Sean Menarguez: Hello everyone and happy Friday here at ISCT. I really appreciate everyone's time today with us. The presentation is Key Considerations in Early Cell Therapy Development to Mitigate Risk and Drive Faster Clinical Success. I’m joined here with Dr. Matthew Porteus, professor of pediatrics from Stanford, and also co-founder for Kamau Therapeutics. We also have Jason Skowronski here, who's the director of CMC [clinical process development] at Kamau Therapeutics as well.

We'll start off the presentation about the great work from Kamau, and it'll dive into a Q&A fireside chat. Any questions from the audience we want to be a constructive dialogue, so please be free to raise your hand or stand by the mic and we'll have that conversation accordingly.

Matthew Porteus: Thank you, Sean, for the introduction and for inviting us to participate in this fireside chat. Thank you all for joining us. So the way we'll do this is I'll give about 10 minutes of how we got to where we got to, and then Jason will take over and discuss the second-generation manufacturing protocol that we've adopted.

I want to start with this slide and point out the importance of monogenic diseases in the world. There's estimated to be at least 6,000, maybe as many as 10,000 such diseases in which people are born with mutations in a single gene that leads to a specific disease. And these diseases span all of medicine, ranging from hematology to pulmonary to immunology, just some of the examples. And while all of these diseases are classified as rare in the United States, in some, they affect about 10% of the population, both in the United States and worldwide. The other thing I like to remind people of is that any individual patient with a serious genetic disease like this affects the community. There's an echo with that disease, not just to the family members or parents but the broader community can be negatively impacted by these diseases. They also tend to impact children. And so one can imagine that if one could cure a disease in a child, you're not only positively impacting the community around that child, but the number of lives here saved is significant.

So the way we approach trying to treat and cure these diseases is through what we call homology-directed repair genome editor, which is a CRISPR-based system in which a double-strand break is created near this mutation you want to correct. And then instead of allowing the cell to repair that break in a mutagenic fashion, we provide a donor template that serves as a template for the recombination machinery to copy the information from the donor into the site of the break. And by doing this, we can make single nucleotide changes or we can insert large gene cassettes. And for today, we'll focus on single nucleotide changes, but there is tremendous versatility.

Now, the overall system is to take our cell type of interest, in our case it's hematopoietic stem and progenitor cells, and put them into the cell cycle because homology-directed repair pathway is most active in S and G2 of the cell cycle. We then deliver our CRISPR-Cas9 nuclease as an RNP complex using a high-fidelity Cas9 protein complex to a synthetic guide molecule with n modifications delivered via electroporation—and obviously we're here with MaxCyte. We then give the donor template on an AAV non-integrating virus. And the reason we chose AAV is it’s a way of delivering the single-stranded carbo that serves an excellent template for combination pathway without being detected by the abundant cytoplasmic sensors of nucleic acids that all primary human cells have. And this system is highly efficient in a wide range of cell types. And as I talk about, it has tremendous versatility.

But for this afternoon, we're going to just talk about the ability to convert a single nucleotide into another nucleotide. The advantage of homology-directed repair is there's no limitation on the change that you can make as there might be in base editing. And the disease that we're focused on is sickle cell disease.

This is the most common serious genetic disease in the world, and every patient has the same thymine instead of adenine and codone 6 of the beta-globin gene leading to a single amino acid change, causing the hemoglobin molecule to polymerize in red blood cells, getting dysfunctional red blood cells that both have hemolysis and cause basal occlusion.

This disease results in decades of early mortality in the United States and in Africa, where it's estimated to be most of the 500,000 [global] cases per year. The average lifespan is in the first decade of life. Now fortunately, genetic medicines have focused on sickle cell disease, and we now have two approved therapies that compensate for the SS genotype that causes the disease. So on the lower left: We have a lentiviral gene therapy, called LYFGENIA that adds an anti-sickling gene to the cell to try to counteract the SS pathologic hemoglobin. CASGEVY was approved as a genetic basis for turning on or upregulating fetal hemoglobin, again, as a way to counteract the SS genes.

What we're doing differently at Kamau is actually not trying to counteract the SS genes, but directly replace S with an A. So we get two effects: We increase A, which is a good thing, and we reduce S, which is also a good thing. And what I'll show you in the first patient treated is this results in a hemoglobin S percentage of less than 10%, which is very differentiated from these other approaches that lead behind close to 50% hemoglobin S.

Now I jumped ahead to the clinical results, but it came from a series of laboratory processes led by some of the people shown here in which we were able to achieve a 60 to 70% gene correction frequency in CD34+ cells derived from people with sickle cell disease. And again, what this results in is that when you take cells that had either been unmanipulated or manipulated, turn them into red blood cells in the tissue culture dish and then measure the different types of hemoglobin in those in vitro derived red blood cells, you see that the pathologic hemoglobin S gets not just compensated by hemoglobin F but gets replaced by hemoglobin A. So it is truly a gene correction process.

As I discussed yesterday, the first patient was treated with nulabeglogene autogedtemcel [nula-cel], the name of the drug, and she was infused in August of 2022. So that was over two and a half years ago. And she came in for her most recent visit and we have the laboratory and molecular values here.

I'll note that she did have a period of prolonged delayed engraftment requiring transfusions and growth factor support longer than we would've wished, but in the end, and now two and a half years later, she has a hemoglobin that is non-transfusion dependent. She has normal platelet and white blood cell counts. She has minimal evidence of hemolysis. And when we measure the fraction, the hemoglobin frac types in her red blood cells, we find that she has 11% hemoglobin A, 9% hemoglobin S, as I said, less than 10%. And the remaining is hemoglobin F. When we look at the alleles that have reconstituted what we find is a sum of her HDR alleles that measures about 4%.

There was an initial period of stability in the first six months. There was then a period of clonal outgrowth that resulted in a single clone growing out with a concomitant proportional reduction in all the other clones. But since the hematopoietic pressure has now resolved, that single clone shown in the orange and yellow is slowly going down over time. And the other clones, including the HDR clones, are now proportionately and equivalently increasing over time. And this has now led to maintenance of a polyclonal engraftment situation, which is a protected issue for long-term engraftment.

Now, as we are translating all of this into the clinic, because it doesn't happen immediately, discoveries were made about the effects of AAV as the donor template in hematopoietic stem and progenitor cells, and the work from Raffaella Di Micco and Luigi Naldini highlighted that AAV, while affected that bypassing the cytoplasmic sensors, does activate a p53 stress response, presumably from the IDR and single-stranded nature of the genome.

And then we showed that, in fact, that stress response was directly proportional to the amount of the AAV that got into the cell. So if there were ways of decreasing the amount of AAV, we could decrease that stress response and presumably enhance the pace of engraftment of both the entire drug product but also the HDR cells as well.

So I’m going to turn it over now to Jason who will talk about how Kamau has implemented a process that achieves a healthier cell product as well as with higher HDR efficiencies.

Jason Skowronski: Thanks, Matt, for setting the scene. Happy to go into our version 1.0 process first, and then we'll go into our version 2.0 process. So our version 1.0 process follows a standard immunomagnetic selection. We bank patient cells until we hit a required CD34+ dose. This then goes into a separate gene editing process, which is five days long in the old process.

You'll see we shortened it with the version 2.0 process. There's some other changes we also made to the culture. We use culture bags. Later on, you'll see we switched to a G-Rex bioreactor-like system. We also use the Lonza electroporation device, where we then switched to a MaxCyte electroporation device. Our process changes, like I just talked about briefly, were summarized with five major changes: the MaxCyte, the G-Rex, the three-day culture and then what Matt was just talking about. We noticed that the cell health might be impacted on an elevated DDR response by AAV. So we worked to reduce it by over fourfold from 2,500 to 625 MOI.

Additionally, in combo with that change, we saw a reduction in HR. So we thought, how can we increase HR while still maintaining cell health, which ultimately led to the inclusion of a 53BP1 inhibitor.

Our version 2.0 process, like we just mentioned, was shortened into a three-day process to promote cell engraftment. The fewer days your cells are in culture, generally the better. So we believe this five day to three day shift significantly helped our process while maintaining cell recovers, which I'll show the data for shortly.

Our day zero is just a cell thaw, a wash, in the culturing to G-Rex vessels for two days, where we then edit with our gene-editing reagents on the MaxCyte, we add our AAV6 donor template post electroporation. And then we seed back into the G-Rex vessels for another day before we g surface cells. On the MaxCyte electroporation device, we saw a direct improvement in HR compared to our previous electroporation system.

This was without significant amounts of optimization. This was just a head-to-head study of static electroporation systems listed here: the CL-1.1 and the Lonzo Nucleocuvette cartridges. Here we ran a pilot study with up to four donors. And what we noticed with the G-Rex and MaxCyte in combination, we saw a significant increase of HR, we saw a better HR to indel ratio, which would help with correction ultimately of the cells in giving the curative allele distribution. Additionally, we also saw a greater than 100% cell recovery, despite shortening our culture duration here.

Now as Matt talked about earlier, we know AAV6 has an effect of increasing the DDR response, so we thought, again, how can we reduce this response but also keep HR at a similar level? You'll see from the graph on the right, we ran an experiment and, of course you add less AAV, you get lower HR.

Our solution here was to introduce the 53BP1 inhibitor. What this does is it biases repair towards the HDR pathway and, ran in a control study here, what we noticed was with 625 MOI our HR was about 38%. We saw about a 50% increase 56% with the inclusion of this booster, so significant improvement, even beyond having a AAV6 with a 1250 MOI without the booster.

This also leads to a reduction in cost, which is of course good for the bottom line during therapy [development]. Here in the next three slides, we did a version 1.0 to version 2.0 comparability study. The reason why we introduced the Annexin-V/7-AAD measurement is it really is able to catch early apoptotic cells. Standard cell counters like the NC 200, the cellometer, they're not often able to catch it and generally require maybe a day of culture to really truly catch and see where your ending viability lie.

With this flow assessment, we're able to see a net 13% increase between the two different methods as well as maintain similar recovery despite two fewer days in cell culture, although you see an increased cell yield by three days of pre-stim versus two days of pre-stim with our current process, we lost about 50% of our cells during the electroporation process, which ultimately led to the recoveries being about equal.

One other measurement we worked to do was basically we plated cells in methylcellulose to do a colony-forming unit assay. And what this is: It's basically a gold standard right now for predictive engraftment of your HSPC products. On that we then ran NGS [next-generation sequencing] on the individual colonies.

What we noticed was a higher frequency of cells carrying at least one HR corrected allele, which, to treat sickle cell, you just need at least one corrected allele per cell type to then get to the hopefully symptomatic-free tool. Additionally, we let a biallelic indel reduction in combination with that HR for allele increase. For predicted engraftment measurements—and then I'll show an in vivo engraftment to directly show an improvement—again, percent CFU assay, we compared it to an RNP-only control. What this is, is without AAV6. So cells are electroporated, you add your RNP complex, you see your cutting efficiency, but you just don't add the AAV6 to your cells.

And what we noticed was a substantial improvement, not only from the version 1.0 process, but we also saw an improvement compared to that RNP-only control, which really showed no significant difference, that the AAV6 was not harming the colonies in our predictive measurement here. We also ran LT-HSC, which was a marker profile that you can measure for LT-HSCs, and we noticed about a 50% increase compared to our version 2.0 process. Now showing the actual engraftment data, we saw for the version 2.0 process, about a four to five x fold improvement in engrafting, including the bone marrow, which shows a high likelihood of success in engraftment of our version 2.0 process.

This table’s really just to show the amount of effort and characterization that went into our version 2.0 process. We didn't just measure viability, we didn't just measure our standard or these criteria, we measured a plethora of other conditions to really characterize and make sure that our version 2.0 process was truly better than our version 1.0. And if you look and you compare each individual value to each other, there's a significant improvement throughout for total cell recovery. While they're about the same, again, you have two fewer days in culture, which has led to some of these other conditions’ improving, like LT-HSC, better CD34+ purity of our products. Those are some other metrics as well.

And then what we did is we transferred our process to our CDMO ElevateBio. We ran re-engineering runs, and what you'll see here is these three engineering runs were very close to the results that we got from our version 2.0 study previously. So there wasn't a substantial change, meaning tech transfer was successful.

And then we compared these to the original patient 1.0 data, and what we saw was a significant improvement across the board in all these metrics—better viability, better editing and reduction of INDELS at one of our off targets. So significant improvement.

Just to conclude, our CMC amendment is approved by the FDA, we're in active manufacturing. Hopefully our next patient will be infused soon. And then just to highlight what we're doing currently, we're manufacturing at our ElevateBio facility, and then we have our two clinical sites, Stanford University and WashU at St. Louis. And numbers for our clinical trial. Then, just to show truly the effort that went into it, not only on the Kamau side, who drove this version 2.0 proven into the clinic but also all the people who worked prior at our former company, or Graphite Bio, to really develop and create this version 2.0 process.

Just wanted to say thank you to both groups of people, and then all our partners as well. Thanks.

Sean Menarguez: Well first off, thank you Matthew, thank you Jason. Congratulations on all the success pushing forward your HDR platform, but also now translating that from a research process into a more robust process. An overwhelming theme throughout this week: I think everyone in this room is very cognizant of the marketing environment that we're all operating in, and it seems that in the industry there's a balancing act of speed to the clinic to getting that first human data then also building a robust processes from the get-go.

And on that, in the talk with Dr. Terry Fry, he brought up how details matter and details matter in manufacturing. You don't want to start off with that balancing act of speed to clinic, while also producing, and then working through robust processes. What are the key considerations that your team takes in that early development?

Matthew Porteus: I think there's a lot of different answers to that question, but a couple that I'd like to highlight, one is maybe what we didn’t but what we should have done. Which is if you're developing a process at a small scale to make sure that when you scale it to a clinical scale, it doesn't take going back to ground zero in terms of having to repeat all the experiments. And I think a highlight of that was when we developed a process in the lab, we were culturing ourselves in place, routine laboratory process, but when we moved to bags, and Jason was there, there was a lot of troubleshooting going on about, wait a second, the cells aren't growing the same in a bag as they were growing in a plate.

But when we switch to the G-Rex, now all of a sudden, we have a bioreactor that scales, and they make that argument and we have experimental data that's justified that. Similarly on the electroporation process, what you want is a process where if you're piloting a number of variables at small scale, which can be done much more cheaply in a more lean caution, then when you move to the larger scale, you don't have to move to anticipate redoing all of that. Again, there was some troubleshooting that was done. And then that leads to the third point, that those partners have all been partners. Yes, we have contracts and you need to write contracts and we have to pay our lawyers, but what's more important than the contract was to feel like the people on the other side are your partners. And that when something went well, you shared the good news. And when something doesn’t go as well as expected, you brainstorm together about how we could solve that problem, or was it a real problem?

And so identifying your vendors and your suppliers and your partners as real partners rather than just as contractors, I think is another important part.

Sean Menarguez: That’s awesome. Thank you, Matt. And Jason, for you, as you look to appreciate the discussion on optimizing to the process 2.0, as you spoke about, to build scalable processes with both systems, et cetera, what are the key attributes that you look for when you're designing platforms that scale-up?

Jason Skowronski: I think Matt touched on one of them, but clear scalability. The process has to look identical as you move from small scale to large scale. It's nice if it's a closed system along the way. I know G-Rex has some closed vessels, so that's obviously nice. Closing the system up as much as possible wherever you can at the start significantly helps your process later on. Matt talked about how we started in six-bowl plates, we went on to flask and even closing it up into something like a cell culture bag, which we later ordered, deciding we wanted a gas-permeable cell culture bag, it was still hard to ultimately harvest those cells. Everything had to be individual, put into a conical tube, connect it, and added hours into the manufacturing process, which people don't often think about. They think you're in a PD lab or a research lab. I can do this in a BSC in an hour, in two hours. But when you go into a clean room and anyone who's in MSAT watching people do a manufacturing process, you'll know it adds many, many hours. There's a lot of red tape and quality that has to go into it. So something as simple as I press a button, I connect the tubing to a G-Rex and GATAREX, and it's all harvested immediately, the benefit is very clear.

So I would say scalability as you move up and hopefully your data doesn't change. I would also add that hopefully your clients are willing to work with you as you encounter issues and can troubleshoot them. They don't just hand you this technology and then say, well figure it out. These are our reagents, this is our equipment that costs hundreds of thousands of dollars, and do a hands-off approach, and not really help you with it. For example, we probably have gas-permeable culture bag issues. There wasn't really anyone to go to and kind of help through those issues, whereas we had successful partnership. You guys develop the therapies as well for the electroporation process. Or like that.

Matthew Porteus: And if I can add one thing. We talked about scale up, but there's the other concept that scale out. And Jerry Cacia, who was CTO at Graphite and he has decades-long experience in the biotech world, particularly at Genentech, I think kept the team focused well on developing processes that would ultimately scale out. Now we're not scaling out at this point in our early-phase clinical trial we’re doing. Every patient is incredibly important and unique, but this is a disease that we ultimately are going to need to treat thousands of patients per year.

So I think in our next stage, and Jason and I talk about it, Jason probably thinks about it even more than we talk about it, is how are we going to take on this version 2.0, which we're very proud about, and create a version 3.0 that will allow us to scale out, not just scale up.

Sean Menarguez: Does modularity play a role in that? As you look to scale these processes, the flexibility with the process and adjacent technology, from that standpoint, is that playing into version 3.0 aspects?

Jason Skowronski: I think it's about having a workflow that really just connects all of the systems together. That's kind of how I’m thinking about it. I think having equipment, necessarily, that does everything for you. But then is a bulky piece of equipment… you have to have one therapy in each device and in culture in that [device] for the entirety of the process. You have 10 patients, you need 10 of these devices, which you then have to culture for a while. That's kind of a bulky process.

Whereas I see you have equipment, robotic arms, whether it's a G-Rex vessel that you can have your entire product into one or everyone's been passing by this large move and having a cube which is a consumable and not necessarily an actual equipment, but you can have multiple running at the same time. So that's how I'm kind of thinking about it. How can you connect, say, maybe a MaxCyte electroporation device to a culture vessel to continue that process.

Sean Menarguez: Thanks for that, Jason. Matt, too, with your work at Stanford as well, I appreciate in the beginning of this session you spoke about those early considerations were small-scale translating to supplier scale. These academic groups that are working on promising therapies, how should they look at manufacturer partnerships early on and value that long-term thinking?

Matthew Porteus: I mean, yes and no. I see us sometimes getting a little too wrapped up in how I'm going to manufacture before I have a really good and robust process done to begin with. Like, don't worry about it too much, too early. On the other hand, the yes is once you have started to get hints or even real data that this seems to be working, then starting to think about it is important.

And what we've tried to do at Stanford is build a cell and gene therapy, which we call the Center for Definitive and Curative Medicine, an ambitiously named center—you know, let's cure patients with one-and-done therapies. But hey, if you don't shoot high, you won't reach it. One of the goals of the CDCM, which is what we call it, is to then provide expertise and infrastructure and introductions to laboratory-based investigators who may not have experience and help them start thinking through what they need to be thinking about. There was a panel yesterday led by Tami John in which Cliona Rooney talked about phase-appropriate manufacturing. You don't want to get too far ahead of yourself before you have clinical results because you're going to spend too much money before you know it.

On the other hand, you need to be manufacturing in a way that if you see positive clinical results, you don't have to go back to ground zero to start all over again. This morning Dr. [Christine] Duncan talked a little bit about the lentiviral vector for ALD. Twenty years ago, they were stuck because they had a vector that showed clinical results, but they also had a vector that had theoretic concerns, which ended up panning out. But at the time, they had to make the difficult decision of sticking with what's new or sticking with what they knew worked, balanced against theoretic concerns versus a time delay if they had had to remake the vector.

So you want experience. I think academically we’re moving to a situation where getting early first few patient clinical results with high quality but perhaps not very expensive reagents is the way to go.

Sean Menarguez: And it's for not only the manufacturing process but also to accelerate development in general with safety assays. What role can safety assays play earlier in that journey to help accelerate [development]?

Matthew Porteus: Sorry, I'll take this again, Jason. I'm going to make a provocative statement here, but I think it's actually true. Early on in the development of CRISPR-Cas9, there were a lot of concerns about the specificity of the nuclease.

And we, of course, always are concerned about that, but I think now we have hundreds of people treated and tens of thousands of mice treated, probably hundreds of NHPs treated with genome editing protocols, and I'm not aware of a single adverse event. So I think, yes, we can identify off targets, but no, they haven't shown to be important, particularly no bioinformatic tools and with specificity assays that we have, the improve reagents… I try to encourage our teams and the people I talk with not to let specificity get in the way of efficacy.

I’ll make another statement here. As you know, when you fly on an airplane, you get exposed to a little bit more radiation than we do at ground level. If you have approximately 3 trillion nucleated cells in your body, we have 30 trillion cells, most of them are regular cells, so who cares about those. You generate around 600 million extra double-stranded breaks per hour flight.

So we all flew—well, a few of us didn't fly here—but many of us flew multiple hours just to get here and we tolerate that. So our cells are really good at repairing double-stranded breaks.

Sean Menarguez: Maybe not just in the manufacturing process, but overall, throughout the industry experience, what kind of key process-related hiccups have you seen that potentially delayed or even derailed programs? Do you have any learnings from that, where developers can anticipate or even eventually prevent them?

Matthew Porteus: I'll let Jason talk about his own experiences in developing nula-cel. What have been some hiccups and then maybe generalize from there?

Jason Skowronski: I don't know if it's necessarily safety, but of course we've had hiccups along the way. We changed our process in our first engineering runs. We went through two engineering runs; we found in engineering run two, we had a cell loss. The cells did not recover. And ultimately, we had about a 20% recovery overall. So we thought at that time, Hey, how can you improve? Ultimately, that led to increase cell culture. And then we went to our five-day process where we saw in engineering three, four, and five significant improvements in cell recovery.

Now, if you remember from the talk, in version 2.0, we're back to a three-day process, which is two days less. So I mean, you have iterations constantly of your development. We obviously thought that was to the culture system overall, and we thought the G-Rex bioreactor system was helping us a lot more. That’s just one very notable change that we made throughout our process. I can go into more details.

Matthew Porteus: I'm going to give Jason PTSD and myself PTSD. Particles. Oh my God. The process is done well, you've manipulated hundreds of millions of cells, you've done this in a way where it's carefully done, the manufacturing specialists are trained, and then you're required at the end of your manufacturing to do visual inspections. Inevitably you see a particle. Right now, I'm going to just say it, I think the guidance around how to interpret what those particles mean and the frequency that these particles and the clinical significance of these particles are not well-managed.

If you can see a 100-micron particle in a bag, I can tell you that is of absolutely no clinical significance. We were talking with a team the other day about particles, and as a data point analogy, she's a rock climber: You get probably more debris into your bloodstream after you fall on a rock that you would possibly get from a drug product. So I think we need to find a better way of handling particles.

Sean Menarguez: I mentioned this in the delivery step as well, to evaluate not only with the AAV platform but also the elctroporation platform, with virals and non-virals, how do you look at that criteria to go in the direction of electroporation or go in a direction from there?

Matthew Porteus: When we in the entire field started the process of genome editing of hematopoietic stem cells, electroporation was the only game in town. That was it. So now, one of the things, we learned, and I think others learned is electroporation, so running cells through a device that puts cells through an electric field and creates holes in the membrane to allow your macromolecules to enter, not totally benign, but honestly, and Jason can speak to this, if you don't have any macromolecules to enter the cell, the cells come through pretty darn well, assuming you get the parameters right, the exposure, you're not using too much voltage, and so on and so forth.

Where the toxicity from electroporation mostly comes from is the macromolecule you're delivering. And that is why using highly, highly purified, high-quality reagent in your electroporation is so important, because if your guide RNA or your Cas protein or your mRNA has contaminants in it, those contaminants will enter through those pores and lead to cell stress. And so this is, again, where working closely with your suppliers about having high-quality reagents is so important. Because you could misinterpret the toxicity of your experiment as coming from the device rather than from your reagent. And deciphering those is really important.

Jason Skowronski:And in some of those controls, like Matt's talking about, I showed in the slide, an RNP complex control, so not adding any AAV to cells. And when we noted that high viability throughout the process, really no toxicity at all. And that's kind of why we decided to benchmark our percent CFUs against that RNP complex to really show there is no difference with our new version 2.0 process while still adding the 625 MOI AAV sets. So I think that’s a very good metric to hit, even better if you can hit untreated without any editing or electroporation at all, but we're all shooting at that target.

Sean Menarguez: Not only reagent selection but also to have those reagents be within the process and the culture of the cell at different concentrations that are used, different parameters that are used throughout the different platforms. Jason, how do you look at the collaboration aspect and service?

Matthew Porteus: Before we get to that, I wanted say this before it gets too far. We’re using AAV, so it's “viral,” but I will disclose this: the cost of goods per patient for our AAV is around $2,000. So this is not a driver of cost of goods for us, and the reason is, is that while a AAV manufacturing for in vivo use requires enormous doses of AAV, for the ex vivo use of AAV, we are getting thousands, if not tens of thousands of doses from the same bioreactor that generates a handful of doses in an in vivo setting. So the fact that AAV manufacturing is now getting better and better and cheaper and cheaper because of the need to drive down costs in the in vivo setting is helping us in the ex vivo setting to make that a reagent.

Sean Menarguez: How do you weigh those specific nucleases or reagents [for packaging into AAV with CRISPR] compared to the non-viral setting approach?

Matthew Porteus: The challenge with, I think, delivering an editing reagent in an AAV is that AAV is assigned to deliver, to express a gene for ideally a lifetime of that cell. Editing has the advantage of being a hit-and-run process, and in fact, the longer you express your edited gene, the more off-target effects you see.

So if you're going to deliver an editor in an AAV, and this is a genome editor. Epigenetic editor, you might add to for kinetics. I think you need to think carefully about how you're packaging that editor into an AAV and what are the potential consequences of now expressing it for days, weeks or even months.

Sean Menarguez: In your talk yesterday, you brought up this vision for cell therapy, making it very futurist. I appreciated the analogy of using the car, the OG car, automobiles that were manufactured hundreds of thousands [times] per day globally, really reaching scale and access. In your perspective, what is that vision of having a factory-type manufacturing in our industry?

Matthew Porteus:Well, I think it's a chicken-and-egg question to some extent. Our suppliers need to make revenue. Right now, to make their revenue, they have to charge a high cost of goods. I will joke that if you have a pizza restaurant and you only sell one pizza a month, you have to charge $10,000 for that pizza. But if you're able to sell a thousand pizzas. Now you, you can charge $10 per pizza. So we, as the end users, need to get the volume to our suppliers so that they reduce unit cost. And so that's going to just ratchet up slowly over time.

And then I think it goes back to the scale-out, which is: How does a piece of equipment integrate into an industrialized manufacturing process? Are we just going to put a bunch of sequencers in a room and do it? That’s how we sequence the genome. Or is it going to be more like an industrial car chain, where the car just moves along and you do parts? I don't think we can do that for cell manufacturing, but I keep thinking about […].

Sean Menarguez: Absolutely. I know we have a few minutes left. Any questions from the audience? Awesome. I’ll lead with this. If there is one big key takeaway that you saw throughout the developments or advice…?

Matthew Porteus:Well, I’ll turn this to Jason since he's been with this product from basically day one. What would be your key takeaway?

Jason Skowronski: That's a good question. I might have a couple. I don't know if I could summarize it into just one. I would say I would establish, while you're developing your process, a better baseline of all your assays, so you can really assess as you make your changes across a variety of things, beyond just viability or purity—just do a standard assessment. We did really look at things like the AAV6 effect on the p53 and then DDR response, which ultimately led to us lowering the AAV. Now we didn't have the 53BP1 at the time to inhibit or to lower it in combination, so we knew that might lower HR. But I would just say better standardized assays, better characterization of your process, the R, the stereotypical by liability veterans, and then HR. So really look at the cell fitness and the overall cell health as you make process changes. And then the other piece: More closed systems, we can automate more, so try to just look at what systems scale before you choose a piece of equipment, [etc.].

Matthew Porteus: And the other thing I will say, and I'll echo what Fyodor [Urnov] mentioned this morning, and I've heard him say it other times, and I say this as well: You learn more from a single patient than you can learn from a thousand mice. And so we absolutely need to respect that this is a human being with a serious disease and make sure that the consent process is done right, but also not being afraid to take what's best at the time in treating a patient or two and learning as much as you can and anticipating that the first time something goes into a patient, you're going to learn a ton.

The exception to the rule is CASGEVY. That didn't happen. It kind of worked right out of the box, but every other technology in the cell and gene therapy space has had to iterate and go back, figure out how to make your AAV better, how to find the dose better. So I think anticipating that error process and doing business planning around that iteration is something maybe I will go back and think differently about: How are we going to plan that version one, might not be, will not be perfect? And we're going to have to plan for version two.

Sean Menarguez: Thanks.

Matthew Porteus: Yes, of course.

Sean Menarguez: Question?

Tami John: Yes, Tami John from Stanford. I appreciate the openness with which you are sharing the trials and tribulations of manufacturing changes and updates and, and so I think it's really exciting to hear not just that there was trouble with the first cases or the first case, but that there's some exciting changes that you're willing to show with the community. My question is once we see a level of success that's exciting and talk about scalability and upscale out, is this a platform that you foresee a quick way to scale out to other orphan diseases? How is your platform set up where once we see success with correction of sickle cell disease, we might be able to quickly move to other diseases, since this is a newer platform gene correction?

Matthew Porteus: What do they say? You go to Midas, you get a muffler. You come to me, I'm going to tell you absolutely that one of the reasons to establish HDR in sickle cell disease is then the ability to apply that across the range of diseases that can be cured by a genetically engineered autologous HSC transplant. Absolutely.

The challenge, of course, or one of the challenges, is that those are much rarer diseases and so in the process of proving it out in sickle cell disease, we can hopefully also prove out how to develop it in a more cost-effective manner and ultimately with a manufacturing process that's cheaper, so thereby, for these rarer diseases, it becomes economically feasible to start treating five, 10 and 100—there's now I think 600 known inborn errors of immunity, almost all of which could be treated, in theory, by this sort of process. And so how do we get there?