A 'switch clamp' small enough to fit inside a virus: one vector opens six genes at once
Gene therapy has an old, much-discussed problem. The most common delivery vehicle, adeno-associated virus (AAV), is like a suitcase with only ~4.7 kb of capacity, while the full CRISPR transcriptional activation (CRISPRa) kit—effector protein, activation domain, and guide RNA expression cassettes—often exceeds that budget. Want to upregulate multiple genes at once? Sorry, that means splitting across two or more viruses, doubling immune risks and manufacturing costs.
On October 28, a paper in *Nature Biotechnology* moved this suitcase problem a big step forward. A team with co-corresponding authors from the Shanghai Institute of Materia Medica / Zhongshan Drug Innovation Research Institute (Guoling Li, Yingsi Zhou, Hui Yang, Hainan Zhang), Guangxi Medical University (Guohe Tan), and the Eye & ENT Hospital of Fudan University (Jinhai Huang) developed smCRISPRa, a single-AAV-delivered in vivo multiplexed gene activation platform, and demonstrated functional improvement in both Dravet syndrome and Duchenne muscular dystrophy (DMD) mouse models.
Finding a clamp that is small enough and packs its own luggage
The team's approach had two steps.
Step one: find the material. From more than 2 million bacterial genomes and metagenomes, they screened 18 uncharacterized Cas12m homologs. Cas12m proteins are type V-M CRISPR enzymes of only ~600 amino acids—less than half the size of SpCas9 (~1368 aa)—and naturally show no detectable DNA-cutting activity. Transcriptional activation doesn't need cutting, so this clamp is inherently safe. Better still, 9 of them retain the ability to process pre-crRNA, so multiple guide RNAs can be strung together into a single transcript that self-processes after entering the cell—the luggage packs itself.
Step two: engineering. The team selected Cas12m18 and, through seven rounds of arginine substitutions plus shortening the direct repeat to 35 nucleotides, obtained denCas12m, then fused an enhanced miniVPR activation domain to build smCRISPRa. It recognizes all 16 NTTN PAMs, offering far more usable landing sites within a limited promoter window than enzymes with restrictive PAMs.
> Tip: the PAM is the "password" a CRISPR system uses to recognize target DNA—just a couple of bases. The more permissive the password (NTTN means any base first, then three fixed T's), the more editable sites are available. The crRNA is the "address label" that guides the protein to its target; stringing multiple addresses together and cutting them apart inside the cell is the key to this compact design.
Two models, two therapeutic logics
Route one: restore a missing dosage directly. Dravet syndrome is caused by haploinsufficiency of the *Scn1a* gene—patients lack normal expression of the Nav1.1 sodium channel. The team packed a dual crRNA array targeting *Scn1a* plus the activation system into a single AAV9 and injected it into the brain ventricles of neonatal mice. Results: cortical *Scn1a* mRNA rose about 4-fold; over a ~100-day observation window, both treatment dose groups survived 100%—comparable to wild type; the threshold for heat-induced seizures increased, and abnormal EEG activity and seizure frequency decreased. For a disease marked by refractory epilepsy with limited drug control, the survival data is a hard endpoint.
Route two: activate a substitute player. DMD patients lack dystrophin. Rather than touching the *dystrophin* gene itself (too large to fit in AAV), the team upregulated its natural substitute, utrophin. Eight weeks after systemic delivery to humanized DMD mice, *Utrn* expression rose to 4-7 fold, with grip strength and exercise endurance improvements lasting six months.
And one "flex" experiment: a single vector carried 8 crRNAs simultaneously targeting six genes—*Scn1a*, *Utrn*, *Fst*, *Sox2*, *Oct4*, and *Klf4*—and 5 of 6 transcripts were significantly upregulated in mouse muscle. One virus opening six endogenous genes in vivo is rare.
| Model | Target | Delivery | Key result | |---|---|---|---| | Dravet (*Scn1a* haploinsufficiency) | *Scn1a* | AAV9, neonatal intracerebroventricular | mRNA up ~4-fold; 100-day survival equal to wild type; higher seizure threshold | | DMD (humanized) | *Utrn* (dystrophin substitute) | Systemic | *Utrn* up 4-7x after 8 weeks; grip/endurance gains lasting 6 months | | Multiplexed activation demo | Six genes incl. *Scn1a* | Single vector, 8 crRNAs | 5/6 transcripts significantly upregulated |
Where it sits in the CRISPR lineage
CRISPRa is not new; the hard part has always satisfying "in vivo + multiplexed + single vector" simultaneously. dCas9-VP64 and CRISPR-SAM work well but exceed the AAV budget; miniaturized Cas12f fits but lacks endogenous pre-crRNA processing, making guide array design harder. Cas12m sits at the intersection of both constraints: small enough, and able to process its own guide RNAs. That intersection was previously empty.
Risks and boundaries. This is mouse data; delivery dosing, immune responses, and off-target activation profiles in humans all remain to be tested. Transcriptional activation does not change DNA sequence, meaning effects are reversible and require sustained expression—long-term dosing regimens are unresolved. The six-gene activation demo includes three reprogramming factors (*Sox2*, *Oct4*, *Klf4*), and the consequences of accidental activation need separate safety evaluation. The paper includes no human or preclinical toxicology data; reading it as a "new gene therapy drug" is premature. Its current identity is an engineering platform that turns "opening multiple genes at once" from multiple injections into one.
Sources and caveats: Paper: *An engineered, compact Cas12m variant for multiplexed gene activation in vivo*, Nature Biotechnology, online October 28, 2026. Figures such as the 4-fold, 4-7x, 100-day, 6-month, and 5/6 numbers come from the paper's reported results; the 2-million-genome screen yielding 18 homologs is from the original text. The sixth gene (*Klf4*) in the multiplexed demo did not reach significance; reports did not explain why.