Scientist Stories: Omar Abudayyeh & Jonathan Gootenberg, CRISPR Gene Editing and Founder-Inventors

Yeah, so I'm just gonna hop in. So, you know, our group at MIT basically, you know, focuses on making better molecular tools for gene editing and cellular engineering. And as many many of you are probably familiar with, we mine natural diversity for new systems, enzymes that we can use to leverage for these applications. And when we first started down this path now almost, I guess, 6, 7 years ago, of course, Cas9 was really the only slice that folks were focusing on for programmable sort of double-stranded DNA cleavage and editing. And we asked the question of whether we could go beyond Cas9, as other folks in this session have also been exploring. Now a lot of the work we do relies on, you know, finding signatures of systems that we would want to find interesting. So, you know, for CRISPR systems, there are a few things that stand out. One is the CRISPR array. The other are proteins that are highly conserved, like adaptation proteins. And we can use those as a seed to mine basically hundreds of thousands of genomes and metagenomes to really find new proteins, which we cluster and can build new families around. And when we did this originally, we found, of course, you know, the Type V and Type VI CRISPR systems, Cpf1 and Cgc2, or Cas13a now it's called, were kind of the first examples of that. And as us and others on this sort of conference have shown, you know, these are really diverse classes, right? Type V systems are— I can't even keep track of the lettering that it's gone down to. I think it's like Cas12K and L. I think someone showed L most recently. So now everyone's trying to find Cas12M. But if we know, and Type VI now is down to, I don't know, I think E, Cas13E. So, you know, we've really learned that there's such a remarkable diversity of proteins and RNAs. And, you know, as Fung has published most recently, the ancestral systems, there's probably hundreds of thousands or millions of RNA-guided nucleases that exist out there. So it's really remarkable kind of the Cambrian explosion of our understanding has been for CRISPR diversity. Now to dive into our diagnostic side of things, you know, a lot of that has been around Cas13 and Cas12, and it revolves around the very interesting activity called the collateral activity. And why Cas13 is unique is because while it can target RNA and RNA genomes of phages and viruses, —as shown here—and can cleave, we actually found that not only did it just cleave specifically, but it could actually cleave all over the RNA, degrade the RNA, would cleave other RNAs in a cell, and eventually could lead to even cell arrest and cell death. We first found that in the tube because we were incubating basically these beacons, which have a fluorescent label. And we found that, of course, if you incubate the guide RNA without and Cas13 without a target, nothing happens. But if you then incubate a target RNA, we found not only did you see cleavage of this black target, but you also got cleavage of these labeled RNAs. And of course we were running these out on gels and we realized at the time, well, hey, these beacons are kind of telling you that there was presence of this target. And so it kind of hints at the fact that perhaps you could use collateral cleavage as kind of an amplification of signal for diagnostics. And that was kind of the aha moment that led us down the path to try to do that. And so instead of using beacons that you run out on a gel, we were interested in something a little bit more scalable and easy to use. And so we switched to quenched beacons, which are a fluorophore quencher on a probe. And when they're cleaved, the fluorophore is allowed to fluoresce. And we basically showed that you could use this to target Cas13s against specific transcripts. And that was diagnostically relevant. Now, of course, these systems are not that sensitive. They go down to maybe a million copies or hundreds of thousands of copies per microliter. To build out later sensitivity, we had to combine it with preamplification. And so we can take either RNA and put that into like RPA or LAMP or all the 3, 4-letter acronym soups of applications that exist, you know, NASBA, SDA, HDA, so on. And that can then be used to transcribe RNA, which then is detected by Cas13. You get really good single molecule sensitivity, the stuff that people can actually use in the clinic or for other applications. So of course, I don't want to dive into everything we've done in diagnostics world. We've shown infectious disease, bacterial antibiotic resistance detection, freeze drying and paper-based testing, genotyping, cell-free cancer detection. I will dive into some of the more recent work from the past year doing COVID detection. And I think it's kind of been a remarkable kind of test of CRISPR diagnostics. You know, multiple groups— this is, I think, when Sherlock Biosciences, a company we co-founded, first got FDA approval for their lab-based CRISPR test. And, you know, of course, you just heard from Lucas, and Mammoth has their own version. Of a CRISPR-based test. And, you know, these are pretty cool because it's the first sort of FDA, you know, clinical use cases for CRISPR, you know, kind of beat the therapeutics out there into the market. But, you know, at the time this was happening, we were excited, but we also wanted to go further. Lab-based testing was not the end-all be-all. Of course, there's a lot of other things that can do that, But what hasn't happened, of course, is, you know, this future of like, what's the next smart home, right? Like, it's cool to have like these devices like Google Home where you can like have it tell you the news and weather every morning and like dim your lights to like a million different colors. But wouldn't it be cool to have like a Google Home that could also like, you know, take a saliva sample and then tell you if you have like the cold versus flu or, you know, COVID or, you know, any other, you know, thing that wakes you up in the middle of the night and you're like, wow, am I sick? Like, you know, it's 2021 and we still use a thermometer basically to diagnose everything at home. There isn't anything too much more than that yet. So, you know, we really are excited about that vision and wanted to make at least a CRISPR-based test that was more amenable to that. Because of course, you know, the versions that these early lab-based tests came out with are complicated. There are multiple steps. You have the pre-amplification and you transfer that to the CRISPR diagnostic test and you have to purify the stuff beforehand. It's not something someone can do in their kitchen. So we spent basically January of 2020 when we started getting a lot of emails from Asia about this virus that was taking hold and started working on this new chemistry. At one point it was, we were like literally the only lab open in our building when everything shut down 'cause we were working on this, but we were able to actually combine using this amplification called LAMP with thermophilic Cas12 enzymes into a single one-pot reaction where everything worked together happily. And we actually could take a swab from your nose, you dip it in an extraction buffer that pops the viruses open. We had magnetic beads that bind the RNA that comes out. And then we could take off the extraction buffer, bind the magnetic beads, and then resuspend in this, what we call Stop COVID Master Mix, which is really just lamp and these enzymes that work well together. And then we could actually use a lateral flow compatible reporter, use a fluorescent reporter, and do real-time detection. And it was really simple. We even had like pictures in, I think the paper, I think they made it in the SUP of like someone using this like kind of like in a kitchen with like a sous vide machine and a water bath because it runs at a single temperature. And it's pretty simple to use. And we tested like a couple hundred patient samples. And showed that the time to threshold using the real-time detection approach actually matched really well with CT values measured by a gold standard qPCR test, which is pretty cool. And we got a total sensitivity with this of about 93%, and specificity was about 98%, as shown here. So it worked reasonably well. Going from there, I think we've spent a lot of time, we've shown multiplexing, we've shown that these things can be low cost and they're quantitative. But as I've said, the really big vision here is taking out these simplified chemistries and getting them to work on a device. And a lot of that involves figuring out how to lyophilize the chemistry, how do you make it room, shelf stable for a long time and then building the device. And I hope in the future, will actually be able to have this. It's come along quite well, but a mini device you can hold in your hand that runs these chemistries, and probably no one will ever know that there's a CRISPR chemistry inside of it, but that's what will hopefully be powering it, and it's something that people can use in any setting. And I think that's hopefully what we're still working towards and trying to build. So, yeah, so on that note, this is, you know, what we've been doing in the diagnostic space. And of course, there's a lot of cellular-based applications. So I'll pass the baton to Jonathan. Yeah, great, cool, perfect.

Well, yeah, I think that's a great kind of one half the perspective of how we can use RNA targeting tools or CRISPR tools for diagnostics. But obviously, one of the amazing things we can do with these tools is we can actually have therapeutic mammalian work. And— you can probably see the TC hood behind me. So how do we actually apply these RNA targeting tools in mammalian contexts? And it's really great because there is a lot of actual RNA targeting work in mammalian cells. So there is a lot of RNA in the cell, as you're probably well aware. And that allows us to do a lot of different things where we can actually target RNA and manipulate it. Of course, there's mRNAs. There's other RNAs in general. So being able to actually target that RNA and manipulate it allows us to do a ton of different applications. So what can we do with that? So obviously, one of the first natural things we can do is knock down transcripts in mammalian cells. So we can actually target in mammalian cells a transcript like a, uh, here a reporter transcript where we have a luciferase a gastric luciferase and a cryptorhinaluciferase. We can target one of them and use the natural RNase activity of this Cas13a enzyme to actually knock down and cut that mRNA. So what we can see is that if we target this, we can measure the ratio of these two luciferases and say, is this active in mammalian cells or not? And luckily it was. So we actually compared the Cas13 family to a kind of gold, uh, or state-of-the-art technique at that time, RNAi, which you obviously are familiar with for knocking down RNA. We found that we could get comparable or better knockdown compared to RNAi. And it's also interesting because you can see that we could try both a cytoplasmic and a nuclear functionalization. So this speaks to the fact that, of course, by putting different nuclear localization sequences or nuclear export sequences on these enzymes, we can actually export them to the cytoplasm or put them in the nucleus where they can target things like nuclear long non-coding RNAs. So this is actually a great proof of principle. Obviously, with these CRISPR-based tools or any of these nucleic acid-based tools, the question is, how specific are they and what's the actual effect on the cell? So what we did is we actually did transcriptome-wide sequencing in these HEK cells And we can compare the effect of a targeting guide and a non-targeting guide to see how true or non-true to each other are these transcriptomes. So if we plot this for something like RNA interference, you might say, how much deviation from the central axis is there that would cause a non-targeting guide to vary from a targeting guide? And that's RNAi. So there's actually a bit of density along that central axis, which is indicative of the fact that there's a lot of off-targets. And if you use one of these RNA-targeting CRISPR tools like Cas13a and compare, you see, wow, it's much tighter. And those two dots over there, those are these basically the targeted transcript of the Gaucillocipherates. So we can do this with multiple Cas13 family members multiple people in these sessions mentioned, it's an ever-growing family. So both with Cas13a and doing the same thing with Cas13b, we can get very precise knockdown in these cell types. Now you're probably saying, wait, that entire first half that Omar talked about was based on this fact that you can target RNA and then activate collateral activity. That's kind of discordant. And when we first saw this, we said, Well, that is kind of strange, but perhaps in this cell type and other cell types that we might want to test, that small amount of cutting isn't a problem. And that's great to see, but that isn't always true. So as these Cas13 targeting tools have proliferated, people have started to test them in different cell types, including in this example, U87 glioblastoma cells. And what groups have seen is that in targeting conditions, there can be reductions in cell growth, there can be toxicity, and there can be— this is, uh, the actual ribosomal RNA that you can run in a gel and see that's degraded. So loss of RNA integrity over time due to this presumptively collateral activity. So it makes you think, well, one, not everything that works in one cell type may be transferable to another, and not everything that works in one animal may be transferable to another. People have shown that in certain applications in mice, the Cas13 works great, but anecdotally, people have had challenges in things like zebrafish. So how do we take an enzyme that has a built-in endogenous RNA targeting capacity, but not have it have this collateral activity? And so you can probably already say to yourself, well, why don't you go back into natural diversity to find more instances of RNA targeting enzymes? So that's what we really asked ourselves. Can we go back to that ever-growing tool of public and private databases, as long as some of information of our own, to actually find another tool? So this is actually a great little piece of CRISPR biology. We talked a lot about Class IIs, and Lucas talked a little about these Class II systems. But really, if you think about how we've used these Class II systems as a source of tools, people have shied away from this. And it's comparable, actually more abundant source of diversity in these Class I systems, because there's a bunch of different proteins there. And no one wants to have to transfect or nucleoffect 6 different proteins. So really, the question is, can we find systems in here that maybe are Class I? Type III systems are known to target RNA, but maybe they have fewer proteins. Maybe there's something else there. So what we did is we actually partnered up with Eugene Koonin to explore crystal diversity. And what we found is that there are multiple different Type III systems. This is a Class I, which is just characterized by having multiple different proteins. But there's something interesting here. The Type IIIe system actually has all of these different proteins fused together into a single effector. So there's actually multiple Cas7 domains 4 of them, and a Cas11 domain in this one protein. And that's something that we started calling Cas7-11. Credit to Eugene on the, on the naming. But the question is, does this enzyme have RNA targeting capabilities, and does it have collateral? So we actually put it into a couple different contexts. We don't have time today to talk about all the biochemistry we've done, but we actually showed that, one, it's a programmable RNA targeting system, and two, it has no collateral activity. This is what collateral activity looks like on a graph. You can see that we're targeting one transcript and we have a reporter in here and we can actually see the fluorescence buildup of that collateral cleavage. There's no actual effect here from Cas711, no collateral activity. And we can actually see that in bacterial cells too. When we target a transcript like RFP with Cas13, you actually reduce the growth rate of the bacteria. It's toxic, and it's actually a programmed cell death mechanism. You don't see this Cas7-11. So this is really great to see. And we said, well, can we take this to mammalian cells? So here's the same assay, a little bit of an updated picture, 2021. But really, we're taking two luciferases and targeting each one. And what we can do is show that now that there are multiple different Cas13s, we actually can be on this target comparable or better, in this case with the Cas13a with this guide, to those RNAi. So Cas7/11 works fantastic. And we can say in these 293T cells, do we see any growth defect? No, we don't. So this agrees with what we saw in our previous papers. In 293FT cells, we don't get a growth defect from targeting with these Cas13s or with Cas7/11. Or with RNAi. But now if we start looking into these different cell models that have been shown to have toxicity, we can say in U87 cells, well, one, do the systems still work? So here you can see we still get good knockdown with these systems. RNAi interestingly does not work. And this is a known fact. It relies on endogenous machinery that these glioblastoma lines don't have. But even more importantly, if we plot the knockdown here or the percent expression, so lower expression here is more knockdown. Versus the cell viability, we can see things that, for example, PSPKAS13B is getting a loss in viability just from expression, but also additionally with knockdown. Similar things with other Cas13s, but with Cas711, we have no— the highest viability measured here, as well as a large amount of knockdown. So this is a great system for having collateral-free knockdown for targeting RNA in both biological explorations where you need to be able to do things in certain cell lines, but also as a therapeutic. So obviously this is only one piece of this larger comprehensive toolbox of RNA binding proteins, um, the Cas13s and Cas11s, as well as other RNA targeting enzymes that we are thinking about. And these allow us to do so many things that we can now modulate translation and expression Groups such as, actually, David Liu published a recent paper using recruitment of RNA-modifying enzymes to modulate the epi-transcriptome. And groups such as Patrick Hsu have shown that we can use these to modulate splicing. So it's really an evolving time. We like to think of it as if you look at the DNA toolbox where now there's so many things we can do and it's growing all the time with base editing and prime editing and CRISPR-A and CRISPR-I. RNA is— we like to think of ourselves a little bit as an underdog here, but it's still definitely a growing toolbox that we're really interested in seeing where it can go. And I know that we've talked today a little bit about RNA targeting, but to take a step back, our group is really focused about how we can mine natural diversity to build these tools for RNA and DNA targeting and essentially assemble a cellular engineering toolbox. So we believe that genome engineering is part of this much broader context of cellular engineering. And so we're thinking about a couple different ways to do this. So our precise RNA modulation, as we just talked about today, and we're working on ways for efficient programmable gene integration, um, without double-stranded breaks, um, using new technologies. So this is obviously a very active area, but can we enzymatically bypass the requirement of relying on endogenous gene repair. And to go a little bit far from CRISPR in general, but how can we actually use RNA-based techniques and engineering of natural enzymes to understand where a cell is and be able to detect that as well as control where it's going? So we're very excited about this future of cellular and part of that genome engineering. And of course, CRISPR is a huge part of that for us. So with that, we'd like to thank everybody. So we're the Abdeh-Gutenberg Lab, and of course we have tissue culture capabilities, but we're at McGovern, and this is actually a little bit outdated, pre-COVID picture of us. We're a bit larger now, but we're actively recruiting postdocs. So if any of these things resonate with you, we'd love to chat. And we'd like to thank everybody that made this possible. Including the first author on the CAS 711 paper, Austin Ozkan, and our collaborators Eugene Koonin, who is a big part of the work on 711, Feng Zhang, who is really our old mentor but really integral in StopCOVID, as well as Jim Collins and Pardi Sabeti, who helped a lot in early Sherlock work. So thank you so much for listening, and we'd love to take any questions. Yeah, awesome.

Thank you both for such a wonderful chat. I think next time actually the challenge is going to be instead of having the BSC in the background, you have to be physically doing an experiment while presenting. Slides. That's something I'd love to see. So we'd actually love to do some Q&A, but you guys have already answered most of them kind of in chat. It's like you guys have, you got some predictive capabilities there.

I tried to get him in the office, but he couldn't, you know, Jonathan didn't want to leave the lab. So always working.

Just one quick question from Anna: are Cas13 knockdowns reversible? And also along with that, how long does the knockdown effect last once cells are transfected?

Yeah, I mean, I can start off on that, but I'm not sure your color— so they are reversible. If you do actually have an inducible system, you can withdraw. And but that knockdown level is transcript dependent, of course. Some transcripts— and of course, some transcripts are degraded in different ways in different cells, and different expression of transcripts will mean that they'll restore to their levels in different amounts. So it really does depend what you're trying to target. But that is something that we find really inspiring because the idea that you can take a reversible enzyme and actually be able to gate it using things like small molecules, you can actually allow for repeated knockdown over time. And you can turn something that's as complex as programmable gene targeting and transcript knockdown into just taking a small molecule. So it, you know, we encourage your systems of interest to try that out, but we've seen a lot of interesting progress with being able to have these as a temporary tool.

Awesome. Yeah, no, I definitely agree. So there's a couple more questions in chat. Maybe you guys have some time and we can just go in and answer those offline as we move on to our next speaker.

Yeah, thanks so much for having us.