Last year’s Cell & Gene Meeting on the Mesa took place at a lush resort in northern San Diego. The coastal October air was cool and crisp, and the attendees were filled with cozy optimism as they clinked glasses of wine and watched a Southern California sunset. Attendees were as excited as children waiting to open a mountain of presents as they awaited the FDA’s approval of Casgevy and Lyfgenia.
The level of confidence was so high that manufacturing-related discussions dominated the conference. It was almost as if cell and gene therapy was a foregone conclusion, and all they wanted was for it to take off. The only remaining challenge was figuring out how to scale production to billions of doses.
However, in a year when some of the most exciting cell and gene therapy technologies appeared to bloom in start-ups, the companies dried up, having burned through their cash with little to show for it. The cuts were not limited to new ventures, affecting some of the biggest players, including Pfizer. Even Bluebird Bio, the company behind Lyfgenia, had to scale back. People have begun to sweat, and it is not because the 2024 Cell & Gene Meeting has relocated to Phoenix, where temperatures have risen above 100°F.
“We have seen lately a lot of [cell and gene therapy] companies go bust,” Alexander Seyf, CEO and co-founder of software company Autolomous, told Inside Precision Medicine. “Some raised [hundreds of millions of] dollars and have been operational for years, and what did they get out of it? What assets have they created? Nothing. Just nothing.”
Instead of sprawling out in lawn chairs with toothpick umbrellas, attendees, primarily founders, C-suite members, and investors, huddled in every nook and cranny of the air-conditioned lobby at the Arizona Biltmore, getting down to business and talking through the real problems they face.
No problem is too small, and Seyf’s Autolomous is a perfect example. They are doing the equivalent of turning photo albums into folders of pixels in the cloud, but in the realm of biopharma, which is essential for streamlining logistics for, say, ordering a cell therapy for a patient.
Seyf said, “When you want to schedule a patient, it’s not like Amazon, where you can say, ‘I want this and deliver it tomorrow’—it takes almost a week. You need to call, email, and text to synchronize with transportation and so forth logistically.”
The theme of this year’s Cell & Gene Meeting on the Mesa was not building empires and shooting for the moon but instead coming down to Earth and breaking things down into manageable, attainable building blocks, making sure not to overlook anything.
Learning from failures
For Hilary Eaton, PhD, chief business officer at AI-first protein design company Profluent, cell and gene therapy isn’t just some exercise in science and medicine—it’s real life with real consequences.
“I found out that my two IVF babies and I all have a rare genetic disease,” Eaton shared. “So, it started feeling a lot closer to home, this idea of what are the options for disorders that take a long time to diagnose, where there’s no specific known genetic variant related to it. It can seem like a very academic exercise to filter through all of these indications and think about which ones match my platform or gene editor. Then, all of a sudden, you’re sitting in a genetic counseling office.”
Around the time Eaton found out about the rare genetic disease in her family, she was approached by Ali Madani, an AI/ML wiz who had just founded and spun out Profluent from Salesforce. Madani, who had caught the bug to pursue curing disease, showed Eaton the high-capacity language model he developed and trained on the largest protein database available (~280 million samples) while at Salesforce, resulting in the ProGen moonshot that advances generative modeling for protein design. She was hooked.
That massive, high-quality dataset would grow into the billions, and some fundamental advances in AI have allowed Profluent to extrapolate and unlock novel protein functionalities that don’t exist in nature. Eaton said that one thing Profluent can do is find evolutionary dead ends where nature didn’t provide the selective pressures to result in a protein, such as a gene editor, that has all the features someone could dream of. She also said that what they’ve developed isn’t a “magic wand.” Things fail all the time. But it is in these failures that historically have been swept under the rug in experimental fields, causing all sorts of havoc (such as the reproducibility crisis), that some of the greatest leaps are made.
Speaking to her experience of failure in experimental labs, Eaton said, “Sometimes after a well-designed experiment, you may be able to say, ‘I didn’t have the right control’ or ‘somebody left the fridge open.’ There are countless experiences of working a 100-hour week, and you have no idea why it didn’t work.”
But that doesn’t have to be the case when using tools like AI/ML. If a design cycle fails and none of the models’ predicted sequences actually have a successful functional outcome, that doesn’t mean that there’s nothing to take from it—instead, something was learned. No one gets everything right the first time around, and knowing what fails is critical in the pursuit of making things work.
Can’t dodge delivery
A major rate-limiting factor for any cell and gene therapy is delivery—no matter how nifty the mechanism to correct a disease, you’re not going to cure anything if the medicine can’t get where it’s supposed to go in enough quantities without harming the patient.
With the success of the mRNA vaccines to SARS-CoV-2 being doled out in the billions via lipid nanoparticles, it seemed that the cell and gene therapy field was moving in the same direction. I’ll be the first to admit that I had hopped on the non-viral train, thinking tools like adeno-associated virus (AAV) would be left for dead.
“If you need a small dose that’s going to activate the immune system by exposing it to a small amount of form protein, that’s fine,” said Dyno CEO and co-founder Eric Kelsic, PhD. “But for gene therapy, you need to deliver billions of payloads to cells all across the body, and the more cells you reach, the more effective the therapy. Efficiency is key. It’s exciting to see innovation in many different areas with LNPs, and one of the ways they’re becoming better is by making them look more like capsids. That’s not surprising to me because capsids are so amazing as proteins and as machines.”
Over a decade ago, Kelsic began developing an AI-backed platform in George Church’s lab to optimize capsids for delivery. While he says the company is a delivery company that will consider all technologies, he’s stuck with his AAV guns and still believes that’s the way to go. A key part of the Dyno platform is what Kelsic calls super-Darwinian evolution to make the best possible AAVs.
“Evolution is just an algorithm—it’s a way of making and recombining mutations according to the fitness of those sequences,” said Kelsic. As part of Dyno preparation, we considered that algorithm and asked, ‘Can we do better?’ In short, we can do better than a random approach to making changes because we know a lot about how proteins work and can look at the data and follow the data to do better than a random approach. We can make more significant leaps in the human space once we have some empirical knowledge, whether from experiments or simply looking at what nature has done before.”
That being said, Kelsic doesn’t think it’s so black and white, as delivery will be either all viral or non-viral. But he also knows he doesn’t need to conquer every aspect of delivery at Dyno—what he needs is to open up new therapeutic possibilities.
“If you can reach more cell types in the brain and, to be more specific, reach more cells at a lower dose, making them safer and easier to tolerate and, ultimately, lowering the cost of goods, the therapies better be cheaper to produce,” Kelsic explained. “This gradual change of making gene therapy work, then making it more applicable to a broader range of indications, and finally more accessible or affordable, can be accomplished just by focusing on delivery for a decade, possibly another decade.”
The regulatory convergence “chicken-and-egg”
Just a month or so after the 2023 Cell & Gene Meeting on the Mesa, Astellas came out with some results from their ASPIRO Study in X-linked Myotubular Myopathy, which was published in The Lancet Neurology, that no company wants to do. While their AAV gene replacement therapy for pediatric patients with X-linked myotubular myopathy (XLMTM) showed signs of ventilator dependence and motor improvements, four of the trial’s 24 patients died as a result of serious adverse events.
However, the notion that every single cell and gene therapy will act as a silver bullet and eradicate rare monogenic diseases on the first try is a pipe dream. Incremental improvements can yield significant benefits. For a disease like XLMTM, 50% of children die before the age of 18 months, and those who do survive are relegated to being reliant on machines to breathe, wheelchairs to move, feeding tubes, and a very low quality of life. The burden on families is enormous.
“When you get beyond the inevitable headlines and safety events, there were 24 boys dosed in that study to this day,” said Richard Wilson, senior vice president at Astellas, who is primarily responsible for leading genetic regulation efforts. “We saw a massive response in patients coming off ventilators. Twelve of the boys were able to sit unassisted. Some of them could stand and start walking. It’s hard to walk away from that.”
According to Wilson, to get to the bottom of this program and find success in the rest of their pipeline, such as Pompe disease and cardiomyopathy associated with Fredreich’s ataxia, Astellas is striving to become an end-to-end gene therapy company, which includes development, manufacturing, and regulatory. However, they will not do so in a fortified silo with no inbound or outbound interactions—quite the contrary. For example, Astellas is already working with Kelsic and Dyno on their quest to optimize AAV-based gene therapies.
According to Wilson, the conversation has to go beyond the business-to-business realm. Open communication is essential for regulatory agencies to engage in discussions to improve and accelerate the delivery of these new technologies to patients. On that note, Wilson stated the concept of regulatory convergence came up repeatedly during the first day of the conference.
“We realized yesterday’s regulatory system was not built for AAV gene therapy or cell therapies,” Wilson explained. “So, while the FDA is very publicly and laudably growing, evolving, and trying to get that message out as much as they can through Peter Marks and Nicole Verdun, we think [Japan’s Pharmaceuticals and Medical Devices Agency (PMDA)] is trying to follow suit and certainly some of the conversations going into the EMA as well. We need to connect with all those groups as they figure out how to change their business so that they can regulate, improve, and accelerate the delivery of these new technologies to patients.”
Wilson continued, “At the same time, we must understand how to adapt in response. So it is a fascinating chicken and egg situation in which things you would not have said to a regulator five years ago may now put you at a competitive disadvantage or prevent you from doing your best for patients. If you don’t go and ask that, you won’t be able to learn from our mistakes. There’s just so much being done that’s novel and innovative. The more people are connected to that knowledge network, the better it is. This is not a finished story, and we do not know who will write the rest. Is it the start of the end or the end of the beginning?”
From what I have seen on the first day of the 2024 Cell & Gene Meeting on the Mesa, it appears that people received a wake-up call, awoke from their dream state, and are ready to get to work.
