Scientist Stories: Feng Zhang, The Future of Gene Editing

Okay, next up is Feng Zhang, who will talk to us on the future of gene editing. Feng is from the Broad and the McGovern Institute at MIT. Hey, good afternoon. It's a real pleasure to be here and to be a part of the celebration for Sterling here at 75. I want to tell you a little bit about gene editing and what gene editing is and where gene editing is headed and how do we develop and apply technologies for modifying the genome with the potential to be able to treat many diseases. In the future. So one of the really exciting advances in biology has been the completion of the human genome, and the cost of human DNA sequencing has, has dropped precipitously, and so now we can sequence the genome of thousands or tens of thousands of individuals. So by comparing the genome of these individuals, we can find genetic differences that may underlie disease. So for example, some of the mutations that we have identified cause sickle cell disease or cystic fibrosis or eye degeneration. But not all genetic differences are necessarily harmful. In fact, some can even be potentially beneficial. And so other ones, including mutations in a gene called CCR5, can confer protection against HIV or decrease the risk for Alzheimer's disease. Or diabetes, and many others. So as we understand more about how genetic differences contribute to human health and also disease, one thought is why don't we take this information and go into the genome of our cells and be able to remove the deleterious mutations and put in the ones that are potentially beneficial. So this is the very idea that many folks have worked on for several decades now, and one of the approaches is based on this idea. This was developed by Gene Haber and Elsa-Maria Jason more than a couple of decades ago. The idea is to use a way to break the DNA. So think of the DNA as a very long string. In order to be able to modify a specific letter within the DNA, One thing that we need to do is to make a cut in the DNA. This is called a DNA double-stranded break. If we know where in the DNA we want to fix or edit, then simply by making a double-stranded cut in that location, we can then trigger repair machineries in the cell to then go and be able to fix those different mutations. So then the challenge becomes, how do you precisely make a cut in where you want to edit. And to do that, scientists have developed many ways to reprogram what's called DNA binding proteins. These are molecules that can be engineered to recognize a specific DNA sequence, and when we introduce that into cells, it can go and try to find the region that you want to edit, and then be able to bring other proteins such as nucleases, to be able to cut the DNA. So what we have been working on for the past number of years is to develop a specific molecule called Cas9, or other variants of it, to be able to edit specific DNA sequences in the human genome. Cas9 is from a bacterial immune system called CRISPR, and it's a system that bacterial cells use to defend themselves against invading viruses. The way it works is that it uses a small RNA to be able to recognize a specific virus sequence, and upon recognition, the enzyme Cas9 will go and cleave that virus sequence, and that circumvents the virus infection. But we can engineer and harness this system to turn it into a way to be able to search for human DNA sequences, and by introducing that into the human nucleus, we can then program this to cleave at specific mutation locations and then be able to effect genome editing. So here is a short video that kind of summarizes the way that Cas9 from the CRISPR system allows us to edit human DNA. And so what you see here, the blue glob, is that Cas9 protein. And we can program it with the RNA molecule that's shown in red, that's supposed to recognize a specific human genome sequence. And by introducing this into the human genome, into the nucleus, this Cas9 protein with RNA guide can search along the human DNA and try to find where the RNA matches the DNA. And so that allows us to localize the enzyme to the site that we want to edit. And upon recognition, the enzyme then cleaves the DNA, which is in blue, into two broken ends. And that is what triggers the repair machinery in the cell to be able to make a repair. And so one way to repair it is called non-homologous end joining. It simply reglues the two ends together, but it introduces a small change that can inactivate a deleterious gene. Even more powerful is you can provide a piece of synthetic DNA called a template DNA. And that template DNA can carry almost any sequence you want to introduce into the genome. And through homology, or through similarity, that new piece of synthetic DNA can get incorporated into the genome to replace the sequence or to fix a disease-causing mutation. And so this is the way that we go about using this DNA nuclease to make precise changes either to inactivate deleterious genes or to be able to introduce genetic sequences that are potentially beneficial for improving human health. In addition to using Cas9 as a nuclease, there are other ways to use Cas9 as well. And that's done by inactivating one of the domains on Cas9 that's responsible for cleaving the DNA. And so when you do that, then what what will happen is that this molecule will now simply find DNA, and then once it finds it, it will sit on the DNA and not do anything more than that. This provides us a way to be able to localize things to a specific location on the genome without actually modifying it, and so then we can bring other types of molecules to that region to turn a gene on or turn a gene off, or to be able to visualize how the genome is dynamically rearranging itself. And so these are the two really main ways that we go about using the CRISPR-Cas9 system for gene editing, either to cleave DNA or to use it as kind of a localizer to bring different things into a specific location. As we're working on Cas9, One of the things that we wondered is Cas9 is a very powerful molecule, but are there more molecules that are like Cas9 or related to Cas9 that may allow us to do other things to be able to study biology and to be able to treat disease? So we went back to the drawing board, and we looked at the diversity of CRISPR-Cas systems. It turns out that CRISPR is a very diverse set of microbial immune systems. In addition to Cas9, which only constitutes a very small portion of the many, many types of CRISPR systems, there are in fact many other types of CRISPR systems. But the other types that scientists have identified at that time were much more complicated. Unlike Cas9, which uses just a single protein to do everything, These other systems use many proteins that have to assemble together in very precise ways to be able to facilitate the recognition and also destruction of viruses. So we thought if Cas9 exists, could there be other simple single protein systems that may also be harnessable to develop different biotechnological applications? So we've been conceptualizing this as follows. We can group the CRISPR-Cas system into two types. There's the Class I system, which are the complicated ones that use many different proteins to do what we want it to do. And then there's the Class II system where Cas9 belongs, and it uses just a single protein. And so one question is, are there other Class II systems that we can find from the natural diversity. So by working with our collaborator Eugene Koonin at NIH, we developed a computational pipeline to search through the bacterial genomic sequences to see are there new types of Class II CRISPR systems. And so by following through a series of steps where we begin with one of the molecules that's very well conserved in all of CRISPR systems, and using that as a bait, to find other putative, yet to be identified CRISPR systems, we were then able to identify a number of candidates that represent potentially new CRISPR enzymes with novel functionalities. So at the end of that search, we found a number of new systems, and this is more or less up to date depiction of what we have found. So there are 3 main types of CRISPR class 2 systems. There's the Type 2, which contains Cas9, but in addition to that, we also found Type 5, which contains Cas12. This is also a DNA targeting system, so you can also develop it for gene editing applications. And then the third subtype is the Type 6 system, which carries a protein called Cas13, and unlike Cas9 and Cas12, Cas13 targets RNA instead of DNA. So RNA is the other molecule in living systems that transmits information from DNA into protein. So we've developed Cas9 and Cas12 for a lot of different applications in gene editing, but I thought what I'll share with you today are some of the applications we've been developing with Cas13 to target RNA. So Cas13 is— is a very diverse family of proteins. In fact, there are 4 different members of Cas13 identified to date, and they're called Cas13A, B, C, and D. And they all target RNA with a guide RNA, and so that means you can programmably target new molecules of RNA, mRNA, non-coding RNAs in cells, by simply giving the Cas13 protein a short RNA guide molecule. And upon recognition, then Cas13 can facilitate modification of that RNA. So one of the applications that we developed using Cas13 is to make it into an RNA editing system. So rather than changing DNA sequences, which can have long-lasting consequences, we can use RNA editing to transiently make an alteration in the cell to be able to achieve a change to the biology. For example, we can rescue a genetic disease-causing mutation, or if you want to transiently alter the function of protein, we can also do it at RNA level. So the way to do this is we alter this Cas13 molecule very much like what we have done with Cas9. You can introduce mutations into the regions of that Cas13 molecule that's responsible for cleaving RNA to inactivate it so that it simply binds to RNA. And so when we do that, we can then attach onto this Cas13 molecule a subdomain that's belonging to an RNA editing enzyme called ADAR. ADAR converts adenosines into inosines, so that allows us to make changes at the RNA level that will affect single base changes. So by taking Cas13 fused with ADAR, we can then program it with a short RNA guide to direct this molecule to a specific mRNA in the cell. And when we do that, then we can specify the conversion of a specific adenosine on that RNA by mispairing that, that base with a cytosine on the guide RNA. And so when you do that, you can then program the system to alter specific RNA molecules. So this is a more detailed version of what that looks like. So on top is the RNA that we want to modify, and then in red, and then extended with this hairpin, is the guide RNA molecule. And so this guide RNA will find the target RNA through base pairing between the guide sequence shown here in red and also the target RNA. And if we want to edit this specific adenosine, we simply mispair that with a cytosine so that we can make a precise change. In that specific molecule. So one application of this could be that this mutation would represent a specific change of amino acid that caused disease, maybe in sickle cell disease or in a congenital eye disease. And by editing that RNA, we can then correct the protein product so that we can rescue the disease phenotype. So then we tested to see how well this system works. And in fact, it can work pretty robustly. So when we design a guide RNA, we can alter the target RNA with over 90% efficiency. But the first version of the system that we developed was not perfect. In fact, it introduces other unwanted changes within the cell. And so by sequencing the cell, even at a shallow depth of 12.5 times the coverage of RNAs in the cell, we find that there are a number of off-target activities. And so what that means is that we're not only making the desired change, but we're also making changes that we don't want to be introducing into the cell. So to solve that problem, we then started to try to understand what's going on. And so by using the same system with the RNA guide that doesn't target the RNA that we're trying to modify, we find that we still get very much the same off-target modifications within that cell. So what that means is that these off-target changes are likely introduced not by the guide RNA, but by the enzyme. And that gave us a path to then go and try to improve this system. So by looking at the crystal structure, the atomic shape of this molecule, we can then design ways to be able to minimize the potential association between this molecule and also the unwanted changes. And so by identifying all of these different residues on this molecule, we can then alter them so that we can try to get it to recognize RNA in a more specific fashion. So we tested a large number of them, and for example, some of these residues, like T375G, can significantly improve the recognition of the desired molecule while ignoring the molecules that we don't want it to modify. And this is the result of that change. So the first version of the system, you can see that in addition to making the desired change, we also get a bunch of other changes in the RNA. Whereas once we engineer this protein to be more specific, we can get it to be much, much more specific. Specific. So we then also profiled the whole cell to see, in addition to making changes on the RNA that we're trying to modify, is it also changing other RNAs that are not recognized by this molecule? And so this is the original version of the system. This is the desired change, but you also get many, many other undesirable changes. But then once we start to engineer the protein, we can really clean this up. So now we get the desired change and very few unwanted modifications in the cell. And in fact, now most of these mutations, they don't really have a functional change to those specific transcripts. So this is part of what we're trying to do to develop this into a system to be able to deliver into the body to be able to treat specific disease-causing mutations. So to take this further forward, one of the things we've been doing is to develop a way to deliver it. And one way to do that is to package this into a virus-based vector so that we can deliver— use this virus to deliver into cells in the body. And so we deliver this into neurons and try to see, can we make precise changes and at an efficient level in the neurons that we try to modify. And so what we find is that indeed, if we want to modify the specific DNA, RNA base in this neuron, we can do it fairly efficiently. So this is one way that we're trying to advance the technology forward, but another way is to expand the capability of what we can do. The ADAR enzyme naturally converts A to I, and that's only one form of a change. But it would be good to be able to make other changes. And so we can engineer ADAR to do that. And so one of the things we did is the original system is called repair, which converts A to I. And if we use this to convert C, it doesn't really alter this cytosine residue. But if we introduce engineered mutations into this enzyme, we can then get it to change C to a U rather than A to an I. And so now we can develop a second system called Rescue that then converts the C into a uracil. And that will allow us to make even further number of changes within the RNA of these cells. Then finally, as we start to engineer the system to make it more robust, more specific, and also more efficient, one of the things that's really helpful is to be able to know how this protein looks so that we can take the structure and then use a rational approach to further engineer the system. And so of all the family members of Cas13, A, B, C, and so forth, we found that the B family, the Cas13B molecules, are the most robust in human cells. And so if you compare their activity, the ones on the bottom left corner are the ones that are most robust, and most of these are the blue ones. So Cas13B. And so to solve the structure, we then used X-ray crystallography, and we solved the crystal structure for Cas13B. So now we know this is how it looks, and we can go in and start to tweak many of these different positions on this molecule to make it more specific so that we can edit RNA in a precise way, and also make it even more efficient. So we're continuing to further improve on the system, make it more specific, more efficient, develop delivery methodologies so that we can use it to edit cells in a very, in an efficient way in the body, and also beginning to test this using animal models to see how well we can really use this system in a therapeutic context. So we have been able to do a lot with CRISPR systems using Cas9, Cas12, Cas13. Then one of the questions that we ask is, can we generalize this idea even more? Can we broaden it to looking at other systems in the microbial world that may be quite interesting? And so one of the things that we're continuing to work on to further push these gene editing technologies also beyond is to look for other biological systems that are putatively programmable. So if you think about the ones that we have worked with so far, Cas9, antibodies, or even the the microbial tail proteins, most of these belong to immune systems. And these are naturally occurring, but rapidly diversified systems that nature uses to be able to defend against different forms of invading species. And so we are now asking the question, are there even more programmable systems beyond the ones that we have identified so far? And furthermore, By looking at microbial-based adaptive immune systems, we may be able to identify some of these new putatively programmable systems. So by taking a computational approach, we began to scour the microbial diversity, and so using computation you can identify many, many different candidates. I'll just show you a few interesting examples. This is example 47 out of 9,000 of them. And you can see that this is another repeat-containing protein that carries regions with variability that's linked to a protease domain. And so one hypothesis is that maybe these repeat residues allow us to develop new proteins that can recognize altered protein substrates to be able to degrade them in cells. Another example are these secreted proteins that may act as immune molecules in the bacterial world. And so these are all highly conserved proteins, and you can see if you line them up, they are largely similar except for a fixed region that's very, very variable. And so by trying to understand what these variable regions may be doing, we can then further try to understand how we might be able to harness this molecule to develop interesting applications to recognize diverse substrates. So this is probably just the tip of the iceberg, because as we sequence the bacterial diversity, we're now accumulating large numbers of genomic sequences. And so this is a plot to date. So as of now, we have about 140,000 bacterial genomes. And as sequencing technology further advances, we're going to accumulate many, many more bacterial genomes. And from that, we'll undoubtedly be able to identify many more interesting microbial physiology and microbial mechanisms. And some of them we might be able to harness and develop into powerful gene editing systems or even for additional applications. So that's where we are. But it's really a very exciting future. But finally, I just want to acknowledge my team. At MIT and the Broad Institute who have worked with me to develop many of these ideas, and also our collaborators and the funding agencies that generously supported our work. So thank you very much.