After Driving an N-of-1 Therapy for Her Son, an Advocate Turns to Helping Others

Yiwei She, founder of the TNPO2 Foundation, discusses how her family was able to treat her son Leo with an experimental ASO with relative speed, the work the TNPO2 Foundation is doing to accelerate the diagnosis of other children with ultra-rare conditions, and its efforts to find accessible and affordable pathways to treatments for others.

Daniel Levine: Yiwei, thanks for joining us.

Yiwei She: Thanks for having us.

Daniel Levine: We’re going to talk about your son Leo’s diagnosis with an ultra-rare neurodevelopmental disorder, how you were able to advance an experimental antisense oligonucleotide therapy to treat him, and efforts by your TNPO2 Foundation to help other children with ultra-rare conditions. Let’s start with Leo, though. He was diagnosed at four months. How did he come to be diagnosed with an ultra-rare condition?

Yiwei She: So, he was born with an abnormally small head and we watched it. It wasn’t growing at the right speed, so we started doing all kinds of workups. He had his first seizure when he was two and a half months old, and that really kicked us in the behind. And we managed to get him involved in a clinical trial at Boston Children’s, which looks at genetics for infants with epilepsy. And that’s where we got Leo sequenced with whole genome sequencing as part of a research study. So it took a research study to get a diagnosis for him.

Daniel Levine: He was diagnosed with a severe neurodevelopmental disorder caused by mutation in a copy of his transportin2 or TNPO2 gene. What were you told about this gene and his condition when he was diagnosed?

Yiwei She: We were told that the gene, there’s very little understood about the gene. They did give us a link to the research paper that had discovered this gene, which was only published about two or three months before he was born. So we were told that science knew very little and to expect to basically just go home and love him and there’s nothing that anyone could really do for him.

Daniel Levine: He’s one of only two known patients with the condition. Is anything known about how the disease progresses, its effects, or what you might expect?

Yiwei She: Yeah. Everything we know about his mutation comes from the one child who had it before him. They’re now about nine years old and they haven’t changed very much developmentally since they were three years old. And right now they have the mental capacity of a little under a 1-year-old child. So, it’s a very severe prognosis and they still suffer from seizures.

Daniel Levine: There have been efforts by a few others before you to develop an individualized N-of-1 therapy for someone with an ultra-rare condition. How did you come to view this as a possibility for Leo?

Yiwei She: So, I think when this happened, we wanted to do everything our family could possibly do for him, and we didn’t think of it as an individualized setting back then. It was more of there’s so little known and what could you possibly do except to target the root mutation. And then once we found the stories of the other families, say at Boston Children’s, like Tim and Julia and others and Lauren, we thought, well, this might be the perfect, I mean, this might be the only thing we could do for him. And we wanted to do everything.

Daniel Levine: You’re talking about Tim and Julia Vitarello?

Yiwei She: Yes, exactly.

Daniel Levine: Some of our listeners may be familiar with Creyon Bio, which we’ve written about for the Global Genes Next report. Chris Hart of Creyon will be featured on an upcoming Rarecast. And how did you come to work with Creyon Bio?

Yiwei She: Yeah, so we got connected to Creyon through a personal friend. We used to live in the Bay Area and one of our friends, they share the same investor as Creyon. And when this happened to us or to Leo, we called everybody we knew and said, look, please help. This is what every family does. And this friend really came through us for us not only connecting us with Creyon, but also members of the Dunn lab and some other major labs. And we were so grateful for her efforts. She was also—we were in the tech industry. We were not in biotech. So she was our only friend who was well-versed in biotech, so she knew what we were facing.

Daniel Levine: You were able to move from the start of the research for an ASO to actually treating Leo with an experimental therapy in a year. The therapy Leosen is named for him. But what enabled this to happen as fast as it did?

Yiwei She: I think capital and the willingness to work or to operate at risk is what. We were willing to operate at risk, meaning that if experiments didn’t pan out, we were willing to walk away from the money and just say, okay, that’s fine. The mouse studies didn’t work out, this other model didn’t work out, that’s fine. And if the ASO weren’t going to work out, we were just saying we made a good faith effort. We’re not sad about the money that we spent. The only thing was trying to do the right thing for Leo and the right thing was to have an option and to see if it was scientifically sound and safe or not.

Daniel Levine: Is it expected that the ASO would slow progression, halt it, or is there any hope of improving Leo’s condition?

Yiwei She: So, we don’t really know, given that there’s only been one case and our expectations were set by the handful or so of other, so-called N-of-1 trials. On one end, I think the most optimistic outcome that I didn’t really put any weight in it, I thought that was of an ASO basically being a miracle for lost abilities. The patients regained a lot of the lost gross motor abilities. On the far other end was say in which they saw some signs of benefit, but there was also a severe adverse event. So, we didn’t know and we had no expectations going in where we would land, I think, on that continuum. Obviously we were hoping for the best. But I would say that in January, I think about six or seven months into the research project, we did get rescue data coming back from brain organoids. And that gave us a lot of hope for not just halting the progress, but also the possibility of rescuing something developmental. But our disease model was very simple and naive. It was the brain organoids that showed a rescue in the size, so the brain organoids also had microcephaly and then dosing with the ASO rescued that a small amount. So we thought if it could rescue one of Leo’s first symptoms, maybe that it could also rescue the development.

Daniel Levine: Well, how’s Leo doing today?

Yiwei She: He’s doing well. I think, as with all of this, things turned out to be more complicated and nuanced than I would’ve hoped, namely that it seemed like he was responding best during the dose escalation phase when he was getting dosed every single month. Since we moved to every three months, things have been a bit slower and we have been waiting for the FDA to give us the go ahead to dose more frequently. Now that the FDA has accepted our new proposed doses schedule, we’re hoping to go back to seeing more clear signs of efficacy.

Daniel Levine: One of the things you’ve done is published the IND documents on the TNPO2 Foundation website. Why is that?

Yiwei She: So, we haven’t fully published them yet, but that is our plan. We’ve made a commitment to do so. The reason is that the FDA and the regulatory steps, they’re one of the biggest hurdles to cross. And having crossed it, I think it’s one of the lowest costs and most impactful actions that we can take as a foundation to help other foundations, to help other families to make this long trek.

Daniel Levine: You were able to move from the start of the research for an ASO to actually treating Leo with the experimental therapy in a year.

Yiwei She: I think the overarching principle was that we had one priority and that was Leo’s clinical wellbeing. And we were not as constrained in terms of resources as other programs. We could operate at risk and parallelize, and that would result normally in what is normally thought of as a lot of wasted effort, but we thought of it as risk mitigation. So, for example, in the disease model side we had a lot of experiments going in parallel, and we didn’t wait for one experiment to finish before starting up another like say, model organism. On the lead side, we did the same thing. We ran many of the steps in parallel and we had even engaged, say, Charles River as a parallel program because Creyon was so untested at the beginning and later on we cut back to just Creyon when their leads were showing a lot of promise. And so I think because of the resourcing, if we had taken donations, it would’ve been very difficult to run the program the way we had done because at the end of the day, we would’ve had to make every single dollar count, whereas the way we considered our resources, failed experiments or experiments that were run in parallel for the purpose of advancing the whole program forward as much as we can from our perspective, every single day Leo got a day older, there was no changing that, but what we could do was every single day, the whole program could move forward faster than that because his yeast could move forward, his mouse could move forward, his worms, his fruit flies, all of these Leo models could move forward today. And on the lead side we could parallelize the progress of more leads. So that’s the basic gist of it. But behind that is ample resourcing. And I think compared to commercial programs, they move faster because more constraints are placed upon them. You need pharmacology, you need them to be commercially viable. So the leads are subject to more restraints like half-life and higher safety margins and things like that.

Daniel Levine: Yeah. I actually want to zero in on that point because I’ve seen other people do this and there’s often a choice people make between doing experiments serially or in parallel. And I’m wondering if you could just expand on that a little and the thinking behind that and the risk that goes with that.

Yiwei She: So, for example, serially versus in parallel. So, we learned about another parent’s—Rohan Seth. Seth has also published his journey and we learned that he screened hundreds, up to over thousand ASOs without finding one that was both safe, at least in in vitro applications. So that told us that if we want to do round after round and come back that will take time. The other option is we could spend money upfront and that will cost more, but we could go through as many screenings in one go. And the risk was, the downside was, well, if we were going to find something in the first batch, we wasted, I don’t know, eight ASOs worth of dollars. But the upside is we didn’t take that risk with time. So that’s an example. And I think we applied that thinking overall over and over again.

Daniel Levine: What did it cost to do this and how was the work funded?

Yiwei She: That’s really, really tricky. So, we funded this work out of our personal savings, so it costs well over a million dollars. And I think Julia Vitarello, I think somewhere in a media article, said that if you accounted for all of the donated services for milasen might’ve cost somewhere almost say $25 million dollars. And so the numbers here are extremely large and there’s only one patient in the middle of it. But I think for us it was our personal money and it was our child and we thought we would regret it so much if at the end of the day we had this money and a child who didn’t have an option when this choice was open to us.

Daniel Levine: In 2022, you created the TNPO2 Foundation. What was the vision for the organization?

Yiwei She: I think the vision is to help other families who aren’t as privileged as us, but who are nevertheless facing the same incredibly difficult situation. And I think having done this with Leo, there’s a lot that I think both I and all of our partners have learned. And I think we could do it again for a lot less money and possibly faster and for a child who might get more out of their personalized medicine. And so the vision is that in rare disease, the economics is often, is often the barrier. How can you resource these N-of-1s for every single child who might need one? And I think having gone through this, I come away thinking that we’re closer to the so-called breakeven than many people give it credit for because on one side is a lifetime of caretaking for Leo, and on the other side is a rather elaborate and expensive drug development program. And these two things are not as far away from each other as you might think at first. So I think the vision is, or we started the foundation thinking there must be a better way than to say to the families, go home and love your child, which we absolutely do no matter what happens. But there might be more Leo to love.

Daniel Levine: One of the things the foundation’s working on is trying to accelerate the diagnosis of children with an ultra-rare disease. This is Project Baby Lion, which you’ve teamed up with Rady’s Children Institute for Genomic Medicine and Stonybrook Children’s Hospital. This is based on the work Rady is done on its Project Baby Bear. You’re sponsoring a small pilot. How will that program work?

Yiwei She: Yes. So I want to say it’s not just accelerating diagnoses. One of the most astonishing things that Rady has done is proven that this rapid whole genome sequence is actually economically viable. So there’s a very positive economic ROI, and that’s what the series of studies have shown. Unfortunately, these studies have not led—well they have. They’ve led to the rapid adoption of rapid whole genome sequencing, but it’s still not widely used. It’s still not standard of care. It’s not standard of care at Stony Brook where Leo has been in the ICU of various sorts many times. So, the goal is to make it this economically scalable part, actually scaled. And then the second question we want to answer is what do you tell the families afterwards, after they get a diagnosis, right? Everybody’s left kind of like a deer in the headlights. So, the foundation, its main goal for Project Baby Lion is to go beyond Project Baby Bear is to go beyond the diagnosis, is to walk families or help build a systematic pathway to—is it a repurposed drug? For some people, they’re lucky. They’re lucky that a repurposed drug could almost cure, if not make significant headway on the symptoms of their genetic disease. And why would we turn, would we throw that away? And if a repurposed drug doesn’t work, we would like to find a more personalized approach and develop, if need be, a personalized medicine for them.

Daniel Levine: Well to that end, you’ve announced two collaborations, one with the Hugh Kaul Precision Medicine Institute at the University of Alabama Birmingham. Listeners may be familiar with that program, which is directed by Matt Might who should be familiar to many RareCast listeners. You’re also working with Every Cure, the nonprofit founded by David Feigenbaum to use AI to repurpose FDA approved therapies, another person who will be familiar to our audience. But how did you connect with these two programs?

Yiwei She: We just reached out to Dr. Matthew Might, we had read his blog. We had read what he had written for his son and his journey and how they came to repurpose drugs. So, when faced with the same problems, we learned from them, and his efforts to be very helpful to us and also that they’re a very open collaborative group. And that’s what we want to encourage and that’s who we want to work with. With Dr. Feigenbaum, we actually connected to David through Creyon. I think he advises Creyon and Chris helped us make that introduction.

Daniel Levine: I should mention that both of these efforts involve the use of artificial intelligence. You’re the engineering co-founder of an AI company. I suspect your involvement with the technology made you open to the potential to use AI to help find treatments for patients. What do you see the in potential for AI in the case of helping patients with an ultra-rare disease?

Yiwei She: Yeah, let me take a step back. I was an early engineer at an AI company, and this was a company before the wave, the current AI wave. And our co-founder, I think at the time said, AI can do what humans can do in about five seconds. And so that was kind of the heuristic he gave us as the engineers and the boots on the ground whose job it was to actually provide AI solutions to our partners. So first of all, I want to pull the curtain back on AI a little bit because there’s a whole lot of hype and marketing and branding. And I want to say as a former AI engineer, what is it? What is AI? And I think if folks with computer science and AI background could bear with me for a bit, classically, computers had these very, very strict algorithms. So, in the 1960s through let’s say the late nineties, the database was built, the algorithms to search a database to organize and compute a database. Those algorithms came out and the trees and graphs, those algorithms came out. But they were in some sense very deterministic algorithms. The current generation of AI is in some sense, what’s at the center are these things called matrices. And it’s actually a very simple concept. And what’s at the center is how are two things related to one another? And that is what the matrix encodes. So, if you take a bunch of different concepts or let’s say variables or in the case of healthcare, say risk factors, how does say each gene, how does each mutation affect your genetic outcome? What is your genetic outcome? So that can be encoded in a matrix as can pretty much any computational problem. What’s happened in the last say 10 to 15 years is that the computation to be able to, the amount of computation you can do with these matrices has truly exploded. And then during the last wave, what we could do was, or rather, what AI could do was it could classify images that were long thought to be impossible. Experts in the field thought it was impossible to classify, to use AI to classify, say, a thousand different objects. And that came out in, say 2013, 2014. And since then, the scale has provided its own, let’s say qualitative advancement. And if you take an example, like something that’s relevant to healthcare, say a decision tree, which is how do you manage the clinical care of somebody presenting certain symptoms and you go down the decision tree, you ask different questions to diagnose somebody. And so this thing can be encoded in a matrix. And what we found is that if you just dump all of this data into an AI algorithm and tell it, minimize the error of the data or minimize the error of trying to predict this data, the AI actually learns the logic behind these decision trees. So, you can recreate in some basic sense human intelligence and human decision making with not a whole lot of software engineering. I mean, it’s still, AI talent is still incredibly expensive for now, but it turns out that AI can actually do a lot of this just by feeding it sufficient data and a sufficiently large matrix. And so what does that have to do with rare disease, I guess? And here’s the thing. So healthcare data is incredibly valuable. And in the case of ultra-rare disease, each individual patient’s healthcare data is incredibly, incredibly valuable, both scientifically, both for the therapeutics to prove that something works, but also scientifically as a matter of biology. And that’s really unique because for common diseases, a lot of people have the same biology and have the same underlying symptomatology. And so, what the information you’re getting out of studying them is not all that—it’s not individually very valuable because there’s so much redundancy. Whereas in the case of rare diseases, let’s say there’s only a handful of patients with a given rare disease. Well, you are learning a lot about that particular gene, that particular biological pathway by those diseases and how you might possibly correct them. And so that is where I think the power of AI can really leverage, or rather the rare disease community can really leverage the technological advancement. That’s AI.

Daniel Levine: It’s remarkable to see the progress you’ve made in such a short time with the foundation and to move an experimental therapy to dose your child. What’s the longer term plan though, for the TNPO2 Foundation? Are you doing things to fund it?

Yiwei She: Yeah, so we are fundraising and we are in talks with a few groups. So, if listeners are interested in supporting our work, feel free to reach out. The long-term vision is to, so we’re starting with Project Baby Lion and we’re starting with this ultra rapid, this rapid whole genome sequencing because it’s a sustainable process. And so, what we want to do is push that further beyond the diagnostic odyssey, into the therapeutic odyssey and see how far we can get. So a hypothesis of mine is that drug repurposing is already hugely positive ROI. So, what we can gain by doing—I would love to do a drug repurposing cohort study like Project Baby Bear and see, do we save hospital visits? Do we save clinical, let’s say healthcare resources and the patients benefit, right? Do they have a better clinical outcome? And if we could do that at a cohort level, there’s data that would push that to become standard of care. Imagine repurposing being standard of care so that we can catch everybody for whom a repurposed drug could help. And then much harder is the next step of, let’s say, can we accelerate a pipeline asset that takes more capital, that takes more resources. But Jacifusen showed us that it was worth it. Jacifusen showed us that that asset went ahead into human trials and now maybe ALS, FUS patients will get a lifesaving treatment years ahead of schedule. And then finally, can we start from scratch and design a denovo drug and in time to save the first patient who has that mutation? And I’m saying, why not? Because I was Leo’s mom, right? And if I was Leo’s mom, I was the person who said, I want to save him. I want to save his development. I want him to grow, I want him to grow up. And for these other babies, for these other children, I want them to grow up too, and we can all work together, find a way to do that so that it’s not just sustainable, it’s accessible, and then finally it’s equitable. Because at the end of the day, having a child grow up and have a relatively healthy life, that’s the most fundamental, one of the most fundamental experiences of life, I think.

Daniel Levine: And given the breadth of the mission, is there any thought of changing the name of the organization?

Yiwei She: Yeah, I guess we should probably should. I think when we started the organization, we had it in mind. We wanted to go beyond TNPO2, but we didn’t see what is the next level, what is small enough that it seems approachable but not sound so grand.

Daniel Levine: And at this point, are you expecting to add programs or staff?

Yiwei She: I think that depends on how well we can fundraise. So, we would love to add staff and help and, I think, only if we’re making a difference. So, if we’re making a positive difference and what we’re doing is helping people, then we would like to do more of it, and that we’ll see if funders like what we’re doing. So, if we get a vote of confidence for funders, we’ll definitely add staff.

Daniel Levine: Yiwei She, founder and CEO of the TNPO2 Foundation. Yiwei, thanks so much for your time today.

Yiwei She: Thank you.

This transcript has been edited for clarity and readability.

The RARECast podcast is made possible through support from the Global Genes’ Corporate Alliance. The members of the Corporate Alliance support Global Genes’ mission and programs, work to meet the vital needs of people with rare diseases, and address inequities they face. To learn more about the Corporate Alliance or how your organization can become a member, click here.