When you choose to publish with PLOS, your research makes an impact. Make your work accessible to all, without restrictions, and accelerate scientific discovery with options like preprints and published peer review that make your work more Open.

PLOS BLOGS The Official PLOS Blog

Sensitive CRISPR diagnostics using RNA targeting CRISPR enzyme

I was recently involved in a collaboration between the Zhang and Collins labs at MIT to use the RNA-targeting CRISPR protein Cas13a/C2c2 to detect either DNA or RNA from pathogens. By combining the use of Cas13a/C2c2 as a detector with isothermal amplification of the DNA or RNA targets, we were able to get down to attomolar detection. You can read the full paper over at Science but here I can give some of my own experience with and views on the platform we’re calling SHERLOCK (Specific High Sensitivity Enzymatic Reporter UnLOCKing).

To be clear, I’m the 5th author on this paper and definitely agree that the four people ahead of me did more of the work. I am not the expert on Cas13a/C2c2. However, I did help some to develop SHERLOCK as a diagnostic system and use it enough to get a handle for how it works with different targets. I also tried it with another diagnostic project and have found it easy to work with.

How it works

Cas13a/C2c2 has two RNA cutting abilities. The first is that it cuts the RNA that you target with theCRISPR guide RNA (crRNA) much like Cas9 cuts DNA that you directly target. The other RNA cutting is more broad and is activated after finding Cas13a/C2c2 finds its target RNA. This broad RNA cutting activity acts like an RNase and will cut many RNAs present in the reaction (or in the cell). In a 2016 Nature paper, the Doudna lab showed that the RNase activity of Cas13a/C2c2 could be used to detect picomolar levels of RNA. They found that Cas13a/C2c2 was able to do at least 10turnovers per target RNA recognized. That catalytic activity of Cas13a/C2c2 gives a strong output signal for even s small input of target RNA.

So Cas13a/C2c2 can be turned into a diagnostic using just a crRNA and a fluorescent RNA probe. After the Cas13a/C2c2 finds its target it starts cutting other RNAs, including the probe, and separates a fluorophore from its quencher. It’s that separation of fluorophore and quencher that gives the fluorescent signal.

To boost the natural sensitivity of Cas13a/C2c2 as a diagnostic, we paired it with isothermal amplification of DNA or RNA. Isothermal amplification methods amplify nucleic acids similar to polymerase chain reaction (PCR) but instead use enzyme mixes that can do the job at a single temperature. SHERLOCK makes use of recombinase polymerase amplification (RPA) that can work between 37-42˚C. This allows both amplification reactions and Cas13a/C2c2 reactions at 37˚C and means that there is no need for expensive machinery to precisely cycle temperatures like PCR.


The level of sensitivity we got certainly jumps off of the page. As mentioned in the last section, Cas13a/C2c2 is itself quite sensitive to RNA molecules and its collateral RNase activity can cut many probe RNAs. It binds to its target RNA and then quickly generates signal through its general RNA cutting activity. As a detector it’s ~1000 times more sensitive than another detector, the RNA toehold switch, that we’ve used in the Collins lab to things like detect Zika.

Similar to the paper-based Zika detection, SHERLOCK was able to be freeze-dried for room-temperature storage, used with minimal hardware, and rapidly reprogrammed rapidly to target almost any sequence. But in addition to the sensitivity advantage, SHERLOCK was able to detect single base mutations. As many important mutations in humans or the pathogens that infect humans are only single bases, the ability to distinguish those small changes would be a major achievement for a cheap diagnostic. Some screening has to be done to find crRNAs that work best for a given mutation, but in general a few variants should be enough.

Potential limitations

This is still early days for Cas13a/C2c2 based CRISPR diagnostics, so there will be more challenges to be addressed in academic labs and in a company setting. While we showed freeze-drying on glass fiber paper and adding RNase inhibitor worked without producing much background signal, samples that contain many RNases could create false positives. A negative control that lacks Cas13a/C2c2or crRNA could inform you of the problem but the RNase containing sample likely couldn’t give you an accurate read of how much nucleic acid is actually present. At the lab bench, we didn’t have problems with background signal but working somewhere like a remote community health center would probably bring less controlled conditions. Rigorous tests will need to be done to make sure that the freeze-dried tests can last out in different conditions.

Other improvements could include a good way to change from a fluorescent output to a color change as the output. This would allow easy readout by eye like a pregnancy test and reduce equipment costs. A color readout can be done by anyone without risk of equipment malfunction. Reducing the equipment and technical skill needed for a diagnostic is key to how easily it can actually be deployed in areas that need it.

Future for CRISPR diagnostics

The variety of CRISPR proteins that target DNA or RNA and can be easily programmed to cut or bind to nearly any sequence. Cas13a/C2c2 is nice because it comes with a secondary activity (general RNA cutting) that can readily be turned into a fluorescent readout. However, CRISPR-Cas9 can also be used for diagnostics when cleverly combined with a way of detecting its targeted DNA cutting. Overall, CRISPR proteins are poised to get integrated into nucleic acid diagnostics as they provide programmable detection and more specificity than traditional nucleic acid amplification-based techniques.


See more coverage over at Science MagazineMIT NewsWashington Post, The Scientist, and STAT News.



Conflicts of interest:

Since I am an author on the paper being discussed, I am mostly certainly biased in favor of its significance. While co-authors may have patent and potential financial stakes in work that follows, I do not personally have a stake in any of these ventures. However, I do work for MIT and have Broad Institute affiliations so I do have institutional conflicts of interest.

  1. Hi Aaron,

    Nice blog post and congratulations on the paper – it’s always nice to see people finding new uses for RPA. I don’t know if you are aware but there are already publications showing how RPA can detect Zika virus with different probe systems at similar levels of sensitivity, but in much less time.
    For example Chan et al showed how you could combine RPA with a mobile phone to detect Zika in <15 minutes:

    Song et al combined RPA with another isothermal amplification method, LAMP to detect Zika and other pathogens in a multiplexed, nested, colourimetric test that showed single plaque forming unit levels of sensitivity:

    Abd El Wahed et al have already taken a real-time RPA Zika assay out into the field with a lab-in-a-suitcase to great success:
    (they've also used the same set-up for ebola: http://dx.doi.org/10.2807/1560-7917.ES.2015.20.44.30053)

    Jauset-Rubio et al have demonstrated attomolar levels of sensitivity with a paper-based lateral flow RPA test:

    Rebecca Richards-Kortum's group at Rice University showed a paper and plastic device for using RPA to detect infant HIV five years ago (http://dx.doi.org/10.1039/c2lc40423k) and showed that RPA reactions could detect 10 copies of HIV DNA using body heat and lateral flow strips (http://dx.doi.org/10.1371/journal.pone.0112146).

    More recently Luke Lee's group at UC Berkeley has developed an inexpensive lab-on-a-chip for digital RPA (http://dx.doi.org/10.1126/sciadv.1501645) and the Weibel lab from University of Wisconsin-Madison have published a low-cost, portable, rapid, and easy-to-use microfluidic cartridge-based system for detecting the ESKAPE (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp.) bacterial pathogens that are most commonly associated with antibiotic resistance: http://dx.doi.org/10.1128/AEM.02449-16

    The Trau group has already shown naked-eye colorimetric detection of single tuberculosis cells using RPA: http://dx.doi.org/10.1021/acssensors.5b00171

    Yamanaka et al have shown low-cost colorimetric SNP detection using RPA: http://dx.doi.org/10.1007/s00604-017-2144-0

    Whilst there is a lot to be said for single nucleotide specificity in assays, there are also lots of advantages to having slightly less specific detection methods, especially for highly mutable RNA viruses such as Zika and HIV. Once you've launched a test it would be a regulatory nightmare to have to go and develop a new assay component specifically for each emerging strain that has a point mutation where your probe is, incorporate it into your existing test, conduct field trials and get approval. One of the advantages RPA has over other isothermal methods is that it is relatively robust to mismatches in primer and probe sequences, see for example this publication on a low-complexity HIV test by Lillis et al http://www.sciencedirect.com/science/article/pii/S0166093416300246

    So, to play Devil's Advocate, what does CRISPR actually add to RPA as a low cost diagnostic? If you want attomolar sensitivity, just do a nested RPA reaction with colorimetric detection or a lateral flow strip. It will be quicker and you won't need to license CRISPR from the Broad!

    There are now over 150 publications showing how sensitive, specific and flexible RPA is. If any of your readers are interested they can find links to them all here:


    Best wishes,


    PS as this is a synthetic biology blog, there's even a publication using RPA for synthetic biology:

    PPS Conflicts of interest: I work for TwistDx, the developer and manufacturer of RPA, so obviously I'm interested in more people finding out about it and trying it out!

  2. Hi Aaron,

    My pleasure. When I talk about lateral flow and RPA I should probably clarify that I’m talking about nucleic acid detection, not traditional lateral flow strips. There are lots of issues with sensitivity and specificity of traditional LF strips, but when you’re detecting amplified nucleic acid rather than protein, at least with RPA and our nfo probes, you can develop assays with single molecule sensitivity.

    Best wishes,


Leave a Reply

Your email address will not be published. Required fields are marked *

Add your ORCID here. (e.g. 0000-0002-7299-680X)

Back to top