.13 “We would not ‘see’ these genes, so to speak, by normal RNA expression,” he says. “So circRNAs could be molecules that tell you something about development or disease that normal molecules do not.”

Specific, circRNA-based biomarkers for several diseases have already emerged from retrospective analyses of patients. In January, researchers described a combination of two circular RNAs, hsa_circ_0124644 and hsa_circ_0098964, that detected coronary artery disease with a specificity and sensitivity rivaling current methods, while presenting a cheaper and more convenient alternative. And other studies in the last two years have highlighted specific circular biomarkers for several cancers, including liver, stomach, and colorectal. Now, these candidates must be validated in studies that predict disease outcome, says Jeck. “There have been a lot of retrospective analyses, and that’s all well and good,” he says. “But I think the next step is to see if people can use circRNA expression in a prospective manner. That would be very exciting and potentially very useful.”

Of course, how circRNAs come to be understood in the lab and possibly one day used in the clinic remains to be seen, as the study of these looped molecules represents an area that’s still young. But if the past five years are any indication, the study of circRNAs is rapidly ramping up. “What’s amazing to me is how fast this field has grown,” says Wilusz, whose lab supplies plasmids expressing circRNAs to other research groups and has recorded a dramatic uptick in requests in the last couple of years. “It’s really taking off.”

Rajewsky, whose group is now focusing on circRNAs’ interactions in the brain, agrees that the best is very much ahead. “We’re really just at the beginning of an exciting journey,” he says. “It doesn’t happen often in molecular biology that you find such a fundamentally new phenomenon.” 


For years, circular RNAs were overlooked, not least because traditional sequencing methods were not designed to identify them. One of the most commonly used transcriptome-sequencing approaches, RNA-Seq, often includes a selection step that picks out only RNA molecules such as linear mRNA that have polyA tails, a posttranscriptional addition that circRNAs lack. To retain circles in a sample, researchers have to skip this step, or deliberately select for RNAs lacking polyA tails instead.

Even when circles are retained, however, their identification is far from trivial. In 2012, Stanford University’s Julia Salzman and colleagues described an approach to pick out exon sequences that had been scrambled in RNA relative to their sequences in the genome, as a mismatched order of exons could indicate a circular arrangement (PLOS ONE, doi:10.1371/journal.pone.0030733). But algorithms that identify these mismatches show little overlap in their predictions. One recent study comparing five current algorithms reported that up to 40 percent of predicted circRNAs were only flagged by one algorithm, and fewer than 20 percent of all the circles predicted in the study were identified by all five (Nucleic Acids Res, 44:e58, 2016).

As a further complication, scrambled exon sequences can result from things other than circularization of RNA, including certain unusual forms of splicing or, more commonly, artifacts from the sequencing technique itself, due to the use of enzymes such as reverse transcriptase. To combat this problem, some groups are working to develop statistical methods that estimate false detection rates, and distinguish real circles from by-products of the approach.

In the meantime, several researchers have pointed out that the challenges of finding circRNAs raise a deeper question about RNA research: If these abundant molecules were all but invisible to earlier RNA detection methods, what other structures could be out there that are currently being overlooked?


  1. N.R. Pamudurti et al. “Translation of circRNAs,” Mol Cell, 66:9-21, 2017.
  2. I. Legnini et al., “Circ-ZNF609 is a circular RNA that can be translated and functions in myogenesis,” Mol Cell, 66:22-37, 2017.
  3. J. Salzman et al., “Circular RNAs are the predominant transcript isoform from hundreds of human genes in diverse cell types,” PLOS ONE, 7:e30733, 2012.
  4. W.R. Jeck et al., “CircRNAs are abundant, conserved, and associated with ALU repeats,” RNA, 19:141-57, 2013.
  5. J. Salzman et al., “Cell-type specific features of circular RNA expression,” PLOS Genetics, 9:e1003777, 2013.
  6. A. Rybak-Wolf et al., “Circular RNAs in the mammalian brain are highly abundant, conserved, and dynamically expressed,” Mol Cell, 58:870-85, 2015.
  7. T.B. Hansen et al., “Natural circRNAs function as efficient microRNA sponges,” Nature, 495:384-88, 2013.
  8. S. Memczak et al., “Circular RNAs are a large class of animal RNAs with regulatory potency,” Nature, 495:333-338, 2013.
  9. J.U. Guo et al., “Expanded identification and characterization of mammalian circular RNAs,” Genome Biol, 15:409, 2014.
  10. Z. Li et al., “Exon-intron circular RNAs regulate transcription in the nucleus,” Nat Struct Mol Biol, 22:256-64, 2015.
  11. R. Ashwal-Fluss et al., “circRNA biogenesis competes with pre-mRNA splicing,” Mol Cell, 56:55-66, 2014.
  12. H. Xu et al., “The circular RNA CDR1as, via miR-7 and its targets, regulates insulin transcription and secretion in islet cells,” Sci Rep, 5:12453, 2015.
  13. S. Memczak et al., “Identification and characterization of circular RNAs as a new class of putative biomarkers in human blood,” PLOS ONE, 10:e0141214, 2015.

Interested in reading more?

Magaizne Cover

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!