ALS Gene Therapy Hits Multiple Targets At

When a protein called TDP-43 abandons its post in the nucleus of your motor neurons, it's basically game over. The protein moves to the cytoplasm, forms aggregates, and your cells start making truncated versions of critical genes. This happens in nearly all ALS patients—whether they inherited the disease or developed it spontaneously. The result is the same: progressive motor neuron death, two to five years from diagnosis to death, no cure.

The field has known about this TDP-43 problem for years. Several biotech companies have raised hundreds of millions of dollars to develop antisense oligonucleotides (ASOs) that target individual cryptic exons—the rogue genetic sequences that appear when TDP-43 goes rogue. Coralus is making one ASO for the STMN2 cryptic exon. Trace is making one for UNC13A. One target, one drug, repeat dosing every few months.

But here's the thing: there are many cryptic exons. Possibly dozens that matter. And current technologies can only hit one or two at a time.

Gene Yeo's lab at UC San Diego just presented research that changes that calculus entirely. They've developed a gene therapy platform that can target up to nine different disease pathways simultaneously, packaged in a single vector. In cell cultures and humanized mice, it works—restoring protein levels, improving axon growth, and potentially solving the manufacturing problems that have kept similar approaches from reaching patients.

The Biology That Breaks

To understand why this matters, you need to grasp what TDP-43 normally does. It's an RNA binding protein—one of thousands that modulate how genetic information moves from DNA to functional proteins. "A fifth of the human proteome consists of RNA binding proteins," Yeo noted in his presentation at the Sanford Stem Cell Institute Symposium. "That's four times larger than the number of kinases."

In healthy neurons, TDP-43 sits in the nucleus, where it acts as a molecular gatekeeper during splicing—the process where introns are removed and exons are stitched together to form mature RNA. TDP-43 binds to certain sequences and blocks them from being recognized by the splicing machinery. This is good. These sequences—cryptic exons—aren't supposed to be included.

But when TDP-43 leaves the nucleus, those cryptic exons suddenly become visible to the splicing machinery. The cell includes them in the mature RNA, creating truncated, nonfunctional proteins. For STMN2, a gene critical for axon stability and regeneration, the truncated version can't do its job. Loss of functional STMN2 causes motor neuropathy. Similar patterns play out across multiple genes.

"In your happy healthy neuron TDP-43 is in the nucleus," Yeo explained. "Over time with age that we have recently shown or with stress situation, TDP-43 moves to the cytoplasm and then through a yet undescribed process then it becomes an aggregate over time and then the neuron degenerates."

The first step—nuclear clearance—is where the cascade begins.

Why Existing Solutions Fall Short

The current approach is to use ASOs—synthetic molecules that bind to specific RNA sequences and block them from being spliced. They work, sort of. You can suppress one cryptic exon. Maybe two, if you use a different technology called dCas13D. But ASOs require repeat dosing (every three to four months), have sequence-dependent toxicity that compounds when you use multiple ASOs, and fundamentally can't scale to address the multiple cryptic exons that matter in ALS.

dCas13D can package two guides in one vector, but the bacterial protein component triggers high immunogenicity. It hasn't been used in humans.

Yeo's team turned to something that has been used in humans: small nuclear RNAs (snRNAs), specifically a type called U7. These already exist in your cells, controlling histone processing. But they can be reprogrammed—essentially hacked—to bind to any RNA sequence you want. They're RNA-only, so theoretically non-immunogenic. They can be delivered via AAV (adeno-associated virus), which means one-time dosing. And crucially, you can pack up to nine different guides into a single vector.

No one had actually shown that last part was possible. Until now.

SURF: Optimized By A Surfer

Trent, Yeo's graduate student (who apparently surfs), led the work. He started by designing guide RNAs targeting different sites around the STMN2 cryptic exon—one overlapping the splice site where it gets recognized, one where TDP-43 normally binds, one where previous ASOs had been designed.

The initial results were modest: 40% increase in mature STMN2 levels. Good, but not good enough.

So Trent went hunting for better promoters. Promoters control how much of the guide RNA gets expressed, and the standard human U7 promoter wasn't particularly strong. He tested promoters from across the mouse and human genomes. Some of the mouse promoters—despite being used in human cells—produced 10-fold more guide RNA than the human U7 promoter.

More guide RNA meant better suppression of the cryptic exon. With the optimized promoter (which Trent named SURF—Suppressing Unfavorable RNA by Functionalizing RBP, because of course a surfer named it that), they could restore STMN2 levels beyond what was lost. "In fact, it's like higher than non-target, non-TDP-43 treated, which is kind of crazy," Yeo said.

In motor neurons derived from induced pluripotent stem cells, SURF outperformed existing ASOs. In the Cleveland lab's axotomy model—where they sever motor neuron axons and watch whether they regrow—SURF-treated neurons with depleted TDP-43 regrew their axons nearly as well as healthy neurons. Without treatment, they didn't regrow at all.

In humanized mice carrying the STMN2 cryptic exon, AAV-delivered SURF restored protein levels in neurons.

The Manufacturing Problem No One Talks About

Here's where it gets interesting for anyone tracking why promising therapies fail to reach patients. When you try to package multiple copies of the same guide RNA with the same promoter into an AAV vector, you get recombination—the viral genome rearranges itself during manufacturing, and you lose your payload. This is allegedly why a Duchenne muscular dystrophy therapy using similar technology hasn't advanced beyond Phase 2, despite showing safety.

Trent tested whether using different promoters for different guides would solve this. It does. Their "TwinFin" design (yes, more surfing terminology)—two different guides with two different promoters—doubled the full-genome packaging efficiency compared to duplicated sequences.

This isn't just elegant science. It's the difference between a therapy that can be manufactured at scale and one that can't.

And TwinFin isn't the ceiling. In cell culture, they've shown they can target two different cryptic exons simultaneously—same suppression as individual guides, but packaged together. "This is the first time we're showing that you can hit two different targets at the same time in one vector," Yeo said, "with the idea that you can hit up to nine which I don't think we need but it's a nice thing to have."

What This Doesn't Answer

There are open questions, and Yeo's team is honest about them. During Q&A, someone asked about using base editing to change the cryptic splice acceptor site directly rather than blocking it. Yeo acknowledged the appeal but pointed to practical barriers: "If you do edits it has to be a one-time thing. So you have to basically use LNP mRNA because you don't want to keep Cas there for any length of time. And that hasn't really quite worked out for CNS, right?"

Lipid nanoparticle delivery—the technology behind COVID mRNA vaccines—doesn't cross the blood-brain barrier effectively. AAV does, but you can't keep a base editor hanging around indefinitely making changes. With SURF, "the whole thing is in so you can control that off target to some extent."

There's also the fundamental question of whether suppressing cryptic exons is sufficient. TDP-43 does many things in the nucleus. Blocking harmful splice sites addresses one downstream consequence of its absence, but it doesn't restore TDP-43's other functions. Will that matter clinically?

We don't know yet. This is mouse data and cell culture data. Humans are not mice, and motor neurons in a dish are not motor neurons in a spinal cord connected to muscles that need to contract for breathing.

Why Multiplexing Matters

But if—and it's still if—this approach translates to patients, the multiplexing capability matters enormously. ALS isn't one broken pathway. It's cascading failures across multiple systems. Being able to address multiple cryptic exons, multiple genes, multiple pathways with a single vector and a single dose isn't just more convenient than repeat ASO dosing. It's a different category of intervention.

Thirty thousand Americans are living with ALS right now. Five thousand will be diagnosed this year. They have two to five years. The field has been chasing this disease for decades with incremental approaches—one target, one drug, modest benefits.

Maybe the path forward isn't incremental. Maybe it's nine guides in one vector, engineered by a surfer, addressing the problem at its source.

—Nadia Marchetti