We don't know how it works, but the mice got better
At Yale, two neuroscientists have secured $40 million in federal funding to pursue one of medicine's most elusive goals: delivering gene-editing tools across the brain's formidable biological defenses. Their platform, called STEP, has shown genuine promise in mouse models of devastating neurological diseases, yet the researchers themselves cannot fully explain why it works — a gap that sits at the center of both the technology's potential and its peril. Science has always advanced through productive uncertainty, but when financial interests, undisclosed conflicts, and a 2028 data-sharing deadline converge around an unproven mechanism, the distance between hope and proof becomes a moral as well as scientific question.
- Families living with Angelman syndrome, Rett syndrome, and spinal muscular atrophy are watching a technology that could rewrite their futures — but the researchers behind it cannot yet explain the mechanism that makes it work.
- STEP's 10-nanometer delivery particles outperformed expectations in mice, distributing gene-editing tools throughout entire brains and producing measurable improvements in learning, motor function, and seizure frequency.
- Serious red flags shadow the results: active Cas9 appears to cause chromosomal damage that would likely disqualify the therapy for human use, and no one has tested whether the approach scales to the vastly larger brains of primates.
- A lead researcher holds an undisclosed equity stake in the company commercializing the technology, while $40 million in NIH funding effectively shields early investors from risk — a conflict-of-interest arrangement Yale manages privately.
- The field cannot independently evaluate the technology because peer-reviewed data does not yet exist, no head-to-head comparisons with existing methods have been published, and the team is not required to share technical details until 2028.
Three years ago, Yale neuroscientists Yong-Hui Jiang and Jiangbing Zhou secured a $40 million NIH commitment around a striking claim: they had built a vehicle capable of ferrying gene-editing tools deep into the brain, potentially unlocking treatments for conditions like Angelman syndrome, Rett syndrome, and spinal muscular atrophy. The announcement was bold. The underlying problem it addressed was real. And the gap between those two facts has defined the story ever since.
The blood-brain barrier has frustrated neuroscience for decades. Engineered viruses trigger immune responses that block repeat dosing. Nanoparticles reach only a fraction of brain cells. Direct injections require invasive procedures. "Delivery is the number one thing holding us back," says UC Berkeley bioengineering professor Niren Murthy. STEP's proposed solution is radical simplicity: package the gene-editing machinery into particles just 10 nanometers across — roughly half the size of conventional viral vectors — and inject them into spinal fluid or brain ventricles. In mouse models of Angelman syndrome, the approach worked. Learning improved. Seizures decreased. Genetic changes persisted for months. "Clearly they're making stuff that works," Murthy acknowledged.
But the mechanism remains opaque. The STEP molecule resembles cholesterol in structure, according to patent filings, yet how it enters cells is unknown. Jiang said plainly: "We don't know right now." That admission carries weight, because what works in a mouse brain may not translate to the far larger and more complex terrain of a primate brain — a gap that has swallowed many promising therapies before. More immediately, the version tested in mice uses active Cas9, which appears to cause significant chromosomal rearrangements — the kind of genomic damage that would likely disqualify it as a human therapy.
Financial entanglements add another layer of complexity. Jiang holds an equity stake in Couragene, the company that licensed the STEP patent from Yale in 2022 and has since raised seed funding for Angelman syndrome treatments. The NIH's high-risk, high-reward funding structure effectively de-risks the technology for private investors — a dynamic one medical ethicist calls a "potentially high-stakes conflict of interest." Yale has filed a conflict-of-interest management plan with the NIH, but its contents remain private.
The field is left waiting. No peer-reviewed publication exists. No head-to-head comparison with existing delivery methods has been conducted. The team is not required to share technical data until their funding ends in 2028. For the patients whose diseases this technology might one day treat, and for the researchers who would need to build on it, the distance between a promising mouse study and a proven human therapy has never felt longer — or more consequential.
Three years ago, Yale University neuroscientists Yong-Hui Jiang and Jiangbing Zhou walked into a $40 million commitment from the National Institutes of Health with a bold claim: they had developed a way to shuttle gene-editing tools deep into the brain, potentially opening treatment pathways for thousands of genetic neurological diseases. The announcement promised to revolutionize how researchers might one day treat conditions like Angelman syndrome, spinal muscular atrophy, and Rett syndrome. But as the field waits to see whether the technology actually works in humans, a more fundamental problem persists: even the researchers themselves cannot fully explain how it works.
The core challenge has haunted neuroscience for decades. The brain is sealed off by the blood-brain barrier, a biological fortress of tightly packed cells designed to keep pathogens out. Getting large molecules—like the gene-editing machinery needed to fix broken genes—across that barrier has proven extraordinarily difficult. Researchers have tried engineered viruses, nanoparticles, direct injections into spinal fluid. Each approach carries trade-offs: viruses trigger immune responses that block repeat treatments; nanoparticles deliver to only a small fraction of brain cells; direct injections require invasive procedures. "Delivery is the number one thing holding us back," says Niren Murthy, a bioengineering professor at UC Berkeley.
Jiang and Zhou's technology, called STEP (Stimuli-responsive Traceless Engineering Platform), takes a different approach. The key innovation is size. Their delivery vehicle packages gene-editing machinery into particles just 10 nanometers across—roughly half the size of the viral vectors that have dominated the field. In mouse studies published as a preprint in November 2025, the researchers demonstrated that STEP could distribute gene-editing tools throughout the entire brain when injected into the spinal fluid or brain ventricles. Mice with Angelman syndrome showed improved learning, better motor function, and fewer seizures after treatment. The genetic changes persisted for months. "Clearly they're making stuff that works," Murthy observed.
Yet the mechanism remains opaque. The STEP molecule itself resembles cholesterol, according to patent filings, but how it actually enters cells is unknown. Once inside, a linker breaks apart and releases the gene-editing machinery, which then migrates to the nucleus and cuts DNA. Jiang acknowledged the gap directly: "We don't know right now." This absence of mechanistic understanding creates genuine uncertainty about whether the technology will function the same way in larger brains, or in humans at all.
Other questions loom larger. One involves collateral damage. The version of STEP tested in mice uses active Cas9, the cutting enzyme, which appears to cause significant chromosomal rearrangements and large deletions in the genome—damage that would likely disqualify it as a therapy. Another involves scale. Gene-editing approaches that looked promising in mice have repeatedly failed when tested in nonhuman primates, simply because primate brains are vastly larger and the delivery becomes exponentially harder. The researchers claim STEP might eventually cross the blood-brain barrier without invasive injection, which would be transformative, but they have not yet published data supporting that claim.
Meanwhile, financial interests complicate the picture. Jiang holds an undisclosed equity stake in Couragene, a company that licensed the STEP patent from Yale in 2022. The company has raised $1 million in seed funding and launched a joint venture focused on Angelman syndrome treatments. The $40 million NIH grant, structured through the agency's high-risk, high-reward Common Fund program, effectively de-risks the technology for investors—a dynamic that Matthew McCoy, an assistant professor of medical ethics at the University of Pennsylvania, describes as "potentially high-stakes conflict of interest." Yale has filed a conflict-of-interest management plan with the NIH, but the university does not publicly disclose its contents, citing sensitivity around financial details.
The field is waiting. Jiang and Zhou are not required to share technical details until 2028, when their NIH funding ends. They say a manuscript is under review at an academic journal, but no publication date has been set. Mark Zylka, a cell biology professor at the University of North Carolina, noted that the researchers have not conducted head-to-head comparisons between STEP and existing viral delivery methods—the kind of rigorous side-by-side testing that would clarify whether STEP represents a genuine advance. "I'd be interested in trying it," Zylka said, but the data needed to make that decision simply does not exist yet. The real test, researchers agree, will come only when the technology moves into human trials—a step that remains years away and whose outcome no one can predict.
Citas Notables
Delivery is the number one thing holding us back— Niren Murthy, UC Berkeley bioengineering professor
We don't know right now— Yong-Hui Jiang, Yale neuroscientist, on how STEP's mechanism works
Unfortunately, there is no way to know if it's going to work in humans unless you do it in humans— Niren Murthy
La Conversación del Hearth Otra perspectiva de la historia
Why does the size of the delivery vehicle matter so much? A particle that's half the size of a virus—does that automatically make it better?
Not automatically, but it changes what's possible. A smaller particle can squeeze through tissue spaces in the brain that larger ones cannot reach. It also means you might fit different kinds of gene-editing tools inside—not just the basic CRISPR, but newer, safer versions that are too bulky for viral vectors.
But if the researchers don't understand how their own molecule works, how confident should we be that it will work the same way in a human brain?
That's the honest question. They've shown it works in mice. But mice brains are tiny. A human brain is roughly a thousand times larger. The physics changes. The immune response changes. You're essentially asking: does a proof of concept in a small system predict success in a vastly more complex one? History says not always.
What about the genome damage you mentioned—the chromosomal rearrangements? Is that a dealbreaker?
It could be. If the active cutting enzyme is causing large deletions and inversions in the DNA, you're not just fixing one gene—you're potentially breaking others. That's not a therapy; that's a different kind of disease. They may need to use a different version of the editing tool, one that's less destructive but also less well-tested.
The researcher has a financial stake in the company commercializing this. Does that change how we should read the results?
It doesn't invalidate the science, but it adds a layer of caution. When someone stands to profit from a technology, there's an incentive—conscious or not—to emphasize promising results and downplay limitations. The NIH knows this, which is why they have transparency rules. But those rules don't kick in until 2028. We're essentially being asked to trust the process.
What would convince you this is real?
Publication of the mechanism. Head-to-head comparison with existing methods. Data from nonhuman primates showing it actually scales. And most importantly: a human trial that shows safety and efficacy. Until then, it's a very promising idea with a lot of unanswered questions.