About a year and a half ago I wrote here about new uses of CRISPR/Cas9 as an imaging tool. In particular, I was excited about the possibility to use enzyme-dead Cas9 (dCas9) as an RNA binding protein for live imaging of mRNA. Unfortunately, in my hands this did not work (the dCas9 has exited the nucleus with non-targeting guide RNA at the same rate as with the specific guide RNA).
Last week, a new CRISPR tool was published in Nature, from Feng Zhang’s lab. Briefly, Zhang’s team found in the past a new type of Cas protein which was named C2c2 and later Cas13. Unlike Cas9, which typically binds and cuts DNA, Cas13 specifically binds and cuts RNA. In this paper, they isolated a Cas13a variant from Leptotrichia Wadei (LwaCas13a), and used it as a programmable tool to knock-down expression. As a tool for knowckdown it is pretty awesome – it is comparable to shRNA in terms of knock-down (~40-90%), but it has three big advantages:
1. Unlike shRNA, LwaCas13a can also target nuclear RNA.
2. It can be used easily for multiplexed knock-down of sereval RNAs.
3. Amazingly, it seems that it has NO OFF-TARGET effect. They used RNA-seq to look at the change in expression level of all RNAs in the cells. Whereas shRNA showed hundreds of off-target knocked-down RNAs, LwaCas13a was very specific and only the targeted mRNA was knocked-down.
They then mutated LwaCas13a to create a dead variant (dLwaCas13a) and asked if it can be used for pull-down assays and for live imaging.
Just like the RNA-dCas9, the pull down efficiency of beta-actin mRNA wasn’t great (2-3 fold over non-targeting guide RNA). However, for Luciferase mRNA, this ratio was better (8-11 fold enrichment). So, this has good potential. Too bad they did not compare it to puul down using MS2-labeling and see if the same proteins are pulled down with the mRNA.
And now I come to my problem with this paper – the imaging.
Just like the RNA-Cas9 paper, they imaged beta-actin mRNA with a single guide RNA and single GFP fused to dLwaCas13a. dLwaCas13a was nuclear localized, but exited the nucleus when specific guide RNA was used. However, they claim only a 3-3.7 fold enrichment of cytoplasmic/nuclear ratio compared to non-targeting guide RNA. They acknowledge that there is neclaer escape due to nuclear off-target binding or other reasons. They added a negative feedback system to improve SNR. Still, the best they got was 3.7-fold. So, if you look at a specific cell – how can you tell if you are looking on RNA-bound or not? you can’t.
They also compared their labeling to FISH staining. Their FISH looks much better than in the RNA-Cas9 paper, really single spots. But to say that it co-localizes? There is more-or less uniform cytoplasmic green staining. Of course it will “co-localize” with the cytoplasmic beta-actin mRNA FISH spots. At the very least they should have done a control with FISH against another highly expressed mRNA (e.g. GAPDH), or perform beta-actin FISH of cells expressing a guide RNA against a differnt mRNA, and not just compare to the non-targeting control.
Last point – stress granules again? I guess it’s just an easy assay to do, that also gets the RNA in a concentrated blob: good if you don’t have single molecule resolution. But nobody really knows what SGs are and only now starting to figure out how and why they form.
They claim that when they induce stress granules (SG) formation, the GFP-dLwaCas13a co-localizes with the SG marker. However, their images are not very good. You see something in Fig 4 (only 2 cells in the control!) but Extended Fig 10 is terrible. Low resolution, small images – can anyone tell from this figure if there is or isnt co-localization, and get some qualitative, if not quantitative sense?
They do provide quantitative data, but although it could be statistically significant, it does not look striking. Really marginal differnece compared to the non-targeting guide RNA (or non-stressed cells).
Is this a good tool for RNA imaging? I don’t think so. Not until you can get single-molecule resolution.