Eric Kelsic, PhD, has spent years thinking about how to target the brain for therapy. Through his career, Kelsic has seen central nervous system (CNS) programs rely heavily on brute force: local CNS injections offer limited benefit, are invasive and unscalable, and systemic adeno-associated virus (AAV) delivery largely fails due to blood-brain barrier (BBB) exclusion, liver sequestration, and, ultimately, toxicity. To Kelsic and others, the bottleneck is clear—get through the BBB and keep clear of the liver.
Many companies have seeded with the mission to engineer AAVs to get through the BBB, including Dyno Therapeutics—Dyno Tx for short—co-founded in 2018 by Kelsic and Adrian Veres, PhD, Sam Sinai, PhD, George Church, PhD, and Alan Crane, PhD.
What sets Dyno Tx apart in the CNS delivery field has been intentionally veering away from screening randomly for a capsid that slips through the BBB. From the beginning, Kelsic and Dyno Tx aimed to carefully develop viruses using machine learning, structural biology, and high-performance computing to answer specific questions: how do capsids attach to BBB receptors, how does that attachment affect their movement through the BBB, and how do surface features impact immune recognition and targeting the liver?
Enabled by a collaboration with NVIDIA, which provides the computational power needed for structure-based capsid design, Dyno Tx can model receptor binding and immune evasion simultaneously, accelerating progress in ways that were not possible even a few years ago. In 2024, Dyno Tx’s approach of fast-forwarding viral capsid evolution to exploit different endogenous transport mechanisms used by the brain itself began to bear fruit, most notably a series of licensing deals with Roche totaling over $3 billion. But the CNS gene therapy field was stopped cold in its tracks in September 2025 when a patient died shortly after being treated with a novel engineered BBB-crossing AAV.
This tragedy brought to light the second pillar of Kelsic’s thesis behind Dyno Tx: CNS gene therapy will not converge on a single universal delivery solution. So, rather than relying on a single solution, Dyno Tx is developing a portfolio of capsids that preferentially cross the BBB through distinct biological mechanisms, each with different tradeoffs and translational risks.
“There’s a few different mechanisms that are promising,” Keslic told Inside Precision Medicine. “They’re different in their expression patterns across regions of the brain. They’re different in the cell types that are expressed there or which region is impacted for a certain indication so it might be that the way we end up is actually not one mechanism that’s the best for everything but different capsids that are the best fit for any specific for different CNS indications.”
The most recent addition to this portfolio is Dyno-yp2, unveiled today at the opening of the 2026 J.P. Morgan Healthcare Conference. Engineered to cross the blood–brain barrier (BBB) via the transferrin receptor (TfR) pathway, Dyno-yp2 further expands Dyno’s growing lineup of off-the-shelf AAV capsids. This portfolio also includes several capsids introduced in 2025: Dyno-bn8 (muscle), Dyno-3hv (neuromuscular), Dyno-4z2 (ocular), and Dyno-ahq (CNS). Together, these offerings position Dyno as the only gene therapy company providing CNS delivery vectors that leverage multiple BBB-crossing mechanisms, giving partners the flexibility to choose the most effective delivery strategy for specific payloads and indications.
Dyno-yp2 joins Kelsic, with modesty matched only by his ever-growing sage-like beard (akin to his former advisor George Church), said, “Very soon, there’s going to be a safe demonstration of CNS delivery, crossing the bloodstream barrier in human patients—the only question is really which mechanism is going to be the first.”
The end of brute-force delivery
Historically, systemic delivery meant overwhelming the liver to achieve modest brain exposure. According to Kelsic, only a tiny fraction was going to the brain, forcing developers to increase doses until “you ended up overloading the liver, which caused liver damage or, in some cases, patient death.” Dyno’s recent capsids suggest that tradeoff is no longer inevitable.
One of Dyno Tx’s capsid features explored with ML-guided directed AAV surface evolution deals with steering clear of the liver and, instead, preferentially targeting the brain. Shifting this targeting preference lowers required doses, reduces immune risk, and improves manufacturability—conditions that must be met if CNS gene therapies are ever to move beyond rare, high-risk use cases. What was once a narrow path has widened into a frontier. Several BBB-crossing mechanisms now show strong performance in non-human primates (NHPs) and appear promising for human translation.
Engineered to cross the BBB via the TfR, Dyno-yp2 builds on a pathway long recognized as biologically plausible but technically difficult. The TfR plays a critical role in iron transport and is expressed on brain endothelial cells—making it an attractive gateway if it can be engaged safely and efficiently. Yet even the most compelling mechanisms, Kelsic acknowledged, still leave open the question of how well all these are eventually going to work in human patients.
Because the human TfR differs subtly from its NHP counterpart, Dyno Tx adjusted its validation strategy, relying on humanized mouse models rather than primates to directly assess binding and transduction. This allowed the team to evaluate not only efficiency but also dosing and off-target risk with a clearer line of sight to human translation.
Head-to-head comparisons against leading external TfR-based capsids revealed a key advantage. Dyno-yp2 delivered higher CNS transduction while dramatically reducing liver exposure. Compared with AAV9, liver targeting dropped roughly 29-fold, and relative to a benchmark TfR capsid, by approximately 80-fold. As Kelsic puts it, the significance lies in the combination, because “not only has the CNS efficiency increased, but we’ve also improved the specificity for the CNS.” That specificity addresses one of the most persistent safety challenges in CNS gene therapy.
“We’re really optimistic about the potential for highly specific, low-dose, highly effective gene therapies through the CNS,” said Kelsic. “We know now that these work in primates, and so the question is what’s going to translate from those models into the human setting.”
Strategic delivery agility
Despite these advances, Dyno does not present Dyno-yp2—or any single capsid—as the final answer. Different diseases involve different regions of the brain, different cell types, and different therapeutic thresholds. As Kelsic noted, “it might be that not one mechanism is the best for everything, but different capsids are the best fit for different CNS indications.” Ultimately, therapeutic success depends on reaching “the right cells and the right regions of the brain.” By continuing to invest across multiple mechanisms—including TfR-based capsids like Dyno-yp2 and other undisclosed pathways—Dyno is deliberately avoiding what Kelsic describes as getting “stuck in a local minimum and miss[ing] out on the full potential of CNS gene therapy.”
Kelsic’s goal of engineering viruses that fit form to function extends beyond the work within Dyno Tx, evidenced by its broader ecosystem strategy, particularly its Frontiers program. Through this initiative, Dyno pairs advanced capsids with payload developers working on highly innovative approaches such as engineered promoters for cell-type-specific expression.
“It’s basically like a ‘push button’ for the developer to come in and say, ‘We’ve got our payload, and we want this much with this capsid,’” said Kelsic. “The Dyno Frontiers manufacturing partners are ready to go with that right away. They get the data to show that they’re highly competent. So unless there’s something different about the developer payload, it’s the most effective way to quickly generate promising preclinical NHP data.”
One example Kelsic highlights is Epicure, a company spun out of the Allen Institute that has published extensively on promoter design and joined Dyno Frontiers to combine its payload technology with Dyno capsids. Dyno’s aim is to ensure that partners are not constrained by early delivery decisions that limit that reach. Rather than forcing partners to commit early, Dyno Tx aims to offer access to capsids spanning multiple high-potential mechanisms within a single partnership.
“I’m really optimistic about CNS gene therapy and the gene therapy field in general,” said Kelsic. “If a few of these pieces come together—many of the pieces are actually out there in different companies—the problem is the integration of them into great products and getting them cost-effectively into the clinic.”
This approach, which Dyno Tx calls strategic delivery agility, is designed to let developers adapt as new data emerges. In an environment where timelines and capital efficiency matter as much as biology, that flexibility can determine whether a program advances or stalls.
“Strategic delivery agility is the ability to switch as information comes along without having to wait probably a year or more to see if you’re going to a completely new mechanism,” said Kelsic. “The investment that requires is going to be challenging for companies who really want to make wise use of investor funds as well as make as many therapies with every dollar they spend as they possibly can. Doing this in a way that minimizes the opportunity cost and enables them to be first to market with the best-in-class drug, all with capital efficiency.”
Genetic agency, patient imperative
Underlying Dyno’s technical strategy is a patient-centered philosophy Kelsic calls genetic agency—the belief that patients should have access to meaningful choices, even when risks remain. Reflecting on recent gene therapy controversies, he has argued that public discourse often overlooks the reality that “in many cases, even if there are some risks, patients want that choice.”
For patients with progressive CNS diseases, waiting for a perfect cure may mean waiting forever. Delivery agility, in this context, becomes not just a technical advantage, but a moral one. “The patient need for more options was being overlooked in the search for a perfect cure that was probably going to take many more years in a lot of cases,” said Kelsic. “If we don’t make those investments now, then we’re never going to get there.”
Looking ahead, Dyno’s ambition is to make delivery a solved problem—ready when patients need it. Kelsic envisions a future in which a capsid is already validated, already manufacturable, and ready to be paired with programmable payloads such as CRISPR-based editors.
“We’re working toward having a capsid ready to deploy as soon as a patient needs it,” said Kelsic. “For example, if a child is born and begins to show symptoms, the parents take them to the doctor, the gene is sequenced, and a causative mutation is identified—there would be a fast path to saying, ‘Here’s a gene therapy that could help.’”
Kelsic continued, “This would be possible even when a therapy hasn’t already been developed for that specific mutation. The goal is very rapid development of new gene therapies, enabled by a platform approach on the payload side—using programmable technologies like CRISPR-based genome editing, base editing, and other genome perturbation tools. That’s particularly promising because, in many cases, you only need to make a small genetic change, while reusing the same underlying components.
Ideally, this would be paired with a capsid that can reach the target cells with high efficiency and strong specificity. That’s the ideal we’re working toward.”
The new CNS gene therapy era
In the end, Dyno-yp2 is not positioned as a singular breakthrough but as another step toward a platform designed for uncertainty. In a field where human data is still emerging and biological complexity resists simplification, adaptability may prove more valuable than any single technical win.
As CNS gene therapy moves closer to demonstrating safe, systemic BBB crossing in humans, the companies best positioned to succeed may not be those that guessed right earliest, but those that built the flexibility to learn, adapt, and change course. Dyno is betting that strategic delivery agility—rooted in respect for biology rather than force against it—will define the next chapter of CNS therapeutics.
