Thank you to everyone who tuned into Decoding Bio’s first Substack Live. A true experiment for us. If you weren’t able to join, we have the recording and the full interview below.
Today, Flagship Pioneering launched Serif Biomedicines, a new company building Modified DNA as a therapeutic modality. We sat down with co-founder and CEO Jacob Rubens to talk about why DNA has been so hard to turn into a medicine, what Serif has built over five years in stealth, and what the early data shows in primates. The company plans to present preclinical data at an upcoming scientific conference, with a peer-reviewed publication to follow later this year. For more information, visit serifbiomedicines.com.
Over the past several decades, each time a layer of biology has become engineerable, it has given rise to enduring companies and transformative medicines. Proteins gave us Genentech and Amgen. RNA gave us Alnylam and Moderna. Genome editing tools like CRISPR have opened up a new frontier for permanently rewriting the genetic code. But DNA itself, delivered as a non-viral medicine that expresses genes durably without altering the genome, has remained out of reach.
Serif Biomedicines, launched today out of Flagship Pioneering after five years in stealth, is betting that this is about to change. The company has developed what it calls Modified DNA: chemically altered DNA molecules that can evade the immune system, reach the nucleus, and durably express therapeutic proteins without integrating into the genome. If it works, Modified DNA would combine the durability of gene therapy with the redosability and scalability of mRNA, while avoiding the limitations of both.
In this conversation, recorded as Decoding Bio's first Substack Live, we speak with Jake Rubens, co-founder and CEO of Serif Biomedicines and Origination Partner at Flagship Pioneering, about why DNA has been so hard to turn into a medicine, how Serif's approach works from first principles, what the early data shows, and where the company is headed.
Our Guest
Jake Rubens is a scientist-entrepreneur who has spent his career at Flagship Pioneering founding and building biotechnology companies around new platform technologies. He co-founded Quotient Therapeutics, Tessera Therapeutics and Sana Biotechnology, before starting Serif in 2021. He received his PhD from MIT, where he worked in Tim Lu’s Synthetic Biology Center on engineered gene circuits for intelligent cell therapies.
A Conversation with Jake Rubens
Zahra: Congratulations on the launch. For people just tuning in, can you give us the short version of what Serif is and what you’re building?
Jake Rubens: Serif is a new biotechnology company, and what we’re seeking to do is make DNA, the most fundamental information layer of biology, into a biotechnology for the very first time.
Zahra: We already have mRNA drugs, gene therapies, gene editing. What’s missing? Why DNA, and why is this the molecule you want to turn into a medicine?
Jake Rubens: The bigger picture here, the backdrop upon which we started Serif, is that at Flagship we’ve had a front row seat to innovations in biotechnology for the past 25 years. But even before that, we saw how when a layer of biology became an engineerable biotechnology for the first time, it gave rise to countless drugs that helped many patients and led to the generation of enduring companies. When we think about proteins, we think about Genentech, Amgen, and more recently Regeneron. When we think about RNA, we think about Alnylam and Moderna. That is the lineage in which we hope to be at Serif, but now focused upon the fundamental information layer in biology, and that’s DNA.
So what does DNA provide? It provides perhaps the very best of mRNA and the best of gene therapy while mitigating some of their limitations. mRNA medicines are incredible because they are scalable to manufacture and redosable. Some of the limitations can be that the half-life of RNA is quite short, and the cell specificity of expression is hard to tune in because, as a messenger molecule, mRNA is translated into nearly every cell that receives it.
On the other hand, gene therapy, especially viral-based gene therapy, we’ve seen striking cures in patients. That includes work like CAR-T, especially ex vivo CAR-T, and gene therapies like for spinal muscular atrophy. But gene therapies are unbelievably challenging and costly, slow to manufacture, and they can’t be redosed to patients, which limits their applicability and their accessibility for the world.
We believe that a DNA medicine can be durable, so longer half-life than mRNA. It can be redosable, unlike gene therapies. It can be cell-specific in expression, much like viral-based gene therapies can be. And it can be scalable, much like mRNA. That’s why we set out to work on DNA.
And to be fair, we weren’t the first people to realize this. This has been a dream for a very long time, arguably since we first elucidated that DNA was the fundamental information molecule in biology. Scientists have wanted to turn it into a medicine, and that’s a hard problem.
Zahra: If it’s so compelling, and people have tried to crack it, why has it been so hard to make this work as a medicine?
Jake Rubens: Back in 2021 when we started the company, we identified two fundamental challenges. And we’re bringing the company out of stealth mode today, unveiling Serif, because we think we’re making progress. We’re not going to declare success until we’re treating patients, hopefully curing patients safely. But the evidence tells us we’re on the right track.
Those two limitations are, first, DNA is highly immunogenic. And second, DNA has a delivery challenge: it has to reach the nucleus of the cell.
On the immunogenicity side, the most ancient part of our immune system, the innate immune system, recognizes DNA when it enters into the cell. The cell thinks it’s being invaded by foreign bacteria or a foreign virus. It sets off alarm bells.
So we began working on solutions to this problem inspired by what worked for mRNA and what worked for siRNA. And that was chemical changes and structural changes to the molecules themselves that permitted them to avoid innate immune detection, have greater stability, and still maintain their coding properties. We found analogous solutions in the realm of DNA after a ton of work.
The second problem, accessing the nucleus, has held back many approaches to turn DNA into a medicine in the past. Here, the solution that we brought is actually inspired by mRNA again. We’re leveraging mRNA to deliver proteins which are expressed transiently, which grab the DNA and help it access the nucleus and localize within the nucleus.
So these two solutions, the modifications (or what we call Modified DNA) and the cofactor, come together to bring the potential solution to the challenges that have held back DNA medicines in the past.
Zahra: Can you paint a step-by-step picture of how a Serif DNA therapeutic would work, from when it’s administered to the patient all the way through to making a protein?
Jake Rubens: If you’ll permit a bit of hand-waving, I’m happy to take a stab at that. We deliver our DNA inside of what are known as lipid nanoparticles. These are fat bubbles. And to this point, they are science’s best solution for delivering complex nucleic acid molecules like mRNA and DNA into cells. There’s been great progress in the field in developing lipid nanoparticles that can ferry nucleic acids to cells like hepatocytes in the liver, to various types of immune cells, T cells especially, and to hematopoietic stem cells.
Once our nucleic acid, the DNA and the mRNA that encodes the cofactor, is delivered into the cytoplasm, that cofactor molecule is translated. It makes a protein. The protein then grabs onto the DNA, helps it access the nucleus. And what’s inside the nucleus, the DNA actually changes form. It is demodified. It goes from being a modified DNA molecule back to a natural DNA molecule. And that process is carried out by endogenous proteins and mechanisms that are present in every nucleus of every cell in our body.
After that, the DNA is transcribed to make mRNA. Then that mRNA is translated into a therapeutic protein, if that’s the ultimate payload.
Importantly, we don’t expect the DNA to integrate into the genome. It doesn’t have to integrate into the genome for its function. We are trying to minimize that possibility. So it exists as an episome, a separate DNA molecule from the rest of our chromosomes. That’s important from a safety perspective. It’s also what we think can be a great feature because it’s unlikely to be permanent forever. We think that our medicines will be re-administered into patients, which is something that we view as a feature rather than a bug for our technology.
Zahra: Does using lipid nanoparticles put any restrictions on DNA size or add to the immunogenicity? An LNP carrier isn’t invisible to the innate immune system either.
Jake Rubens: There are some absolutely seminal studies from the last five years that show that the immunogenicity of a nucleic acid is not determined just by the nucleic acid itself, but how it’s delivered into a cell. Certain lipid nanoparticles are more immunogenic than others. So this is something that we spent a lot of time thinking about. And it may be the case that the best lipid nanoparticle for delivering mRNA may not be the best one for delivering DNA. So there’s a lot of opportunity for engineering in that realm.
Zahra: Talk to me about how you found and screened for the right modifications and the demodification steps. How did that work?
Jake Rubens: A lot of brute force, a lot of effort. We began by testing different forms and structures of DNA. Think about linear DNA versus circular DNA, open-ended, closed-ended DNA, double-stranded versus single-stranded. And then we layered on top of that different potential chemical modifications: backbone changes, base changes, and then testing those in combination and in different positions. You can see how quickly this multi-parametric space scales into the thousands.
We kind of systematically went through all of the space to find modifications that permitted us to have minimal innate immune stimulation while maintaining as much gene expression as we could. And you probably won’t be surprised to hear that we found many ways to make our DNA immune silent. Unfortunately, most of those were also expression silent. So the trick was finding modification approaches that permit high expression and minimal immunogenicity.
We’ll share a lot more about that specifically at a conference in the not-too-distant future, and subsequently, we hope, in a peer-reviewed publication.
Zahra: Is there cell-to-cell variability in how the demodification step works? And with the episomal DNA, does it dilute as the cell divides?
Jake Rubens: We’ve tested a variety of cofactors and we have some that we find are working pretty well consistently across cell types. So I don’t want to rule out that there’s cell-to-cell variability, but so far our favorite approaches work best across cell types.
Your second question is a great one. What you’re getting at in asking about dilution is: what is the durability? I don’t want to steal our team’s thunder ahead of the scientific conference we’ll be at soon. But we see quite striking durability in vivo in non-dividing cell types like hepatocytes in the liver. And we see less durability in dividing cell types like T cells. That’s not surprising because what we expect is happening is that the DNA is being diluted as T cells expand and divide.
But that’s not necessarily a bad thing. In fact, for certain applications, and I’ll just talk about in vivo CAR-T for a second, it may be that the best solution is not the short-lived durability that mRNA provides or the permanence that genome engineering or gene therapy provides, but rather a Goldilocks zone of expression where the drug lasts and T cells can be engineered to express a CAR for a meaningful period of time, sufficient to reset the immune system or target various cancer cells. And if necessary, have the flexibility to re-administer the drug to the patient at a time where the pharmacology is suitable, such that it’s safe and can still be potent. That’s the way we think about the durability of the drug, and we think it provides us great flexibility.
Zahra: What are some key factors you use to pick the ideal targets for Serif? How do you decide where Modified DNA wins?
Jake Rubens: There are probably two or three main criteria. First, we want to choose technical problems that we’re confident are solvable. So we aren’t starting with delivering Modified DNA to the central nervous system. That’s not because there isn’t a need there, but delivery is very challenging there. So we focus on: where are we confident we can develop or access lipid nanoparticles that will deliver to target cells? That’s the first question.
The second question is, of course, is there unmet need? And then if so, where are we confident that the target gives an extremely high likelihood of curing the disease definitively if we can hit our technical goals? So that’s another way to think about biological risk.
And the third is differentiation. Where do we believe that Modified DNA will actually capture the win state for a target and create best-in-class products, superior to what’s possible with mRNA or with other approaches to genetic medicine like gene therapy or genome engineering?
Zahra: Can you walk us through where Serif is starting in terms of programs?
Jake Rubens: We haven’t shared too much on exactly what we’re pursuing yet, but we are excited about delivering Modified DNA molecules to the liver to address a variety of different diseases where we’re quite confident that the protein we express can address that disease, whether it’s a liver disease or otherwise. And the other is around immune cell programming, where we see that there’s probably still a tremendous amount of opportunity, following in the incredible pioneering footsteps of other companies that have gone after in vivo CAR-T, in vivo macrophage engineering, in vivo CAR-NK engineering, and all these other cell types.
Zahra: You’ve been in stealth for five years. At a high level, what have you seen so far?
Jake Rubens: What gives us confidence to bring the company out of stealth mode today, five years after we started, is we wanted to see some proof points ourselves that gave us reason to believe. And where we felt that it was time to get feedback from the broader scientific community. So we’ll be sharing some of our work at a conference soon.
What I can share is that we’ve demonstrated that our Modified DNA is safe in monkeys. We’ve demonstrated that our Modified DNA drives durable expression, dramatically outperforming mRNA-based approaches inside of rodent models at levels that give us conviction that we’re going to be able to make drugs, or at least have a good shot at making drugs. These are the things that lead us to want to bring the company out of stealth mode today.
Zahra: You mentioned Modified DNA is the core invention, but it sounds like you need a broader system around it. Can you summarize what else you’re building, and where AI might help?
Jake Rubens: This is hard. The Modified DNA and the cofactors that address what we believe to be the two fundamental limitations of DNA medicines, it’s not enough to make a medicine. We also need to optimize the DNA sequence of those medicines, we need to optimize lipid nanoparticles, and we need to be able to manufacture.
Since the beginning of Serif, we have been putting in place capabilities that address all five parts of that stack: the modifications, the cofactors, the DNA sequence engineering, the LNPs, and the manufacturing.
AI has been a great tool for us. I wouldn’t call us an AI-native biotech company necessarily, compared to much of what you otherwise cover, Zahra. But we recognize the transformative potential of AI as a tool. So we leverage AI heavily for DNA sequence engineering and lipid nanoparticle design and screening.
We’ve developed in vivo screens where we can administer to a single animal a barcoded library of 10,000 different combinations of promoters and enhancers. By sequencing the RNA in target and off-target organs, we can determine which of those promoters and enhancers had the best expression or the highest specificity of expression in the cell types that we’re seeking to deliver to. That creates a pretty rich data set that can be combined with everything else that’s known publicly to learn and design the next round of DNA sequences. We take an analogous approach for discovering and optimizing lipid nanoparticles, although the throughput isn’t quite as high due to limitations in lipid nanoparticle manufacturing.
Zahra: How many programs are you hoping to develop? What is the $50 million going toward?
Jake Rubens: We hope over time to have many, many drugs at Serif. For now, that’s a dream. Today we have zero drugs at Serif. But we wouldn’t be doing this if we didn’t think that we could get there.
We think we’re going to get there fastest from the help of partners. When we look at the history of some of the enduring companies I mentioned at the outset, Genentech, Regeneron, Moderna, Alnylam, one of the common threads through all of them was the role of strategic partners. Partners that gave those small startups access to capital and know-how that helped them build their platform, figure stuff out, and bring their first drugs into patients. We think that strategy is a sound one for us to pursue at Serif as well. I hope that in the next year or more, we’ll be announcing some of those partnerships.
Zahra: What are some of the biggest unresolved technological challenges that worry you?
Jake Rubens: Look, there are a variety of unknowns. That’s why we don’t want to overhype the progress we’ve made, because we may find out that there’s not just two problems we need to solve, but four. If that were the case, it would be probably impossible to foresee today. But science is, of course, littered with these types of stories.
I don’t worry so much about the known unknowns. It’s more about the unknown unknowns that keep me up at night, and I think should keep any developer of a new biotechnology up at night.
Zahra: But if you had to point to one part of the approach, whether it’s LNP delivery, getting into the nucleus, the demodification, stability, payload size, where would you say you really have to prove things out?
Jake Rubens: Delivery. Delivery is always the answer. The reason is just the throughput of experiments around delivery is fundamentally slower because oftentimes they have to be carried out inside of non-human primates to draw strong conclusions. So it’s probably delivery, but we’re aware of this and we’re putting a lot of resources on it for that reason.
Zahra: The competitive pressure from China keeps coming up across the industry. How does that factor into how you’re building Serif?
Jake Rubens: A little bit. The innovation coming out of China is obviously remarkable. People oftentimes reference how companies in China are developing me-too molecules, like small molecules and biologics targeting known targets. But I’m not sure that’s right. We’ve seen a lot of competition emerge from China on complex genetic medicines, both mRNA and especially in the in vivo CAR-T space.
So I don’t think about competition from China much differently than how I think about competition in the United States and in Europe. We have to win, and we’re going to do our best to win.
In many ways, though, I like competition. Competition is oftentimes the first validation in science. It’s a lot scarier to be out there being the only one doing something than to see someone else pursuing the same hard problem and maybe even arriving at similar conclusions. That’s validating. So we welcome competition. We think it’ll make us better.
Zahra: The genetic medicine field has also changed quite a bit over the past years, with immense progress but also some big setbacks. Does that shape your view of how Modified DNA fits in?
Jake Rubens: We’re always aware of what else is going on in the field. If anything, we’ve been quite excited by the incredible momentum we’ve seen in genetic medicines over the past couple of years.
Just to highlight two of the most exciting stories in all of medicine: in vivo CAR-T and cancer vaccines. In in vivo CAR-T, we’ve seen striking impacts in both oncology settings and autoimmunity, immune reset settings. Data is early, but we’ve seen pharma companies put out $7 billion deals, $2 billion deals, $1.5 billion deals, all based upon just hints of activity inside of non-human primates and some safety data. That’s very exciting to us.
And in cancer vaccines, I think the world hasn’t yet quite woken up to the potential for using information molecules to program our immune system to seek out and eliminate cancer. Deservedly so, over the last week, Revolution Medicines has received a ton of attention for their KRAS drugs because they’re treating patients who otherwise didn’t have much hope. But what hasn’t received enough attention was the data that came out of BioNTech’s personalized cancer vaccine clinical trial with Memorial Sloan Kettering, where they saw something like seven out of 16 patients with survival up to four to six years in pancreatic cancer. That is remarkable. These are both just small hints of what’s going to be possible and what is likely to be the future of many other areas of medicine, and that’s information molecules.
Zahra: Serif started five years ago. What did it look like back then? I can’t imagine it was a linear path.
Jake Rubens: It never is in science. We started the company with a couple of my co-founders, Eric Keen, Louisa Helms, Geoff von Maltzahn, Noubar Afeyan, and an idea. Flagship started what we call a protocol to get the company off the ground. We saw some signs of progress and felt that the plans were concrete enough that we wanted to wade further into the uncertainty and see whether we could actually develop Modified DNA as a medicine.
I’m unbelievably grateful to the support afforded to us by Flagship and the belief and just dedicated hard work by our team at Serif. We’ve built capabilities across five different important aspects of developing a DNA medicine. We’ve done that with a small team in a short period of time. We’ve worked really hard. We’ve run into some incredible roadblocks and found ways to overcome them. It’s a testament to the dedication of that group that we are where we are today.
Zahra: Was there a data point or a trigger where everyone thought, this is the moment, this is where we need to scale up?
Jake Rubens: I could think of two, and these both came in within the last couple of years.
One was the first results from monkeys showing that unmodified DNA was very poorly tolerated, but Modified DNA was perfectly well tolerated. The difference in the clinical signs in these animals was just night and day. That told us that the rodent models and the human cell models actually accurately model primates. That was big.
The second was the first in vivo experiments with our cofactor. That’s our lead one, where we saw four orders of magnitude increase in expression. I recall specifically bursting into a meeting just to validate that the numbers they sent me were correct because they were so mind-boggling. That’s what a breakthrough looks like. You don’t get many of those moments in your career. Both felt really special.
Zahra: Your tagline is “Writing the Code of Life in Distinctive Fonts.” What does that mean to you?
Jake Rubens: We love our name. We’re obviously biased, but we think it’s a really nice analogy for what we do. If you think of unmodified DNA, natural DNA, as sans serif, we add serifs to it. We change the font slightly of the code of life so that the words are still the same, but the meaning is a little bit different. The interpretation is a little bit different, just like changing fonts changes the way that we interpret the written word.
Zahra: Do you have any advice for aspiring biotech builders, people who want to change the game in how patients are treated?
Jake Rubens: Just do it. What I mean is, take a shot. Hard things are hard. And you got to try them to see how hard they truly are. I think sometimes you’ll find that what you think is hard is just actually uncertain. And sometimes it’s easier when you start working on it. But you won’t know unless you just do it, unless you try.
Zahra: Thank you Jake for this wonderful introduction to Serif. We are excited to follow along as the company progresses.
