Scientist Stories: J. Keith Joung, Reprogramming the Genome & Epigenome with CRISPR

My name is Keith Jun. I'm a professor of pathology at Harvard Medical School and an investigator in the Department of Pathology at Mass General Hospital, where I run a research lab that's focused on technology development. I want to thank the organizers for giving me a chance to tell you about an as-yet unpublished story that we and the folks in my lab are very excited about. So as you can see, it's about defining impacts of human genetic variation on CRISPR gene editor off-target effects. So I'm required by my institution to show you this conflict of interest slide, so I've done my duty there. So my lab for the last 15-plus years has worked on gene editing technologies, and these are broadly applicable technologies with many, many applications for research and therapeutics. And as many of you know, there's been a tremendous amount of excitement in the field about them. I wanted to define first what they are. So these are technologies that allow us to be able to go into living cells and in a targeted customizable fashion, be able to make genetic changes in living cells or organisms. As this slide shows you, gene editing technologies is not just one platform. There are actually multiple modalities. Many of you know that in the early days of the field, we had technologies like zinc finger nucleases and meganucleases, followed later by the TALENs. But over the last 9 to 10 years in particular, there's been a rapid development of the CRISPR-based technologies, first the nucleases and then second generation, technologies like the base editors and the prime editor technology, the latter two developed by David Liu's lab at Harvard. So all of these technologies, as I say, allow us to go into cells or organisms and make targeted changes. But what they also have in common is that they can all induce off-target effects. So what do I mean by this? Well, we've gotten very good at being able to put these changes at specific locations in the genome, shown here with the green box, but we know also that all of them are at least capable of making what we call off-target off-target changes, so changes to the genetic material at other locations in the genome. These are unintended changes. And what we know from a lot of different studies is that off-targets are not random. They typically occur at sites that resemble the on-target site. And generally, the more mismatches a site has in its DNA sequence relative to the on-target site, the higher the frequency of off-target mutations that you see at those sites. Now, fortunately, there's been a lot of progress in the field over the last 6 or 7 years with the development of assays that allow us to be able to define these gene off-target effects. These are some examples shown here, including the GUIDE-seq method originally developed by my lab, as well as Circle-seq, an in vitro method that we described a couple of years ago. However, all of these different assays suffer from a challenge, which is that they don't have the ability to take into account human genetic diversity, the idea that all of us are different in our DNA sequences when looking for off-targets of gene editors. So what do I mean by this, or what's an example of this? So all of the assays assays described to date are really designed to be done on a single genome at a time. And so typically, people are doing them on a reference genome or a particular cell line. And when you practice one of these current methods, you— obviously, you know the on-target site, and then the assays tell you what the off-target sites are. And so in the way I'm showing you things here, I've highlighted in the off-target sites those base positions that are mismatched or different from the on-target sites. So you can see, for example, in this top example, there are 3 mismatches, in the next one, 4 mismatches, and in the next one for mismatches. So the challenge comes when you start to think about applying these editors to the global population of people and their different genomes. And that's, again, because we are all different. We all have small differences. And for example, a single nucleotide polymorphism can potentially make a difference in off-target. So what do I mean by this, and what are some of the examples? So again, here's the same on-target site, but it's possible that in a particular individual, or set of individuals, there's a SNP or a single nucleotide change or difference at one of these positions. And you can see it's in this same site here, this off-target site that has 4 mismatches, but this change causes this site now to have this position look like the on-target site. And so you go from a situation where you have 4 mismatches to only 3 mismatches. And recall that I mentioned earlier that the fewer mismatches you have, the higher the frequency of the off-target editing or mutations that you see at that site. So here's an example where the presence of a small single base difference in an individual can actually potentially lead to an increase in off-target editing. Here's another example. Now, take this site that has 3 mismatches. In this case, a SNP lands at one of the mismatched positions, but instead of taking it back to a base that's the same as it is in the on-target site, it's now a different base. So it's not the C that's here in the original off-target site. It's not the G in the on-target site, but it's actually a T. So this is a 3-mismatch site becoming a different 3-mismatch site. And so here, it's a little bit unclear what happens, right? But there is a difference between these sites. And we know from nuclease studies that this type of a change can make a difference in what you see. So for therapeutics in particular, we're worried about these types of differences that can then lead to a change in off-target frequencies. Now, increasingly, there are efforts to sequence larger and larger numbers of human genomes. So for example, this is a paper describing the final results from the 1000 Genomes Project, in which they sequenced 2,504 individuals from a globally distributed set of people that is distributed throughout the globe. And what this highlights, then, is the limitation of all existing off-target assays. So, for example, if you wanted to determine what the impacts of all the different genetic differences in these 2,504 individuals were on the off-target profile of just a single gene editor, you would actually have to do 2,504 individual assays. And that's because all existing assays are really set up to be done one at a time. Individual goes into a single reaction tube. So if you have 2,500 different genomes, you'd have to do 2,500 assays. And again, that's for every single gene editor that you wanted to determine off-targets for. So we developed a new method that we call OneSeq. And the big advantage of OneSeq is that it's designed to be able to handle the input of multiple genomes. And that's because we take a novel approach where we computationally analyze these genomes looking for the most closely matched sites. And then we leverage high-throughput oligonucleotide synthesis to be able to synthesize all the different possible off-target sites that exist in these individuals. And so we can take all the data, for example, from 1000 Genomes Project, 2,504 individuals, and in a single reaction tube now, assess what the impacts are of all of those on the off-target profile of a given gene editor. Now, the other thing I wanted to mention about ONCE, you can emphasize, is that it's actually a universal off-target assay. We believe that it will work with all gene-editing nucleases regardless of the type of break that they induce. And we've also demonstrated that it's actually better than existing methodologies for detecting off-targets of base editors. So let me give you some examples of how OneSeq actually identifies these types of sites. So we did OneSeq with two different Cas9 nucleases programmed by these two different guide RNAs shown here. And OneSeq actually identified two different SNPs or variants in the genomes of individuals from the 1000 Genomes Project in which a single base change changes the number of mismatches in the sites. So for example— and I should say in vitro also leads to a difference in cutting activity. Activity. So for example, here in the top, I'm showing you the on-target site. I'm showing you the reference genome mismatch site. And then I'm showing you the site that OneSeq identified. So you're going from 3 mismatches to 2 mismatches, and so on for all the others. And so the amazing thing is, is that these sites show increased editing activity in vitro in the OneSeq assay. But when we looked in human cells, they also show increased evidence of off-target mutations. So these are bona fide off-targets where the presence of a SNP is actually increasing the frequency of off-target effects. Now, the other thing that's interesting to look at is how are these variants distributed in the global population of people that were sequenced in the 1000 Genomes Project. And so this heat map attempts to represent each of these 4 different variants, the frequency of which is found in different human populations as well as across different superpopulations. And so the darker these bars, the more frequent the mutation is. And so you can see that, for example, these 2 variants for the DMX1 guide RNA are actually enriched in people from the African superpopulation. This one from the FANCF guide RNA is enriched in a population within the admixed American superpopulation, that is within the Colombian population. And then this variant is actually enriched in two different superpopulations, the admixed American and the European population. So the important takeaway from this is that one-seq can identify these variants and show, that actually show increased target mutation frequencies in human cells. And also, it's interesting to be able to look and see that some of these variants are actually enriched in specific human populations or superpopulations. And this becomes increasingly important as one starts to think about applying all gene editors to larger patient populations. I just want to touch on briefly the commercialization of OneSeq. We filed IP to cover the method. We think it's ideal for commercialization, both because it's universal and also because it has the unique ability, as I just showed you, to account for genetic variation. And finally, because the assay is experimentally and informatically complex. We think there are multiple broad markets for this type of assay in research, as well as preclinical evaluation and screening of different editors, clinical trial filings, and long-term, ultimately, for individualized testing. And I do want to just mention quickly that we are spinning out OneSeq and some of the other off-target assays developed in my lab into a startup new co-presently. So I'll stop there. I want to acknowledge the folks in my lab who did the work, in particular Vikram Patnaik, who's a junior faculty member in my lab who led the effort to develop OneSeq, and Karl Petrie, a postdoc who worked with Vikram. A lot of different individuals contributed to this project, and in particular, I want to acknowledge DARPA primarily for funding it, as well as NIH and the MGH research scholars. And so I'll stop there, and I'll be happy to take any questions you might have.