Imaging with CRISPR/Cas9

The hottest buzz-word in biology today is CRISPR: an adaptive immune system in bacteria and archea. At its basis is a nuclease, named Cas9, which is targeted to DNA by a short single-guide RNA (sgRNA). This turned out to be a very useful system for genome engineering in any organism due to its specificity (provided by the sgRNA) and its simplicity (all you need is to express the Cas9 and sgRNA in the cell). However, this system can also be used for other purposes. One such use is modulation of gene expression, for example by targeting a nuclease dead Cas9 (dCas9) fused to a transcription activator or repressor to promoter regions. Another such use is for imaging.

Here, I’ll described how Cas9 can be used to visualize specific DNA loci or specific RNA transcripts in fixed and live cells.


Cas9-mediated fluorescent in situ hybridization (CASFISH)  was developed by the HHMI Janelia branch of Rob Singer’s lab (in collaboration with Tjian lab). It was developed to visualize DNA loci in fixed and live cells. Why do we need CASFISH if we already use FISH  that is based on oligo probes? Beacuse FISH for DNA has two main disadvantages: 1) It is a very elaborate protocol (many many steps of washes with all kinds of solutions) that takes a very long time (hours to days(!)) and 2) it requires heat (typically >65°C) and high formamide concentration to achieve total DNA denaturation (and probably sample malformation).  CASFISH is a rather simple protocol that takes <1hr (and they even got it down to 15min!) at 37°C. Also, it can be modified to use for live cell imaging.

Becasue Cas9 binds a specific DNA sequence in a tertiary complex with the sgRNA, one can label the dCas9 and/or the sgRNA with a fluorescent dye, thus visualize that specific locus. However, to get a high enough signal above the background, one needs to have multiple dCas9 proteins at a single locus. This is why most of this paper deals with imaging repetitive sequences – it is easier to image becasue with a single sgRNA you can get multiple binding of dCas9 to these loci.

The protocol for CASFISH requires pre-assembly of the dCas9-sgRNA complex  in vitro and then applying it to the fixed cells. So, the main disadvantage to my mind is that one needs to clean dCas9 protein. But, I’m sure a biochemist can do that easily enough and at large enough quantities.

They started by using labeled dCas9 (fused to Halo tag – and using Luke Lavis’s  fluorescent dyes) together with labeled sgRNA. However, labeling the sgRNA is expensive, so they continued with dCas9-labeling only.


The principle behind CASFISH and examples of CASFISH of pericenteomeres & telomeres, as well as imaging in live cells. Adapted from Deng W. et al. (2015) PNAS 112:11870.


At first they tried imaging sites with 100s to 1000s or repetitive sequences like telomeric regions and the murine major satellite (MajSat) region at pericentromeres. Then they tried regions with a smaller number of repeats, like the human MUC4 gene that contains ~400 repeats in Exon 2 and only 45 (other) repeats in exon 3.  They were able to use two color sequantial labeling to detec both exons and these were co-localized in 94% of the cases. that is amazing!

Two color labeling is fairly easy.  All they needed to do is differntially label two batches of dCas9 and pre-assemble each batch with a different sgRNA. Sequential hybridization is easy and they showed that the tertiary complex (dCas9-sgRNA-DNA) is strong and does not disassemble, even in the presence of 100-fold excess of target DNA. Actually, since there are 4 possible dyes for dCas9-Halo (and more dyes can be added by labeling the sgRNA), multiplexing CASFISH with barcoding should be possible.

An attempt was made to image a non-repetitive sequence. 73 sgRNAs were synthesized against the non-repetitive sequence of MUC4 and dual CASFISH was performed with a second color probe agains a repetitive region. Though it worked, they noticed that this was less efficient and had higher background compared to the probe agains the repetitive region, for reasons yet unknown.

Last was an experiment to image DNA loci in tissue slices. This worked very well, and is therefore a major advantage over regular DNA FISH.

In summary, this seems to be a good method to visualize specific genomic loci. I wonder if it can be combined with RNA FISH. That would be cool, in particular for studies on transcription – since transcriptions sites will be positively identified, including inactive sites.

It will also be useful to study the relationship between DNA copy number (e.g. gene multiplications, plasmids, viral DNA – either integrated or non-integrated during infection) and RNA copy numbers. I’m sure there’s a lot to learn there on the regulation of gene expression.


RNA-targeting Cas9:

The best way to image RNAs in fixed cells is single-molecule FISH. However, imaging RNA in live cells is trickier. I reviewed those methods in some detail in my Nature Review paper. So far the best method to use is the MS2 system. However this requires introduction of multiple aptamers to the RNA (a long repetitive sequence, not without problems. I’ll write about it more in the next post)  and expression of the MS2 coat protein (which binds these aptamers) fused to a fluorescent protein. Cloning is somewhat trickier and there is always the concern of perturbing the RNA behavior.

Recently, it was shown that Cas9, targeted by a sgRNA, can bind RNA in vitro. For optimal binding, this requires a third component- PAMer: a DNA oligomer that contains the NGG  motif and hybridizes to the target RNA adjacent to the sgRNA. Dave Nelles, from Gene Yeo’s lab at UCSD decided to use this system for imaging of mRNAs in live cells.

So, he transfected cells with plasmids encoding dCas9-FP (GFP or mCherry) and a specific (targeting) sgRNA or a non-targeting sgRNA in combination with a stabilized PAMer (targeting or non-targeting). The dCas9 has a nuclear localization signal (NLS), so in the absence of a targeting sgRNA it remains in the nucleus, whereas in the presence of targeting sgRNA it is found in the cytoplasm – presumably bound to the target mRNA (GAPDH, which is highly expressed). Interestingly, the PAMer was not essential for the labeling with Cas9, but did slightly improve it.


The principle of RCas9. Nelles D. et al (2016) Cell: 165:1-9.


Also, even in the absence of sgRNA and PAMer, a small fraction of cells (12%) showed a cytomplasmic signal greater than nuclear signal. Which means that this system is not all-or-none and may still require optimization, as we can see later.

Next, they showed that this system can be used for RNA immunoprecipitation, using an anti-GFP antibody against the dCas9-GFP. The enrichment wasn’t great (~4-fold at best) but I guess it can be improved by optimizing the protocol. For instance, maybe the sonication step destabilizes the complex? Dunno, but worth to check. Other tests verified that the target mRNA and its protein product levels are not affected.

So, can we actually visualize and analyze mRNA localization using this system? Yes, but not at single molecule level. First, they compared RCas9 to RNA FISH and found a nice correlation of the two in the presence of the targeting sgRNA.

(I should mention that the FISH images are not as good as I expected, but this is probably due to the method of imaging – using a confocal microscope and maybe other parameters).



upper panel: ACTB FISH and RCas9 co-localization from Nelles et al. (2016) Cell 165:1-9. I think the colors just disturb the visualization. I can’t really see single molecules, or distinguish transcripiotn sites. Lower panel – ACTB FISHimage from my bench. To my eyes, the single molecules and transcription sites are much more visible in my image.


Then, to show that one can use this system to follow changes in localization in live cells, Dave treated the cells with sodium arsenate. This treatment is a known trigger the formation of stress granules (SG) – RNA granules which are distinct from P-bodies and typically contain mRNA with translation factors (including, I think, the small ribosomal subunit). Using a known SG marker fused to RFP,  it could be shown that RCas9-GFP accumulates in SG only when targeting sgRNA is used, in a time and arsenite dose dependent manner. Hence, RCas9 seems to work well in tracking the labeled mRNAs (encoding beta-actin, cyclin A2 and transferrin receptor) moving into SG.

SG RCas9

RCas9 signal co-localized with the stress-granule marker G3BP1. NTC-1: non-targeting sgRNA control. Adapted from Nelles et al. (2016) Cell 165:1-9.


As I mentioned, this work is still not at the single molecule level, since there is only one Cas9-FP per mRNA molecule, which is difficult to detect above background fluorescence.  One way to achieve single-molecule resolution is by increasing the number of sgRNAs & PAMers  which target a specific mRNA. From experience with the MS2 system, probably a minimum of 6-8 RCas9 will be required, if we use a good fluorescent mark (e.g. Halotag). However, one must be carful not to mask cis-elements which can affect the mRNA behavior (obviously, one is also limited to the UTRs since targeting the ORF will probably affect translation. The UTRs are rich in regulatory cis-elements…).

A second option is to increase the fluorescent signal of a single Cas9 protein by fusing it to multiple GFPs (or Halotags…). However, how can we distinguish an RNA from a free Cas9?

The major advantage of this system is that one can label endogenous RNAs in a very simple manner. Actually, you only need to design the sgRNA and the PAMer. No need to label the mRNA with multiple aptamers or other means.

Furthermore, just like the MS2 system, it can be used to manipulate the RNA. For example, tether the RNA to another location or tether a protein to the mRNA at a specific position. One can use it to mask a cis-element (e.g. a splice site, a zipcode, an ARE sequence etc…)and see how it affects the mRNA behavior.

Can we detect multiple mRNAs? Perhaps a combination of MS2 with RCas9, to detect two mRNAs bundled together in the same RNA granule, or localize to the same cellular location.

If we use split GFP, one can use two adjacent sgRNAs to target the same mRNA and thus detect only cells that express that specific transcript (e.g. splice variant).

I’m going to try RCas9 soon. I’ll begin with Dave’s plasmids, then see how it goes from there. I’ll report here how it works. Deng W, Shi X, Tjian R, Lionnet T, & Singer RH (2015). CASFISH: CRISPR/Cas9-mediated in situ labeling of genomic loci in fixed cells. Proceedings of the National Academy of Sciences of the United States of America, 112 (38), 11870-5 PMID: 26324940

Nelles DA, Fang MY, O’Connell MR, Xu JL, Markmiller SJ, Doudna JA, & Yeo GW (2016). Programmable RNA Tracking in Live Cells with CRISPR/Cas9. Cell, 165, 1-9 PMID: 26997482


2 responses to “Imaging with CRISPR/Cas9

  1. Pingback: New promising method for in situ labeling of genomic loci in fixed cells | HCS-pharma

  2. Pingback: New CRIPSR-based RNA imaging tool | greenfluorescentblog

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s