Precision BioSciences, Inc. (DTIL) Earnings Call Transcript & Summary

[Presentation] Good morning, everybody, and welcome to Precision Biosciences Gene Editing R&D event. This is the first time that we're going to gather to focus on our gene editing pipeline. My name is Alex Kelly, and I'm the Chief Financial Officer at Precision BioSciences, and our team is really excited about what we're going to do today. So our company has been around for 15 years. And during that time, the dream has always been about creating genome editing products to help patients. And we've come a long way. We've got 4 products in the clinic that are focused on cell therapies. But today, we want to put a spotlight on gene editing and the great products in our pipeline. So we're very excited about what's going on at the company and looking forward to sharing with you more about this pipeline. So first, we're going to have a few things on the agenda today. But we're really excited because Derek Jantz is going to give you a deep dive on ARCUS. And in that discussion, you're going to hear about really 2 key features about ARCUS and what makes it special and differentiated. It's about precision and it's about versatility, and Derek will tell you all about that. Next, you're going to hear about our pipeline. And you're going to hear that over the next 3 years, we're going to file 3 INDs or 3 clinical trial applications for our programs, starting with PCSK9, PH1 and hepatitis B. The third thing that we're really excited about today is a new collaboration with iECURE, and that iECURE collaboration is going to help us in a number of ways. Number one, it's going to expedite the clinical development and validation of PCSK9 as a gene editing target. Number two, we're going to focus on gene insertion. And I think also at the end of the day today, you're going to hear from Dr. Jim Wilson at University of Pennsylvania, talking about some great new data that his team has generated that focus on gene insertion. And they're going to show you that with ARCUS, you can generate -- you can insert genes into a nonhuman primate, first-ever data. So we're looking forward to that. But let me first introduce Derek Jantz, who is our Co-Founder and Chief Scientific Officer. Derek?

Thank you, Alex.

Thank you.

Good morning, everyone, and thank you very much for joining us. We are not going to talk at all today about CAR T cells. That's not because we don't really like CAR T cells, we do. We're still very interested in that. But what we wanted to do today was really focus a spotlight on our in vivo gene editing programs. So our agenda for the day looks like this. We're going to start off with a pretty deep dive into the ARCUS gene editing technology. I'm going to tell you what it is, where it came from, how we use it and why we think it really is the best-in-class technology for in vivo gene editing. At that point, we're actually going to pause and take a few questions. So if you have any burning questions on the ARCUS technology itself, you probably want to queue up in advance of that. Then we're going to talk about our gene editing pipeline and strategy. We're going to highlight 4 programs this morning. We're going to talk about PCSK9, primary hyperoxaluria type 1, chronic hepatitis B and Duchenne muscular dystrophy. And at that point, Cindy Atwell, who's Precision's Head of Business Development, is going to join us to talk a little bit about our partnering strategy to talk about our partnership with Eli Lilly and to talk about our brand-new partnership with Jim Wilson's new company, iECURE, that we just announced this morning. And then we're going to wrap things up, talking about a couple of things that are sort of on the horizon. We're going to talk about mitochondrial gene editing. And then as Alex mentioned, at the very end, Jim Wilson is going to join us from the University of Pennsylvania to share some brand-new data demonstrating targeted gene insertion in nonhuman primates. So we're definitely going to finish on a high note. So let's go ahead and get started. The story of ARCUS really started in a pond somewhere tens of thousands of years ago. The green stuff that grows on the surface of a pond is primarily an organism called chlamydomonas reinhardtii, which is an algae. It's a single celled aquatic plant. I wanted to start here because I think in order to understand ARCUS and to understand how it's differentiated from the other technologies, it's important to understand where it came from. And this algae actually figured out how to do gene editing thousands of years before people did. ARCUS is based on a naturally occurring gene editing enzyme from chlamydomonas algae that's called I-CreI. I-CreI is a member of a larger class of enzymes called homing endonucleases. Homing endonucleases are nature's gene editing technology. So nature doesn't use CRISPR or zinc finger or talen to do gene editing. Nature uses homing endonucleases. And in the next few slides, I'm going to show you why that is. The natural job of I-CreI in the algae is it gets expressed in the cell and then it hunts through the genome until it finds a particular 22 base pair DNA sequence that shows up from the 23S ribosomal RNA gene. When it finds that site, it binds to it and cuts it. And very, very importantly, as you can see by the red scissors in that figure, I-CreI cuts the top strand of DNA and the bottom strand of DNA, offset by 4-base pairs. What that does is it creates a 4 base, 3 prime overhang, so a short stretch of DNA that's single stranded on the 3 prime end of the cut. Remember that because it's going to be very important a little bit later in the story. After I-CreI cuts its target site, a new piece of DNA actually gets inserted precisely into that location in the genome. And that new piece of DNA, in fact, encodes the gene encoding I-CreI. So I-CreI is an example of a selfish DNA. It exists for the sole purpose of inserting a copy of its own gene precisely into this location in the algae genome. So the real takeaway here is I-CreI is an enzyme that evolved for the purpose of inserting a new gene sequence into a defined location in the genome of a eukaryotic cell with a really, really big genome. The DNA sequence that I-CreI recognizes that I showed you on the previous slide, as far as we know, that sequence doesn't show up anywhere in nature other than the algae genome. So if we want to use I-CreI to do gene editing in genes that we're interested in, we need to reengineer it to make it recognize a completely different DNA sequence. The wild type enzyme is a homodimer. So it has a left monomer and a right monomer. The left monomer recognizes the left half of the target site, and the right monomer recognizes the right half of the target site. Because the left monomer and the right monomer are actually identical to one another, they recognize the same DNA sequence. So the left half of the target is basically a mirror image of the right half. What that means is in order to reengineer this enzyme to make it recognize a completely different DNA sequence, we actually have to engineer 2 proteins. We have to reengineer a first monomer on the left, that's the green one on the slide. That recognizes the left half of our new target. And then we have to engineer a completely different second monomer, which is the purple one, that recognizes the right half of the new target site. We then take those 2 halves of the protein, and we fuse them together into a single protein that is expressed from a single gene. And that's what we call ARCUS. All of that protein engineering is really pretty involved. And that's because a very significant percentage of the surface area of I-CreI is, in fact, involved in recognizing and cutting its DNA target, which is a good thing. That means our protein engineers have a lot of surface to work with to optimize the nuclease for its particular function. In fact, we know which parts of the protein, we know which amino acids are responsible for controlling the catalytic efficiency of the enzyme, which determines how quickly or slowly the enzyme cuts. We know which immuno acids are responsible for binding specificity, which determines what sequence it binds to and what sequence it cuts. And we know which amino acids are responsible for binding affinity, which determine how tightly the enzyme binds to its DNA target and how long it remains bound to the target after it cuts. Those 3 parameters, efficiency, specificity and affinity, can actually be fine-tuned independently of one another to optimize the nuclease for its particular function. So for example, if we're interested in making an ARCUS enzyme that we're going to deliver using AAV. We know in that case, the AAV is going to persist in the cell for a long time. So the ARCUS is going to be expressed for a long period of time in the cell, days, weeks, months. So we're going to err on the side of making an ARCUS enzyme that cuts very slowly and very carefully because we want to minimize the likelihood that, that enzyme is going to find some off-target site in the cell to cut. Conversely, if we're using a lipid nanoparticle to deliver our ARCUS enzyme, in that case, we get a very short burst of expression, just a few hours or maybe a day. So we're going to err on the side of making an ARCUS enzyme that's a lot more active, that works a lot faster because it has a lot less time to get the job done. Our overall workflow for making a new ARCUS nuclease looks like this. It starts with lead generation, and the first step in that process is we identify a site or more often a group of sites in a gene that we're interested in modifying and those sites that are amenable to being cut and edited by an ARCUS nuclease. We then use 2 different protein engineering techniques to reengineer the surface of I-CreI, that I showed you on the previous slide, to make it recognize a new DNA sequence. Those 2 techniques are in silico protein design, which involves modeling structures of the protein on the computer; and directed evolution, which is an experimental method in which we screen large mutant libraries to find mutants of I-CreI that recognize the DNA sequence we're interested in. We use those 2 techniques to engineer the surface, and that gets us to a first-generation ARCUS nuclease. The first-generation enzyme will generally cut its intended target site with high efficiency but will also cut some number of off-target sites in the genome that we don't want. At that point, our lead nuclease goes into the lead optimization process, and this starts with a very thorough characterization of off-target editing. And based on what we find, we introduce additional changes into the first-generation enzyme to improve the specificity and eliminate the off-target gene editing. That gets us to a second generation ARCUS nuclease. We then repeat this process 2 to 3x generally. And usually about Generation 4 or Generation 5, we have an ARCUS enzyme that has the desired therapeutic index and that is our clinical candidate nuclease. The clinical candidate is an enzyme that cuts the intended target site with high efficiency but doesn't cut any concerning off-target sites to a degree that we can detect. The first part of the process, the lead generation, everything prior to getting to that first-generation ARCUS, that takes us about 5 weeks. And then the lead optimization process generally takes anywhere from 6 months to a year depending on how many cycles we have to take it through. So a question that I get asked all the time is, if ARCUS is so great, why does everybody use CRISPR? And the reason is shown on the screen. There's a very significant upfront investment in time and resources to make every ARCUS nuclease, and I'm pretty sure we're the only group that can actually do this. So it's a technology that just isn't accessible to the vast majority of the research community. It's not nearly as easy as just changing out the guide RNA on a CRISPR, and there are dozens of online services that will make a CRISPR in a couple of days. But what I'm going to show you in the next few slides is why that upfront investment in ARCUS is so worthwhile. A couple of themes you're going to see throughout the rest of the talk. One is precision and the other is versatility. Precision, what we're talking about here is safety and specificity or the avoidance of off-target gene editing, which obviously is very important for any editing-based therapy. But probably less obvious is versatility, and what I mean here is there are a lot of ways to knock a gene out in the liver. You can do it with ARCUS, you can do it with CRISPR, with zinc finger, base editing, prime editing. But there's a limited number of diseases that we can treat by knocking out a gene in the liver. For everything else, you need an editing technology that has the properties of ARCUS. In particular, 2 properties that I want to highlight. One is ARCUS is easy to deliver. Lipid nanoparticles are terrific for delivering to the liver. For everything else, at least today, the only real option is AAV. So you need a technology that's compatible with AAV, and ARCUS is. The other property that ARCUS has is it's able to perform complex edits. So it's not just a tool for knocking genes out. We can use ARCUS to knock genes in. And ultimately, that's what we want to be able to do. We want to be able [Audio Gap] of something that the cell needs. What we want to be able to do is provide the cell with a copy of that gene that's defective to compensate for the one that's in its genome that isn't working. So let's talk first about safety. ARCUS is unique among the gene editing technologies in that it has the ability to turn itself off. The enzyme actually exists in 2 different configurations. There is a closed configuration, which is what's shown on this slide. That's the inactive form of the enzyme. In this configuration, the active side of the enzyme is actually tucked up inside the middle of the protein so it can't access or cut the DNA. But in the presence of its target DNA site, the enzyme opens up, and now the active site becomes exposed and the enzyme becomes active and able to cut the DNA target. And then once it cuts and edits its target, the enzyme closes back up again. It goes back into the default inactive form. The reason this matters is it allows us to express an ARCUS nuclease in a cell for an extended period of time, for example, from an AAV vector without having to worry about that cell accumulating off-target editing over time. And we have demonstrated this over and over again in long-term nonhuman primate studies in which we use AAV to deliver ARCUS to a primate and we track the editing profile over time, the on-target editing as well as the off-target editing. What we find is the editing profile that we see at 3 years following administration of the vector looks more or less the same as the editing profile that we saw at 2 weeks. So despite the continued expression of the ARCUS nuclease in the cells, we don't see any off-target editing accumulating over time, presumably because the enzyme is being made, but it's being made in the default inactive form. On the topic of off-target editing, I want to get on my soap box for just a moment here and give Precision's perspectives on off-target editing and the importance of being able to detect it. The number of off-target sites that are identified for any gene editing nucleus, whether it's ARCUS, or CRISPR, or zinc finger, is a function of how well those off-target edits can be detected. And off-target sites are very, very difficult to detect for most gene editing platforms. This is particularly true of something like a base editor that doesn't leave much of a signature behind at the off-target sites that can be detected after the fact. And if you can detect better, you can engineer a better editing enzyme, meaning we can't fix the off-target editing that we don't know about. We are very, very fortunate that ARCUS, in fact, does leave a very clear signature behind when it cuts off target sites in the genome, and that allows us to find them. And more often than not, when we find off-target editing, we're able to redesign the enzyme to fix it. The method that we use in-house to detect off-target editing is something that we developed that we call Oligo Capture. And the way it works is like this. We transfect a population of cells with a gene encoding an ARCUS enzyme as well as a very high concentration of a DNA tag. The ARCUS enzyme gets made in the cells and it cuts its sites in the genome. It cuts the on-target site, and it also cuts any off-target sites that it might have. And then that DNA tag ends up getting captured at the sites of all of those breaks, whether they're on-target or off-target. We can then use NGS to figure out where in the genome that tag landed, and we can deep sequence those sites one at a time to figure out which ones represent bonafide off-target cut sites by ARCUS and what the frequency of those off-target cuts actually is. This looks a lot like a method called GUIDE-seq that was developed for CRISPR. It's really the gold standard method for detecting off-targeting with CRISPR. The really, really important difference, though, between Oligo Capture and GUIDE-seq is, as you can see on this slide, the DNA tag that we use has 4 base, 3 prime overhangs that are complementary to the overhangs made by ARCUS. What that means is that DNA tag tends to be captured preferentially at sites of bonafide off-target editing by ARCUS as opposed to being captured at sites of DNA breaks that just happened to be in the genome because of UV radiation or cell division or sort of everyday metabolism. A DNA break that's introduced by a CRISPR is more or less indistinguishable from a break made by UV radiation. That gives the CRISPR methods a really high degree of background and a really poor signal-to-noise ratio. With ARCUS, the signal-to-noise ratio is much, much better, which gives us the ability to detect off targeting with much higher efficiency. Our collaborators at the University of Pennsylvania actually demonstrated this. They used both techniques. They used GUIDE-seq, which is the gold standard CRISPR method; and Oligo Capture to detect the sites of off-target editing in nonhuman primates that were treated with an ARCUS enzyme directed against the PCSK9 gene. What they found was GUIDE-seq was able to detect one off-target site. Whereas all of the capture was able to detect that off-target site and 6 more. So the gold standard method for CRISPR, in fact, failed to detect 6 out of 7 legitimate off-target sites. This is not to say that the CRISPR groups aren't doing everything they possibly can to identify off-target gene editing. What I am saying is that technology has an inherent blind spot that makes off-target editing hard to find. With ARCUS, we are very fortunate that we can better understand, better characterize our product. And at the end of the day, that's what we want because if we know where the off-target editing is, more often than not, we can fix it. Let's talk now about the versatility of the technology. ARCUS is small and that makes it very easy to deliver. It's by far the smallest in the gene editing enzymes. It's a fraction of the size of even the new micro Cas9s. The gene that encodes in ARCUS enzyme is about 1,000 base pairs long, meaning that even with promoters and polyadenylation signals and all the parts that we need to express the enzyme in a cell, we can fit multiple ARCUS enzymes into a single AAV vector and still get high-titer, full-length virus. That is very, very difficult and, in most cases, impossible with the other editing technologies. And so what you typically see insofar as AAV is used to deliver and of the other technologies, the nuclease has to be split between 2, sometimes even 3, AAV vectors, all of which have to be delivered to the same cell at the same time, which obviously is a significant complication for any therapeutic program. All right. One final unique property that ARCUS has that I want to spend some time talking about. When we introduced a DNA cut into the genome using a gene-editing nuclease, that DNA cut can be repaired by 2 different mechanisms. The first mechanism, shown here on the left, is called nonhomologous end-joining or NHEJ. This is the cell's sort of quick and dirty DNA repair mechanism. When a break is repaired by nonhomologous end-joining, what we typically see is small, sort of random insertion and deletion mutations that we call indels get introduced into the cut site. That is a terrific way to knock a gene out, and all of the gene editing technologies are good at doing that. But the more interesting pathway is the one shown here on the right. That's homology-directed repair or HDR. In this case, we actually make a DNA molecule that we call the repair template, and we give that to the cell. The HDR process then fixes the DNA break by recombining the chromosome that we just cut with the repair template, such that the sequence of the repair template gets inserted permanently into the chroma cell. That is a great way of inserting new genes into the genome or repairing a gene that's defective. ARCUS is really good at doing the one on the right. Remember, I told you in the beginning that the I-CreI enzyme that ARCUS is based on, its job in nature is to insert a new gene into the allergy genome. The way it does that is it stimulates DNA repair by HDR, and the way it stimulates HDR is it makes a very particular type of DNA cut. It makes a cut that has a 4 base, 3 prime overhang. Three prime overhangs act as a kind of a signal to the cell that, that DNA break needs to be repaired through HDR instead of NHEJ. And this effect is already pretty well established in the scientific literature, but one of our scientists came up with a very elegant way of demonstrating this, and I want to show this to you because it's cool data. So what he did is he used Precision's process for making CAR T cells. He started with a population of activated T cells and introduced into them an ARCUS enzyme that recognizes and cuts a DNA sequence in a gene called TRAC. He then introduced into the cell a repair template encoding a gene called a CAR. That's the purple sequence that you see in the repair template. And 49.4% of the time, the DNA break in the TRAC locus was repaired through HDR with our repair template such that the CAR gene was integrated stably into the genome in TRAC. With lower frequency, 20.7% of the time, that DNA break was repaired through nonhomologous end-joining, and an Indel was introduced into the TRAC gene. So DNA breaks with 3 prime overhangs are repaired preferentially by HDR. He then repeated this experiment, exact same experiment with one change. He added an enzyme called the 3 prime exonuclease, which is the purple Pacman thing on the slide. That purple Pacman eats 3 prime overhangs. So now he introduced this into the cell, and the ARCUS enzyme cut the TRAC locus and made a 3 prime overhang, but then the exonuclease shooed the overhangs off to give us blunt ends that look like a DNA break made by a CRISPR. Now the frequency of homology-directed repair dropped off very significantly. Only 6.6% of the cells had DNA repair by HDR and acquired the CAR gene in the TRAC locus. The vast majority of sales, 58% had the DNA break repaired by nonhomologous end-joining and the introduction of an indel. So blunt end cut repaired primarily by nonhomologous end-joining. The exact same experiment, the only difference is the presence or absence of those 3 prime overhangs, and what you see is it has a huge impact on the editing outcome. What we care about at the end of the day is the ratio of HDR to NHEJ, so the ratio of what we do want to what we don't want. And there's about a 20-fold enhancement in favor of HDR when we use ARCUS and make cuts with 3 prime overhangs. Therefore, we really do think that ARCUS is the perfect technology for making complex edits like gene insertion. And at the very end of the event, Jim Wilson is going to join us, and he's going to show this at work in nonhuman primates. Precision has 2 in-house delivery platforms. We have LNP and we have AAV. We like to use lipid nanoparticles for delivery to the liver because they give us a nice, short burst of expression of the nuclease, which is all we really need to get long-term gene editing. And we in-license our LNP technology from Acuitas. We also have AAV. And as I mentioned earlier, we like to use AAV for everything else outside of the liver. We also use AAV to make our repair templates for HDR. Both technologies, both delivery technologies are compatible with ARCUS. Here, we're showing a biodistribution study in nonhuman primates using lipid nanoparticles. What you can see is the ARCUS mRNA, which is shown by the red staining, gets delivered pretty uniformly to hepatocytes throughout the liver. We see a lot of mRNA in hepatocytes in the first few hours. Following LNP delivery by about 48 hours, you can see most of the mRNA is now extracellular. And by 7 days, it's basically gone. With AAV, we see the same thing in terms of distribution. So we see a lot of ARCUS mRNA being expressed across hepatocytes in the liver. We see a lot at day 17, and then it's largely gone by day 129. So both technologies do a great job of delivering to hepatocytes in the liver. The difference is, as we would expect, the AAV sticks around for a lot longer. At Precision, we do have our own GMP manufacturing facility. We have 3 platforms: one is CAR T; one is mRNA, mRNA being the active ingredient in a lipid nanoparticle; and we have an AAV platform. And we are actively manufacturing clinical trial materials now to support our ongoing CAR T clinical trials. So hopefully now, you're all as excited about ARCUS as I am. And if there are any burning questions, would love to take a few minutes to answer them. So the instructions for calling in are on the screen.

[Operator Instructions] Your first question is from Ben Burnett with Stifel.

Two quick questions just about the DNA tag method that you spoke about just in terms of defining off-target editing. I guess it seems like this method relies on the assumption that ARCUS always cuts at a specific 4-base pair sequence. I guess is there any capacity for ARCUS to cut outside of that 4-base pair sequence?

Yes and no. So every time ARCUS cuts, it produces a 4-base, 3 prime overhang. But the exact sequence of that overhang can vary from enzyme to enzyme and from site to site. So the tag that we introduced doesn't have a fixed 4-base sequence as was shown in that slide. It's actually a mixture of tags that have different 4-base sequences, and what we see is the tag with a given 4-base sequence will tend to find its off-target site in the genome if ARCUS makes one. So we do vary the exact sequence of the tag. Great question.

Okay. Okay. Very helpful. And then you mentioned that you use ARCUS for CAR-T. I guess do you use the same off-target assessments in that setting? And if so, have you received any feedback from the FDA regarding this method of off-target assessment?

Yes. So we do use the same method for the CAR T cells. And obviously, we have 4 programs in the clinic. The FDA has generally been very, very positive about the Oligo Capture method. And in fact, some members of the FDA have asked us if we could adapt the Oligo Capture method to make it applicable to CRISPR which, unfortunately, the answer is no, because that 3 prime overhang really is critical to making it work.

Your next question is from Maury Raycroft with Jefferies.

First one, just for the HDR:NHEJ ratio, does this change based on the DNA target sequence? And does this variable ultimately influence the gene that you decide to target in that end?

Yes, it does. So that ratio can change not only based on the exact target sequence and the location in the genome. So we have found that some locations, the genome, for reasons that we don't entirely understand, are more amenable to HDR than other regions. Also, certain cell types are more amenable to HDR than other cell types. And in particular, cells that are actively dividing like hepatocytes or T cells are more prone to HDR repair, which is why the experiment that we've showed is so important because we kept all the variables the same. The only thing that we changed was whether or not that 3 prime overhang is there.

Got it. That's helpful. And then the other question I had is just if you could talk a little bit more about ARCUS activation. What regulates the activation? And what other factors influence the activation besides the DNA sequence?

Yes, so it is primarily the DNA sequence. So essentially, what happens is the enzyme gets made in sort of that closed clam shell configuration, and it sort of floats around in the cell as the inactive form of the enzyme. Then it opens up in response to binding to its DNA target site. So what that means is the target site interaction with the protein has to have enough binding energy that is able to overcome the energy necessary to open it up. So essentially, what that means is there's a threshold in binding affinity that the enzyme has to have for a given target site before it can open up and assume the active configuration. It's that threshold. I mentioned earlier that we have those 3 parameters that we can fine-tune to optimize the nuclease, and one of them is DNA binding affinity. That's where that becomes so important because what we can do is we can engineer an enzyme that has a high-enough affinity for its intended target that is able to open up, but it doesn't have enough affinity for any of the off-target sites in the genome that they're able to open it up and activate it. So as it flows around in the cell in the inactive form, as it sort of bumps into the DNA, it doesn't cut.

Your next question is from Raju Prasad with William Blair.

I just had a question on the cutting and the 3 prime overhang you mentioned. Is there any worry exonuclease activity or degradation of those arms before the repairs can be made?

Yes. And again -- all right, we talked -- I talked about those 3 parameters that we can optimize, and I talked about how 1 of the parameters that we can optimize as DNA binding affinity. What the DNA binding affinity determines, in addition to whether or not the ARCUS enzyme gets activated, it determines how long it remains bound to the target site after it cuts. What we have found is that if we increase the DNA binding affinity for its target, the enzyme actually remains stuck to the target after it cuts. And we think what it's doing is it's protecting those 3 prime ends from getting degraded until something coming along the DNA like a polymerase eventually knocks the enzyme off and the break gets detected and fixed through HDR. I-CreI has an extraordinarily high affinity for DNA, and we think the reason for that is it evolved this ability to remain stuck to the DNA to protect the 3 prime ends after it makes them. Great question.

Got it. And then we talked a little bit about the Oligo Capture method, and apologies if I missed this. Can you maybe just go into a little more depth on how you use this method to kind of proactively look at potential sites? Is it based on kind of the sequence of the overhangs and if there's sequence homology down close by? Or I mean maybe just a little more in depth into the Oligo platform and how indels or potential sites are identified, that'll be great.

Yes. Thank you for that. That's a very, very important point. No. The Oligo Capture method is entirely sequence-agnostic so we don't limit our search to DNA sequences that look similar to the sequence that we're trying to cut. Other methods, that's what they do. So they base their hunt for off-targeting on the assumption that the enzyme is going to cut the sequence that they want to cut. We don't do that. So the Oligo Capture method is entirely sequence-agnostic. We put the nuclease in the cell. It cuts wherever it's going to cut in the genome. The tag goes into those cuts, and then we figure out after the fact where the tag went. So irrespective of whether or not an off-target site is similar to the sequence of the intended target site, we can still detect it.

Your next question is from Tom Shrader with BTIG.

Thanks for the thorough review. This -- I know you got plenty on your plate, but I'm curious, can you make ARCUS nucleases that don't cut and then use them to localize things like gene regulatory proteins and base modifiers? That's a big part of the CRISPR world now. Are you interested in that? Is it a good match for your technology?

Can we do it? Yes. So we just make 2 amino acid changes to an ARCUS enzyme, and it stops cutting the DNA. Instead of being a DNA cutting enzyme, it becomes again a binding enzyme. We can then fuse that ARCUS to transcription activators repressors and use it as a transcription factor. We can attach it to base editing machinery and use it as a base editor. So you're right. All of the different flavors of inactivated CRISPR fused to something else, we can do the exact same thing with ARCUS. Frankly, we aren't that interested in doing that because there aren't that many things that you can do with a base editor, to be perfectly blunt. We really do want to highlight the versatility of ARCUS, its ability to do all of these different things by taking advantage of making DNA breaks and promoting DNA repair through HDR. That's really where we see the greatest potential and the ability to do the most good for patients. Anything else?

No, I'll stop asking about it.

Your next question is from Eric Joseph with JPMorgan.

Derek, thanks for the overview. I'm just curious to know how much of a gene -- or a target primary sequence might dictate the type of edit that can be made and ultimately with high enough efficiency and specificity. I guess if you come across any target genes of interest where kind of doing a stable knockout approach has not been feasible with ARCUS. And then secondly, right now, it seems like from the [ lecture and ] discussions that the potential to do stable gene insertion for the purposes of restorative expression using any of the various editing approaches might be a little bit limited right now. Can you maybe contrast how ARCUS is more or less suited to do or to edit longer stretches of a DNA sequence? And any kind of proof of concept work you've done in this regard?

Yes. Sure. So to the first question, it's less the primary sequence of the DNA target that impacts editing efficiency than it is the location of that target in the genome. So it is absolutely the case with ARCUS as well as all the other editing technologies that where a site is in the genome has a very significant impact on our ability to edit that target, presumably because the DNA is wrapped up in chromatin [Audio Gap] we generally don't pick 1 site upfront to focus on in the gene of interest. We'll pick 3 or 4, sometimes 5 or 6 different sites in the gene of interest, design first-generation ARCUS nucleases for all of those sites and test them experimentally to figure out which enzymes are able to edit sites that are accessible in the genome. That's how we pick the lead nuclease that then moves into optimization. As for using ARCUS for the complex edits like gene insertion, we're going to see probably the most compelling data in a few minutes when Jim joins the call and shows us gene insertion in primates using ARCUS. Thanks for the question, Eric.

Your next question is from Soumit Roy with JonesTrading.

On the -- in scenarios where you insert a corrective gene, have you seen longitude? Do you provide a promoter region or you use indigenous promoter? And have you looked at longitudinally how if the expression is affected over time?

[ Fraction ] and that's kind of nice because that allows us to insert gene -- sort of pick a particular landing pad in the genome that we want to go to repeatedly with a transgene with its own promoter. And in fact, that's what Jim Wilson's new company, iECURE, is doing. They're inserting genes consistently into the same location in the genome that have their own promoter. We've also done projects and have one significant project underway already in which we are inserting a healthy copy of a gene into the native locus. So we're basically hijacking the existing promoter and existing regulatory machinery for a gene and using that to drive expression of a therapeutic transgene. As for stability of expression, again, Jim is going to show some NHP data at the end where we show stability out past a year. Okay. We have time for one -- go ahead.

And last question is, in case of NHP models when you use lipid nanoparticle. Could you give us any color on like how deep the lipid nanoparticle penetrated. Is it just restricted to couple 2 layers of cells or it goes deeper than that?

As far as we know, we hit pretty close to every hepatocyte. So I showed multiple biopsies taken from different regions of nonhuman primate liver, and we do see pretty uniform distribution throughout the liver. So we're fairly confident that the LNPs do a very good job of penetrating very deep into the liver. Okay. We've got time for one more question.

Okay. And that question is from Andrea Tan with Goldman Sachs.

Just wondering, you've spoken about the iterative process for the ARCUS optimization. Just wondering if you have a sense as to the level of off-target frequency by that fourth or fifth generation that gives you the confidence that the enzyme would be therapeutic grade.

Yes. So it does vary from enzyme to enzyme. So sometimes we'll make a first-generation nuclease that looks actually remarkably clean. Other times, we'll make one that has a lot of off-target editing and requires a lot of optimization cycles to get to what we consider to be the clinical candidate. I'll actually show you some data a little bit later on for our primary hyperoxaluria nuclease sort of showing the difference between the generation 1 nuclease and generation 6. And what we're looking for in the clinical candidate nuclease is, number one, obviously, it has to have a high level of editing at the intended on-target site. But then number two, insofar as we are able to detect off-target cutting, we want to make sure that, that off-target cutting is in a region of the genome that we would not expect to be problematic. So typically, we would assume that if an off-target edit is kind of in the middle of nowhere in a region that doesn't encode a gene, it's probably less of a safety concern. Something that we definitely still need to consider, we still need to keep our eyes on, but it's less of a concern than, for example, if an off-target site is in a gene coding sequence.

Got it. So there's not like, I guess, maybe a specific threshold that you're looking for, just to be more about the overall like where it's located?

Yes, exactly. It's the combination of what is the frequency and where is it.

Okay. Terrific questions. Thank you so much. Now let's talk about Precision's gene editing pipeline and strategy. Starting with our overall strategy in gene editing, it has 2 steps. Step 1 is we want to achieve clinical validation with ARCUS using each of our 2 delivery platforms, AAV and lipid nanoparticle, and we're doing this using programs involving knocking genes out in the liver. So I just spent 30 minutes telling you how ARCUS is really great at doing everything else. But even for us, even for ARCUS, the most straightforward thing that we can do is knock a gene out in the liver. And we made the decision that for our first programs, that's what we want to do to maximize the likelihood of clinical success and maximize the likelihood that we achieve those near-term value inflection points. But at the same time, we're already working on Phase 2 of the strategy, which is pipeline expansion, and that involves making complex edits like gene insertion in multiple tissues outside the liver. So essentially, going into those white spaces where we think the other technologies can't really go. So if we look at our pipeline, it looks like this. The first program that we expect to have into the clinic is PCSK9 for the treatment of familial hypercholesterolemia. And we expect to file the CTA for this program next year, so 2022. We're going to use AAV for delivery in that program, and that is being conducted in partnership with iECURE, and Cindy is going to join us in just a few slides and explain the mechanics of that agreement. Our next 2 programs in the clinic will be primary hyperoxaluria type 1 and chronic hepatitis B with INDs and CTAs expected in 2023 and 2024, respectively. And both of those programs will use LNP for delivery. After that, we have our 3 programs partnered with Eli Lilly. One program has been named, and that's Duchenne muscular dystrophy. The other 2 programs have not been named, although 1 program is in the liver, the other program is CNS. You may notice that absent from this list are TTR amyloidosis and retinitis pigmentosa despite the fact that we've recently shared some very impressive large animal data for both of those programs. We had to make some very hard decisions in prioritizing our programs. Unfortunately, we can't do everything that I want to do, and we decided that these 6 programs represent the best balance of risk and reward, and in each case, give us the opportunity to be first into the clinic. Plus we do think that there will be significant partnering opportunities for both TTR and RP based on the large volume of data that we have generated already. So now to go back to the strategy slide, I'll show you where all of these programs fit into the overall strategy. We have the 3 liver-directed programs, PCSK9, PH1 and HBV in sort of that first clinical validation phase. And then we're already working on the 3 programs with Eli Lilly that fall into that second pipeline expansion phase. So we're already working on those programs with them that move us into those white spaces where we think the other technologies can't go. So with that, I'm going to invite Cindy Atwell, who is our Senior Vice President of Business Development, to join us and talk about Precision's partnering strategy. She's going to tell us about our partnership with Eli Lilly as well as a brand-new partnership with iECURE. So Cindy?

Thank you, Derek. So you've heard a lot about ARCUS. You've heard about our pipeline. You've heard about our strategy. Now let's take a moment and talk about partnering. We partner in order to fully unlock the value of ARCUS. There are many, many different applications of ARCUS because as you've heard, ARCUS is versatile. We can't address all of these on our own. There's over 7,000 monogenetic diseases. These affect various organ types and that may require simple edits such as knockout or even more complex edits such as gene insertion and repair. Now when we think about partnering, we think about a few different elements: speed, speed to clinical validation; the economics; capabilities in terms of being able to access new tissue types and new targets; capacity and the partner's disease expertise; as well as market knowledge. Now let's walk through 2 of our strategic in vivo gene editing partnerships. We signed a deal with Eli Lilly and announced the deal last fall with Eli Lilly. This is a transformative gene editing collaboration for Precision. We -- Lilly brings deep development and commercialization expertise. Additionally, the programs we're working on allow us to apply ARCUS to multiple different tissue types. In return, Lilly receives access to ARCUS. We are working together on 3 initial collaboration programs, including DMD, a CNS target as well as a liver target. Plus Lilly retains the right to select up to 3 additional gene targets. Precision received $135 million upfront. We are eligible for up to $420 million per target plus royalties. But most importantly, we're working together in order to treat challenging diseases, and you'll hear a lot more about this a little bit later in a conversation with Derek and Dr. Ruth Gimeno from Lilly. Now hot off the presses. We announced this morning a collaboration with iECURE. This collaboration provides us with the potential for clinical validation for the in vivo platform. Additionally, it provides validation for gene insertion. In return, iECURE receives rights to develop and commercialize the PCSK9 nuclease to insert 4 -- up to 4 targets in rare disease, PKU and OTC. Additionally and very importantly, iECURE will advance the PCSK9 product through clinical proof of concept, and Precision retains the rights to cardiovascular disease, including familial hypercholesterolemia. So now why partner with Precision and why choose ARCUS? Safety, versatility, being able to access multiple tissue types and apply ARCUS in many different edits, complex, that is, to simple edits as well as intellectual property. Precision owns and controls the ARCUS platform. It is proprietary to Precision. So we're just on the start of our partnering journey, and we hope you stay tuned. Now with that, I'll turn the presentation back over to Derek.

Thank you, Cindy. All right. Let's go ahead and dive into the PCSK9 program. Familial hypercholesterolemia is a rare genetic disease that is caused by mutations that lead to an inability of hepatocytes to pull LDL cholesterol out of circulation. So what happens is cholesterol ends up staying in circulation where it can form plaques and lead to very early heart disease. PCSK9 is a very well-validated target for cholesterol control. We know that if we suppress or knock out PCSK9, what happens is it leads to a significant increase in the amount of LDL receptor expressed on the surface of hepatocytes, and that allows them to better be able to pull cholesterol out of circulation and into the liver. We have been working for several years with Jim Wilson's group at Penn to demonstrate long-term editing of the PCSK9 gene in nonhuman primates. It started with a first-generation ARCUS against PCSK9 that we call M1PCSK9. And we did a dose response in that case in nonhuman primates, 3 different doses of AAV to deliver the nuclease. And what we saw was a very nice dose response of reduction in PCSK9 that reached a maximum of 82% reduction in the case of the high dose, and those reductions in PCSK9 were stable out to several years. So then we turned our attention to a second-generation ARCUS called M2PCSK9 that recognizes the same target sequence. And we focused our efforts there on the middle dose, 6e12 vg per kg. And in that case, we saw an average reduction of 60% in circulating PCSK9. That second-generation nuclease, M2PCSK9, that is our clinical candidate. With the reductions in PCSK9 came stable reductions in LDL cholesterol. Again, in the case of the first-generation nuclease, we see these nice dose-responsive reductions in LDL cholesterol, reaching a maximum of 62% reduction in the case of the high dose. In the case of the clinical candidate nuclease, we see an average of 42% reduction in LDL cholesterol, stable across multiple years. So we've decided for this program, we're going to use AAV for delivery despite the fact that our target is the liver. And the reason for that is this. The disease that we're interested in treating, familial hypercholesterolemia, is caused by an inability of cells in the liver to take up lipids out of circulation. And the mechanism that is deficient in the disease is the exact same mechanism that the liver uses to take up a lipid nanoparticle. So we reasoned that the disease itself would significantly impair the ability of a lipid nanoparticle to be delivered to hepatocytes in the patient. And importantly, there could be significant patient-to-patient variability in terms of delivery, with patients that have the more severe form of the disease less able to take up the lipid nanoparticle and therefore requiring a higher dose than patients that have a less severe form of the disease. And that can significantly complicate our ability to determine a dose for Phase II. AAV works completely independently of lipid uptake. So we decided for this program in particular, we wanted to use AAV for delivery. The research vector that we used to generate the data that I showed you on the previous slides is shown on the upper right. The anticipated clinical vector is shown on the lower right of the slide. There are a couple of significant differences between those 2 vectors. The first is the research vector was an AAV8. The anticipated clinical vector is AAVrh.79. Two reasons for that: Number one, we think that the rh79 capsid will have a lower frequency of neutralizing antibodies. And then the second reason is intellectual property. We think that this has blockbuster potential, and so we would rather not step on someone else's IP. The other change that we made to the vector is we changed the promoter and we removed the expression enhancer. What this does is it reduces the total amount of ARCUS enzyme that's getting made in each hepatocyte. And what we found is that sort of gets us to this right threshold level of expression where we cut the on-target site efficiently but have significantly reduced off-target gene editing. And again, we expect that this program, we will be filing the CTA next year. Moving on to primary hyperoxaluria type 1. This is a rare disease that is caused by defects in the glycine biosynthesis pathway. Specifically, mutations in a gene called AGXT block the pathway one step too early. And that leads to the accumulation of a metabolite called oxalate, which then forms crystals of calcium oxalate in the kidneys. That leads to very severe kidney stones and ultimately, end-stage renal disease, so a disease with a very high unmet need. Our therapeutic strategy is to target a gene called HAO1 and knock it out. HAO1 acts one step upstream in the pathway. So rather than accumulating the toxic metabolite, oxalate, we accumulate a nontoxic metabolite called glycolate, which is soluble and easily excreted. HAO1 is the same gene that is targeted by Lumasiran, which is Alnylam's approved siRNA therapy for PH1. So it is a very well-validated target, and Alnylam has done a very nice job of blazing a trail for us to follow into and hopefully through the clinic. This program, we will use lipid nanoparticle for delivery. I'm going to show you some proof-of-concept data from nonhuman primates in which we used a first-generation nuclease called M1HA01 and we deliver that using AAV. The actual clinical formulation is going to be a sixth-generation ARCUS enzyme called M6HAO1, and it will be delivered via lipid nanoparticle. Proof-of-concept data in NHPs looks like this. We showed that we can use the first-generation ARCUS enzyme to very efficiently knock out the HAO1 gene in nonhuman primates. In fact, we see about a 98% reduction in HAO1 mRNA. That then leads to a similar 97.9% reduction in glycolate oxidase protein. Glycolate oxidase is the protein that is encoded by the HAO1 gene. So we knock out the gene. That then gives us a reduction in mRNA. Reducing the mRNA gives us a reduction in protein. As we knock out the HAO1 gene, we do see the accumulation of glycolate, which is that nontoxic metabolite, in the serum of the animals. And in fact, we can benchmark our studies in primates against published data from Alnylam using the approved siRNA in primates. And what we see is, in fact, a superior metabolic profile following a single treatment of ARCUS on the left as compared to multiple administrations of a high dose of the siRNA on the right. This is great. This tells us that we have a very good therapeutic index and actually have the ability to go down in dose if we want to while still having an efficacious therapy. So that was data with the first-generation ARCUS enzyme. This is data comparing the first-generation research enzyme to the sixth-generation clinical candidate. Looking first at potency, what we did is we titrate the nuclease mRNA in cells and we look for editing at the on-target site in HAO1. What you can see is the 2 nucleases are very similar in terms of potency. In both cases, we see saturating amounts of on-target editing somewhere around 100 nanograms of mRNA. Looking at off-target editing with the candidate nuclease. At that dose of 100 nanograms of mRNA, we really don't see much. So we sequenced the top 32 sites identified using the oligo capture method, and we're only able to detect off-target editing at one site and it's at a very low frequency, 0.1%, which is very close to background. And that particular off-target site is in a region of a genome that we're really not concerned about. We do expect -- thank you -- we do expect, again, that this program will reach IND in 2023. Turning now to chronic hepatitis B. The real challenge with hepatitis B that has made it very difficult to cure is that the HBV virus persists in chronically infected hepatocytes as an extrachromosomal genome called cccDNA. And unless we are able to eliminate cccDNA, there's always the potential that the virus can reactivate. So the strategy that we're using here is using a lipid nanoparticle to deliver an ARCUS enzyme that recognizes a conserved site in the HBV genome. That nuclease then cuts its target site in cccDNA or in cases where cccDNA was able to integrate into the genome. It inactivates the virus, leading to durable antigen loss and, hopefully, a functional cure. We generated some data in primary human hepatocytes in collaboration with Gilead a few years ago, and we've shared this previously, basically showing that we can introduce ARCUS on either a lentivirus or mRNA into infected primary hepatocytes. We see significant reductions in cccDNA. We see the introduction of a significant frequency of indel mutations that inactivate the virus in the remaining copies. And then that combination, reduction in total cccDNA and the introduction of inactivating mutations, the combination of the 2 then results in significant reductions in secretion of HBV S-antigen. So this is a very nice data set. But then we sort of ran into a science brick wall because there is no good model of human HBV infection. The standard models in HBV are woodchuck and duck. But woodchuck has to be infected with woodchuck HBV and ducks have to be infected with duck HBV. And those viruses are actually quite a bit different from the human virus, and the target site for our nuclease is not conserved in the other viruses. So the project was sort of stuck because we didn't have an animal model that we could actually do dose-ranging studies on. I'm very pleased to say that we very recently figured out a way to overcome this challenge and have developed an entirely new model for human HBV infection. And to do this, we borrowed a page from our collaborators at Penn. We recognized that HBV actually has a lot of similarities to AAV. In particular, they both infect hepatocytes and they're both able to establish latency as extrachromosomal circular DNA elements. So what we did is we cloned the open reading frames from hepatitis B into an AAV vector. We can then introduce that AAV vector into an animal, and that can be either a mouse or a nonhuman primate. The vector goes to the nucleus and hangs out there as a circular piece of DNA. That circular genome produces S-antigen mRNA, which then secretes S-antigen protein into circulation in the animal. We can then come in with our ARCUS lipid nanoparticle, which cuts the virus DNA, leading to a loss in production of S-antigen mRNA and reductions in the amount of S-antigen protein that gets secreted into circulation. So we actually have 3 different biomarkers that we can look for in this model. We can look for reduction in S-antigen in circulation, we can look for a reduction in total viral DNA, and we can look for the production of inactivating indel mutations in the virus open reading frames. Here's what that looks like, first in a mouse. When we run this model in an immunodeficient mouse, we see that animals that are treated with the ARCUS LNP have a significant reduction in total viral DNA, and the introduction of a very significant number of inactivating indel mutations into the virus. Those 2 together, reduction in virus and inactivating mutations, then give us very significant reductions in secreted S-antigen. So we see about a 95% reduction in S-antigen production in the animals treated with ARCUS, so a very, very nice animal model, a very good demonstration of LNP delivery of ARCUS. We can also run the model in a nonhuman primate. In this case, we see even more significant reductions in total AAV copy number, so total viral DNA and the introduction of those inactivating mutations into the genome. In this case, the nonhuman primate is immunocompetent. So the animal's immune system actually neutralizes the S-antigen before we can detect it. So S-antigen isn't going to be a useful biomarker in primates, but we really don't need it because between AAV copy number and indels, we have very good biomarkers that we can use to do dose-ranging studies in primates. So we feel like we've overcome the major challenge in front of us to generating the IND-enabling data that we're going to need for this program. It is full steam ahead on HBV. We expect to file the CTA for this program in 2024. All right. The last program I want to talk about but the one that is probably my favorite because it is a really cool demonstration of the power of ARCUS and just some awesome molecular biology, that's our DMD program. What we're doing here is we are using a single AAV vector to deliver 2 ARCUS nucleases at the same time. Those 2 nucleases recognize different sites in the dystrophin gene. And if those 2 nucleases cut, they actually chop out a 0.5 million base pair fragment of the dystrophin gene that encodes exons 45 through 55. We picked that particular reason to delete because it's a gene mutation hotspot in dystrophin that is responsible for more than 50% of DMD cases. So if we can just delete the entire hotspot entirely, we potentially have a therapy that's applicable to more than half of DMD patients. If we do this, if we delete this fragment, it creates a new in-frame fusion between exons 44 and 56 of dystrophin, which then results in the expression of a slightly truncated form of the dystrophin protein. This particular mutation actually exists already in the human population. So there are people walking around who don't have exons 45 through 55 of the dystrophin gene. They have a very, very mild phenotype, so almost asymptomatic. So what we can potentially do with this is have a therapy that is able to convert a very severe form of the disease to a very, very mild form of the disease in more than 50% of patients. We've shown that this works in patient cells. So if we introduce RNA, encoding those 2 ARCUS enzymes into myoblast from a DMD patient, we see that, in fact, in about 30% of those cells that 0.5 million base pair chunk of the gene gets deleted. If we correct the DNA, that results to -- in corrected mRNA. And if we correct the mRNA, we now see the expression of dystrophin in cells that previously were unable to express any. So very exciting, I'm very happy with the progress that we've made so far with this program as well as our other Eli Lilly partnered programs. What I wanted to do at this point was provide Eli Lilly's perspectives on the collaboration. Ruth Gimeno, who is the Lilly Vice President responsible for our collaboration, is taking a much deserved vacation right now, but I was able to sit down with her last week and asked her a few questions that I imagine are probably on your mind. And we're going to share a short video of that right now. [Presentation]

Thank you so much for joining us. I know how busy you are.

Hello, Derek. Good to be here.

So if you don't mind, I thought what I would do is ask you some of the questions that I get asked a lot so that they can hear the answers directly from you.

Absolutely.

All right. First question, probably the one that I get the most, I talk a lot about why Precision wanted to work with Eli Lilly and how this collaboration has been so great for us. What a lot of people ask me is why did Lilly choose Precision. Why were you interested in working with us?

Yes, Derek, there were really 3 things that excited us about the partnership with Precision. First, the science, and that always comes first. We were very impressed by Precision's gene editing platform, the progress that you've made using that platform ex vivo and then the preclinical data in vivo with PCSK9. We thought this was an exciting opportunity to work with you, take the next step, bring this platform in vivo into the clinic and do some good things for patients. That brings me to the second point. We saw a big upside for patients. For example, we had in the past tried to address Duchenne's muscular dystrophy indirectly with a small molecule in a large Phase III trial, and we were very disappointed when this failed. We know that our chance for success is so much higher when we address the root cause, underlying disease, and gene editing certainly is the way to do that. So as we learn more about the Precision platform, we saw an opportunity to go back and maybe be successful in this space, so really harness this technology. And finally, we also felt that Precision was the right partner. As we went through the process of getting to know each other, there was a very open and science-based dialogue between the 2 teams. I think a lot of trust did develop between our 2 organizations. And this is just a really important ingredient for a successful collaboration. So we saw this in Precision, and I think everything we've seen over the past year has borne out that this has really been a great collaboration so far.

Despite significant medical advances, there really are a large number of genetic diseases for which no suitable treatments exist. So we'd love to hear your thoughts on how you think ARCUS might contribute to a next generation of medicines.

So the goal of gene editing is to correct disease at once and be done with it. So we'd expect much better durability with the gene editing therapy. And then finally, as we think about gene editing, we keep in mind the technical challenges and we need to address them. One of them is to get the genes into the -- edited genes in the right tissue and enough cells to be therapeutically meaningful. We also need to have a high degree of specificity to ensure safety. Safety is always high on our minds. One of the attraction of Precision's ARCUS platform was that it does have the potential for a unique safety profile that we found very attractive.

Yes. You used the phrase, once and be done with it, earlier to describe gene editing and I think this might be a perfect example of where that's going to be really important. So looking beyond DMD, we're also working with Lilly on 2 other indications, and we disclosed today that one of those indications is in CNS and the other is in liver. Would you mind commenting a little bit on why you picked the indications you did?

Yes. A number of criteria when we selected projects for our collaboration was actually quite a systematic process. And first and most important of all, we wanted to make sure we worked on indications where there was a potential for a large benefit for patients. And interestingly, some of the indications we prioritized actually were ones that Precision that also prioritized. We want to make sure we address the significant unmet medical need. We also wanted to find indications that were uniquely suited for gene editing and ideally ones where the ARCUS technology provides an advantage. So this is one of the reasons we selected Duchenne's. Now at the same time, we wanted to pressure test the Precision technology. The programs we decided to work on were not necessarily the easiest, but they allow us to evaluate the ARCUS platform for different types of editing in a number of different tissues we are very interested in. As you know, we have a strong interest in CNS and neurodegeneration. We've worked in muscle for quite a long time, and liver is also a tissue we are very familiar with and very interested in. So we see potential for additional applications. And if our initial 3 projects are successful, it would actually set us up well to potentially expand our collaboration.

So one last and related question. Can you comment on how Lilly is thinking about the future of gene editing and where you can see this going in sort of the middle and long term?

Yes. I think in the near future, we see gene editing as a way to address genetically defined diseases that are devastating to patients. And as you pointed out, there are quite a few of those that really are not being addressed right now. As we gather more information on safety and we learn how best to use gene editing, I think we'll start to extend the indications that are suitable for gene editing. We go into a broader range of diseases. Now ultimately, gene editing could become a standard tool in our drug discovery toolbox that we can bring to bear to help patients wherever it makes sense from a scientific perspective. You could envision, for example, in cardiovascular disease. Instead of having to inject the PCSK9-lowering agent on a regular basis, one could delete PCSK9 via gene editing and allow patients a lifetime of low cholesterol levels without having to worry about chronic medications. So simplicity, adherence to therapy, durable benefit, those are some of the unique features that gene editing could bring to the table if we can learn how to do it safely. So you could envision a future where gene editing is just another modality we are using. We're not there yet, but certainly, we are hoping to be able to get there.

I'm definitely looking forward to the day that gene editing is as routine as antibody therapies are.

That would be a great vision.

Thank you so much to Ruth and the rest of the Eli Lilly team for the ongoing support and a terrific collaboration so far. With the last few minutes, I would like to talk about a few things that are sort of on the horizon. And the one that I want to start with is mitochondrial gene editing because this is just super cool. So the nucleus isn't the only place in the cell where there's DNA. Mitochondria, which are responsible for producing the cell's energy through aerobic respiration, also have their own genomes. And much like the nuclear genome, mitochondrial genomes can have mutations in them that lead to disease. And in fact, a large number of genetic diseases are associated with mutations in the mitochondrial genome. So unlike the nuclear DNA, the mitochondrial genome actually has a bunch of copies -- or the mitochondria matrix has a bunch of copies of its genome. And in the case of mitochondrial disease, what you typically see is a mixture of mutant mitochondria genomes and wild-type mitochondrial genomes. And it's the ratio of mutant to wild type that determines the severity of the disease and which tissues the disease manifests itself in. So we've been working with Carlos Moraes at the University of Miami to demonstrate that we can use ARCUS to selectively eliminate mutant mitochondrial genomes. It works like this. So we deliver an ARCUS enzyme to the cell that is engineered to recognize a unique sequence in mutant mitochondrial genomes. And we fuse that ARCUS to a mitochondrial targeting sequence. So then the ARCUS protein gets made in the cell. And because it has the targeting sequence, it gets imported into the mitochondrial matrix. Importantly, mitochondria can import proteins. They cannot import nucleic acids. Meaning the guide RNA for a CRISPR cannot be imported into mitochondria. This is something that is really best done with ARCUS. Once ARCUS gets into the mitochondrial matrix, it binds to the mutant genome sequence, cuts it and degrades it. Mitochondria, unlike the nucleus, don't have the ability to repair DNA breaks. So if we cut the circular mitochondrial genome, it just gets degraded and lost. And then what happens is as that mitochondria continues to grow and divide, it ends up preferentially replicating the wild-type genomes. And over time, the wild-type genomes predominate and we lose all of the mutant genome copies. We just published these data for the first time a couple of months ago. Here, we're showing that, in fact, if we fuse an ARCUS enzyme to a mitochondrial targeting sequence, it does localize to the mitochondria. You can see the ARCUS stained in green and Mitotracker stained in red. You can see that they basically overlap, showing that ARCUS is, in fact, being imported into the mitochondria. So we tested this in a cell model of mitochondrial disease. And the way this works is we have a cell line that we made that has a mixture of mutant and wild-type mitochondrial genomes. And in fact, on day 0, when we started the experiment, about 95% of the genomes were mutant. That's the light blue trace in the figure. Only about 5% of the genomes were wild type, which is the pink trace. We then delivered ARCUS mRNA to the cell. And you can see in 24 hours, we almost completely eliminated all of the mutant mitochondrial genomes. And then over the next few days, you can see the number of wild-type genomes goes up as the wild-type genomes get replicated. And after about a week, we have a mitochondria that is now basically wild type. With this shift in mitochondrial genomes, we also see a restoration of mitochondrial function measured either as basal respiration, maximal respiration or total mitochondrial ATP production. So we, by selectively eliminating the mutant mitochondrial genomes, were able to restore function to the mitochondria in the cells. This also works in mice. Our collaborators at Miami took an ARCUS enzyme that we designed for a mutation in mouse mitochondria. And they introduced that to a mouse model of mitochondrial disease using an AAV9 vector. And what we can see is in all of the tissues that AAV9 gets too well, so muscle, kidney, liver, we see very significant reductions in the amount of mutant mitochondrial DNA in those tissues. So very, very excited about the prospects for this, looking forward to getting this into the clinic. All right. To wrap things up, we have a very special guest star. Jim Wilson is hopefully going to be joining us from the University of Pennsylvania, if the technology collaborates, to walk us through some brand-new data demonstrating targeted gene insertion into the PCSK9 locus in nonhuman primates. So do we have Jim on the line? There he is.

Great. All right. Can you see me, Derek?

Yes. Jim, how are you?

I'm doing fine. Thank you. And Derek, are you going to control the deck?

Yes. I am the man in charge of the clicker. So you just tell me when you want to go to the next slide.

Yes. I need to see the first slide here.

Do we have the first slide for Jim?

Perfect. Okay. Great. Well, Derek, thanks for inviting me to your R&D Day. Derek and I have talked and worked on gene therapy for many, many years, in vivo gene therapy. And the goal here is to try to leverage our experience here at Penn and the gene therapy program that we learned from in vivo gene therapy to in vivo genome editing. And as was brought up earlier, we were able to forge a collaboration between Precision and a company that we just launched called iECURE whose goal is to use genome editing to treat liver diseases. And so we heard some nice applications of ARCUS for in vivo editing, where what was needed was targeting a break and then some repair that led to splicing out about -- out of a sequence or introducing an indel to knock something down. But the holy grail from my perspective is not only to create the break but to drive integration or modification of a genome to treat loss of function diseases. So that's what our initial programs are at iECURE in collaboration with Precision. And so the question is, can we actually achieve a gene replacement -- stable gene replacement in a primate liver with the goal of standing this up for our lead programs, ornithine transcarbamylase deficiency and then subsequently PKU. So Derek asked me to share some of the early data that we generated in an attempt to achieve targeted insertion of a gene into a primate liver. I'll say at the beginning that attempts to do this in any species in which the recipient is an adult have not been very successful because the targeted integration requires that the cell is dividing. So we focused on newborn and infant nonhuman primates, which turns out to be the target population for many of these lethal liver metabolic diseases. So on this slide just summarizes our approach for the use of ARCUS, where we have a 2-vector strategy. One vector expresses ARCUS, and this is in -- on a newly discovered AAV vector that came out of our lab. That would create the break much like what Derek had described earlier. And then we co-inject a donor that would then drive integration into the site that we created. And rather than try to correct the specific mutation which is the way in which many programs are moving forward, we wanted to develop and approach that with mutation-agnostic. And so that the goal here is then to integrate into a break a fully functional gene that would reconstitute function. So in order to best illustrate this, we've used, human factor IX. And we've swapped that out and we can set anything in and out. We have experiments underway using the OTC gene, but factor IX allows us to measure the expression of the integrated gene over time. So again, dual injection in either newborn monkeys or we've done this in infant monkeys that were 3 months old and then follow them for Factor IX expression. If we go to the next slide, Derek. Important controls here, which you will see in subsequent data, are introducing the donor without a nuclease. So that would be GFP to determine whether there's any ability to integrate this genome without a site-specific targeting or editing approach, and you'll see those data. And then in addition to that, we have done a lot of work attempting to leverage the new technology of CRISPR/Cas in the same application for which we've had some success in mice. But we also attempted the same experiment using the Staph aureus Cas9, which is one of the endonucleases that can actually fit into an AAV. And in this situation, then the guide would be on the donor DNA. So again, the experiment is ARCUS and donor. The controls are donor without a nuclease. And then we attempted to try to achieve some level of success with Staph aureus Cas9. Next slide, please. So these are the data that we generated in these experiments. The green is the experiment without a nuclease, but the donor -- the blue is an experiment with the Staph aureus Cas9. And the orange are the data in an animal with ARCUS and the donor. And in the upper left, we followed the level of factor IX over time, and this is out to about a year. These are newborn monkeys that were -- which were treated with the 2-vector system, actually achieved levels that were significantly above normal. Again, this isn't meant to be interpreted that we're going to propose to treat hemophilia in this way, but you could see that we could follow expression of the transgene over time. The negative control, which is in green, gave us the expected result of factor IX expression, which is we had a little bit at the beginning because it's an AAV vector. But as the liver began to proliferate that we lost expression to the point of it being 0. And then blue was the results with Staph aureus Cas9, where we did achieve editing above baseline, but this is a log Y axis, and it was several logs lower than what we saw with ARCUS. And in the middle panels, we've tried to follow the PCSK9 because we decided to leverage the experiment -- experience we had with PCSK9 as a site into which we would drive this new gene. And you can see at baseline, newborn monkeys have a wide range of PCSK9. So it was hard to determine really whether anything was significantly decreased. But the goal was not to treat the animals. The therapeutic end point wasn't to decrease PCSK9. That was used as a target to drive the integration of our gene. And then over on the right, the 2 panels, the ALT and AST, which are really important, is if you look at our program, our product, which is ARCUS with donor, which is orange, that there was really no very mild transient increase in transaminases that were lower than when we used GFP as the control. And for reasons we don't understand in this animal, the Staph aureus led to a pretty significant increase in transaminases. So from the perspective of liver toxicity, we were quite pleased with this result. So these are the data with newborns. We go to the next slide. The question really is how old can a recipient be for the recipient to be receptive to targeted integration? And you can do studies in mice, but it's hard to extrapolate primate. So we're starting to further evaluate that window. In other words, can we wait a while before we dose the recipient with a vector and still get targeted integration? So here's an experiment where we dosed an animal that was 3 months of age, a primate, with either the Staph aureus construct in blue or the factor IX ARCUS construct in orange -- or brown. Again, upper left, we're evaluating the factor IX expression. Again, it went above normal and then reached a steady state. That was -- would be therapeutic and just a little bit below normal. In this case, the PCSK9 levels were a little more stable. But as we saw in newborns, the Staph aureus construct really didn't work very well, that there was no detectable factor IX after a while. And interestingly, in the newborns, there was a slight reduction in PCSK9 with the ARCUS nuclease. But over on the right, we followed the weights of the animals, which is a proxy for health, and the ARCUS-treated animal continued to gain weight appropriately. And again, upper right, really very little, if any, evidence for liver toxicity. So, so far, really encouraged by at least the immune toxicity when you treat newborn animals or infant animals. So the question really is what is the developmental window where we could still achieve efficient targeted insertion? And we're exploring that with an expanded number of nonhuman primate experiments. Next slide. As I showed at the very beginning, we do biopsy these animals at day 84 and then at 1 year, day 356, and then conduct in situ hybridization to determine how many cells actually have been edited and are expressing the transgene. And if you look at the far left, we have day 84 with -- in the newborn animals with ARCUS, which shows many -- each spot there is a positive cell. When we use ARCUS with Staph aureus, it's much lower. And control, there's a random cell. So this is consistent with the expression based on serum levels of the secreted protein. And then we were delighted we've been able to biopsy at least a newborn animal at 1 year, and it looks like there's no reduction in the level of a number of transduced cells. Then when we go over on the far right, the infant that received ARCUS, really an impressive number of positive cells at day 84, consistent with the expression data that Staph aureus constructed and worked. If you go to the next slide, it just quantifies this where we've done quantitative morphometric analysis in -- for the different components of the system, in particular if we look at purple for factor IX, that the newborn animals treated with the ARCUS nuclease was 12.1%, 11.6% over a year, so really no significant decline. And we believe 10% would be really helpful and therapeutic in patients with these severe forms of metabolic diseases. But the day 84 sample from the infant was almost 20%. So I must say that I did not expect it to work this well. Being a scientist, you always manage your expectations. So maybe that's just me. But here, we have evidence in vivo in a primate for targeted integration of a functional gene in 20% of the hepatocytes. So with that, we're delighted to move forward with the collaboration with Precision. And just to summarize in the last slide. The conclusions are that the transgenes were targeted efficiently in the PCSK9 locus using ARCUS in both newborn and infants; that the gene addition appears to be stable at least for a year, again requires more of a follow-up. But the reason that we're excited about moving this forward and the new venture iECURE is it may represent a universal approach for treating rare diseases, genetic diseases caused by loss of function. If you can treat infants or newborns, it really is plug-and-play. We continue to use the same ARCUS nuclease, building on the safety profile but then plug-and-play a different donor for different diseases. So with that, Derek, I think my time has come. And again, thanks for inviting me to your R&D Day.

Thank you so much for joining us, Jim. That was awesome. So that's going to conclude our event. And Jim, hopefully, you can stay on the line for Q&A.

Okay.

Hopefully, we have convinced you that ARCUS is the best-in-class technology for in vivo gene editing. We're looking forward to proving that with 3 INDs in the next 3 years and then really pushing the boundaries of gene editing and going places that we don't think the other technologies can go. So with that, we will open the line to questions.

[Operator Instructions] Your first question is from Maury Raycroft with Jefferies.

Maybe just tying the background info on ARCUS together with the NHP PCSK9 data you just showed, if you can talk a little bit more about what you're seeing on the HDR to NHEJ ratio in the NHP studies and if age influences that ratio to change.

Yes. So the -- you saw the vector had homology arms, and we're doing the molecular analysis to understand how exactly the vector went into the genome and the extent to which that was driven by HDR. The assumption is that at least to a large extent, integration was mediated via the homology arms driving the HDR process.

Got it. And then for the -- so you mentioned Alnylam's Lumasiran as a benchmark or reference for PH1. Would you view Alnylam's developed inclisiran as the right reference for PCSK9? And for timing for the trial starts, will PCSK9 be first because you're furthest along? Or is this more about prioritizing PCSK9 as a value driver or proof of concept?

Yes. So we would reasonably expect to benchmark against patisiran. That siRNA gives about a 40% reduction in LDL cholesterol, which is similar to what we were seeing in nonhuman primates treated with the clinical candidate nuclease at that sort of intermediate dose. So that is approximately where we think the threshold might be. As for prioritization, I would say the PCSK9 program is both very, very interesting and ahead of the others. So both of those contribute to kind of where it is in the queue.

Your next question is from Ben Burnett with Stifel.

So I guess regarding the PH1 program. As I understand it, this patient population includes like a spectrum of ages, both kids and adults. So I guess I'm talking of any regulatory differences. Are there any technical or like scientific challenges that are unique to developing this in children and adolescents?

Yes. So we don't so. Obviously, we are going to do our preclinical studies in animals representing a range of different ages. And we would expect to start the trial in adults and then gradually move younger over time. But it's a good question and definitely something that we're going to have to pay attention to.

Excellent. Okay. And then just a follow-up on the previous line of questions around PCSK9. When -- after iECURE runs this Phase I proof-of-concept study and you guys are looking at it, is there a threshold of PCSK9 reduction that you need to see in the clinic in order to continue to be excited to move this forward? Is that basically inclisiran-like efficacy?

Potentially, yes. So that is sort of the target that we've set as we would like to meet -- match or beat the siRNA therapy. An argument could be made that even if the efficacy is a little bit reduced relative to the siRNA, we are talking about a onetime treatment as opposed to repeated administration of the siRNA. So we'll cross that bridge when we come to it. But at least right now, the thinking is we want to at least match what's on market.

Your next question comes from Raju Prasad with William Blair.

Maybe if we can talk a little bit about the AAV platform. Can you maybe discuss a little bit about the integrated DNA and some of the surrogate markers that you might use to kind of assess cccDNA and integrated DNA reduction? I think it's been a discussion point in the space as of recent.

Yes. Definitely. So it is very important that we targeted -- with our nuclease, we picked a site in the genome that is conserved in both cccDNA and integrated HBV because we recognize that we need to inactivate both of those because they're both potentially sources of viral antigens. The most obvious biomarker that we're going to use in the clinic is S-antigen, and the expectation is that S-antigen will give us a good read of total intact viral load.

Great. And if Dr. Wilson is still on the line, I would love to get your thoughts on the use of the AAV delivery mechanism for PCSK9 in light of the adcom last week and his thoughts kind of on utilizing kind of that mechanism -- or that delivery mechanism and dosing potential parameters with PCSK9.

Sure. Can you guys hear me?

Yes.

Great. Yes. It's a question we had at the beginning, and we continue to ask that question. But the one thing that I learned about AAV when I began to work with Derek and was sort of a eureka moment is what AAV can do for liver at least in primate is it's pretty efficient at modest doses. The problem is, is that its expression is extinguished pretty quickly over a couple of months. So it's only a few percent of the cells. So it does give -- it's a very good way to get transient expression. That's not the way that I started the field and hope that it would continue because I had hoped that the efficiency would be durable. So with really modest doses of vector in primates, we can achieve the majority, if not all, of the hepatocytes expressed in the ARCUS and then it's turned off. And then if there's any kind of foreign nature to the transgene, it's turned off completely. And we showed that in our publications. So AAV in this application actually may have a role because it's a very efficient way to get transient expression. Now in terms of the off-target effects, that's a function of anything that's going to edit the genome. You'll have non -- NHEJ and indels. But the other area that we felt comfortable in moving forward with AAV in this application is we have an enormous amount of clinical data in the use of these same types of vectors for gene therapy from an acute toxicity standpoint. So we still think that there is -- there would be a role in this application.

Your next question is from Gena Wang with Barclays.

This is Tom for Gena Wang. My first question is for Jim. I just wanted to get your thought on the long-term safety using AAV to deliver gene editing tool. I think those gene editing tool, in theory, should be as episomal inside the nucleus for long -- as long as the cell lives. So for a company, what is your longest safety data from mouse model using AAV to deliver ARCUS?

Right. Yes. It is a good question. But as I commented on the previous speaker that for primate liver, it's not stable. And in fact, if you express anything that's foreign, you lose almost entirely logs of expression over literally a couple of months. So in this application, AAV is not stable. And therefore, the concern about ongoing expression of a nuclease is not something that we worry about. The problem with mice, they only live about 1.5 years or 2, but we have monkeys that are several years out, and we continue to follow them -- that received the ARCUS nuclease, and all those animals will continue to be followed.

Okay. Got you. And my next question is for the PCSK9 program. If I understand correctly, you have decided to use the AAV vector as the delivery vehicle, right, over the LNP. So the LNP is not the option for the IND?

That's correct. So PCSK9 will be AAV.

Okay. Great. Just for clarification. And for DMD program, is there any concerns on 2 cuts since it will like generate many more unwanted editing, like -- such as like AAV self-integration? Can you comment on that?

So at least so far, the major edit that we see in cells at least is exactly the edit that we want to make. So we make 2 cuts and then we delete the intervening fragment. We do, with some frequencies, see some other things happening. So for example, occasionally, what will happen is we cut out the 0.5 million base pair piece of the genome. It flips around backwards and goes right back in. That happens every now and then. So we do see some editing effects other than the one that we're necessarily trying to create, but predominantly, we see the deletion that we're trying to make.

Your next question is from Eric Joseph with JPMorgan.

Great. Maybe just picking up on an earlier question with respect to off-target editing. I guess is there -- at this point, have you been able to look at editing fidelity using [ guide seeker ] oligo capture? Is there any -- is there a general comment you can make about the propensity for off-target editing following -- sorry, for gene insertion application versus gene deletion or gene silencing?

Yes. I don't know that we have any data necessarily, although I wouldn't expect given that the off-target editing is a function of the nuclease that we're delivering and whether we're knocking a gene out or knocking a gene in, we're using the same nuclease and sort of the same dose of vector. The expectation would be that the off-target editing profile is the same between the 2 applications but something that we do need to look at.

Okay. Got it. I'm also curious to know whether -- for the gene insertion applications, whether LNP media delivery would be something that's feasible or what you need? Really AAV [ or adeno nuclease ] -- for the AAV [ adeno nuclease ] will kind of be expressed for a longer period to facilitate efficient editing insertion.

Yes. That's a great idea, Eric. Yes. You're exactly right. In the case of Jim's data, he showed using 2 AAV vectors, one delivering a nuclease and one delivering a transgene. This potentially could also be done using a lipid nanoparticle to deliver the nuclease and AAV to deliver the transgene.

Okay. Well, it sounds like maybe more work in progress there behind the scenes. Maybe just a final question here strategically. I'm curious, Derek, to kind of get a sense of how you arrived with iECURE as the appropriate partner for the PCSK9 program. I know Dr. Wilson has been key to the program developing from its inception. But I'm just curious to know whether there were other strategic alternatives that were explored, whether there was an upside, competing interest potentially from larger pharma akin to the partnership that you struck in DMD with Lilly.

I look for any opportunity to work with Jim, and this was a great example. You're exactly right. So Jim and his group have been involved in really running the PCSK9 program from the very beginning. So it made a lot of sense for us to allow that to continue to move forward. And at the same time, what the partnership with iECURE allows us to do is not only do we get access to Jim and all of his resources to push the PCSK9 program forward. But also, what iECURE is doing is targeted gene insertion, which, as I've sort of said repeatedly, is something we're very, very interested in doing. So we sort of we win both ways and iECURE wins both ways. So I think really, really terrific. A little bit of an unusual collaboration but definitely one where I think both sides win, and we're really looking forward to moving forward with it.

So maybe, Cindy, I don't know if you want to, maybe talk about where -- what you're interested seeing from partners and what things they're interested in?

Sure. Absolutely. So there's a broader variety of interest across different tissue types, different types of edits in terms of gene insertion, knockout gene repair. So we're seeing quite a bit of interest in different areas.

Your next question is from Gena Wang with Barclays.

This is Tom again for Gena Wang. I just have a follow-up on the safety data. [ Perhaps ] what is your longest safety data from the mouse model using AAV to deliver ARCUS?

From a mouse, as Jim mentioned, mice don't live that long. So probably longest safety data we have in-house is 1 year, 1.5 years. Probably the more meaningful data is the nonhuman primate data that Jim referred to earlier. We actually published a paper with Jim's group showing 3 years of follow-up on the nonhuman primates treated with the PCSK9 nuclease. And as Jim said, those animals are still being followed. So we're now out past 4 years.

Your final question is from Soumit Roy with JonesTrading.

In light of your peer, bluebird, showing some hematological malignancy as safety issue, is FDA asking for any specific kind of assays to be monitored or safety database to be looked into?

Well, so we're using AAV for the PCSK9 program, which is a non-integrating virus. So we wouldn't necessarily expect to see the same sorts of things that bluebird had. So in terms of what we think regulators are expecting is sort of the interface between safety studies for an AAV gene therapy and safety studies for a gene editing therapy. So we have to look at things like -- on the one hand, think about complement activation and AAV integration. On the other hand, we have to think about nuclease expression and off-target editing.

Derek, I think that's our last question. I don't know if you have any maybe closing thoughts before we end the show.

No. I'd just like to thank everyone for tuning in early to watch the show. And I hope it's obvious that we're all very, very excited about the future of gene editing at Precision BioSciences.

Ladies and gentlemen, that concludes today's conference call. Thank you for participating. You may now disconnect.