Our work serves as an important reminder that rigorous validation of off-target effects are essential for screens, particularly ones involving circRNAs.

Overall, our findings show that Cas13 is neither as specific nor as effective for circRNA knockdowns as previously thought. It also demonstrates how rare functional circRNAs are relative to other linear RNAs, at least in the context of a broad phenotype like cellular growth.

In addition to defining this off-target effect, we also used existing sequence-based models that accurately predict Cas13 guide activity to show that most circRNA junctions are poor substrates for Cas13 guide activation.

Notably, this included circFAM120a, which was the main circRNA hit that the original authors had selected for further characterization in their study.

Given the similarity in guide designs, we wondered if certain published screens also suffered from similar issues (doi.org/10.1038/s415...). By re-analyzing their data + conducting our own RT-qPCR experiments, we confirmed that many of their top hits were likely confounded by off-target effects.

This guide position is problematic because it provides the longest homology between the guide and the off-target linear junction (at the 5’ end of the Cas13 spacer). Despite using a 30 nucleotide spacer, only 22-23 nucleotides of homology are needed for Cas13 activation.

This suggested to us that these guides may also be targeting the linear isoforms of these circRNAs. Through a series of single-guide RT-qPCR experiments, we validated that the observed growth effects were caused by unintended knockdown of cognate linear isoforms, and not by circRNA knockdown.

We found that most circRNA-targeting guides did not have an effect on cell growth, especially when compared to the effects of targeting linear mRNAs. Of the circRNA guides that did appear to alter cell growth, we noticed a concerning pattern: nearly all started at the same position along the BSJ.

We were eager to adopt this method ourselves, so we conducted growth screens in the well-studied cell line, K562. Using Cas13d, we targeted 900 of the most highly expressed circRNAs in K562, along with their matched linear isoforms.

Some studies seemed promising at first because they not only described a reliable method for specifically degrading circRNAs at scale, but suggested that an appreciable fraction of circRNAs are biologically functional.
doi.org/10.1038/s415...

This constraint is problematic for tools like RNAi, which have well-documented off-target effects that require greater design flexibility to circumvent.

Recently, CRISPR-Cas13d has been of interest due to reported higher efficiency and specificity in circRNA screens.

Hundreds of thousands of circRNAs have been identified, yet it is unclear how many are functional. A key reason for this uncertainty is that circRNAs are difficult to perturb independent of their linear isoforms. CircRNA-targeting designs are often restricted to a narrow window around the BSJ.

CircRNAs are alternatively-spliced isoforms that derive from genes that typically produce linear mRNAs. They consist of the same sequences as their linear counterparts, but are back-spliced to produce closed RNA loops, which are marked by a unique sequence known as the back-splice junction (BSJ).