Gene Editing Weekly — July 17-24, 2026: a new clinical entry point, large-DNA insertion, and a governance failure

The week's signal

This week's strongest clinical addition is not a new efficacy readout. It is a new route into the sickle-cell race: the RESTORE Phase 1/2 study is enrolling adults for nula-cel, an ex vivo CRISPR-Cas9 therapy designed to correct the beta-globin mutation itself. Alongside it, a Nature paper reports large DNA insertion without double-strand breaks, Scribe Therapeutics priced a $128.7 million IPO, and a Science/Retraction Watch investigation exposed a fatal, previously undisclosed Chinese gene-editing experiment.

For readers triaging the field, the evidence separates into four levels:

Clinical entry: RESTORE is a safety-led, early-stage study. It has no efficacy result to read through yet.
Platform science: KNIT, a miniature base editor, and an engineered Cas9 all remain preclinical or preprint evidence.
Capital and infrastructure: Scribe's financing and the Broad-Apertura consortium put money and delivery capacity behind in vivo and personalized programs, but neither is a clinical validation event.
Governance risk: The China case is a reminder that a technically novel intervention can move faster than public trial reporting and independent oversight.

Clinical: direct correction enters a crowded sickle-cell field

Children's Hospital Los Angeles said on July 20 that it is one of three pediatric centers participating in RESTORE, a first-in-human Phase 1/2 study sponsored by Kamau Therapeutics. The single-arm trial is enrolling adults with severe sickle-cell disease and is expected to expand to patients as young as 12 later this year. Nula-cel collects a patient's own blood-forming stem cells, edits them outside the body, and reinfuses them after chemotherapy. Unlike Casgevy, which reactivates fetal hemoglobin, nula-cel is designed to convert the disease-causing beta-globin sequence toward normal adult hemoglobin. 1

That is a meaningful change in the competitive question, but not a clinical win. The public announcement reports the design and enrollment plan, not patient outcomes, safety events, or a primary efficacy readout. The trial still carries the familiar ex vivo burdens: myeloablative conditioning, cell manufacturing, and treatment at specialist centers.

A separate July 23 article revisited Intellia's Phase 3 HAELO results for lonvoguran ziclumeran, or lonvo-z. The article reports 62% of treated patients were attack-free without maintenance therapy versus 11% on placebo, with an 89% decline in on-demand medication use and a 91% reduction in moderate-to-severe attacks. Those are the previously reported Phase 3 data, not a new dataset this week; the new information is the possibility of regulatory submissions in the United States and Europe within roughly a year. 2

Paper triage: three different attempts to remove a platform bottleneck

KNIT targets large insertions without a double-strand break

A July 22 Nature paper introduces CRISPR kilobase-scale nickase-targeting, or KNIT. It couples a Cas9 nickase to a DNA-donor recruitment system, enabling programmable insertion of fragments from 0.7 kb to more than 10 kb. The authors report up to 89% insertion efficiency across loci and cell types, with lower indel, translocation, and off-target rates than the comparison systems. They also show non-viral CAR-T engineering in primary human T cells and antitumor activity in cell and mouse models. 3

The read-through is about insertion chemistry, not a ready-made therapeutic. The strongest evidence is still in engineered cells and mice, and the paper does not establish human safety, in vivo delivery, or durability. This is the paper to pull for readers working on gene replacement or CAR-T manufacturing, especially if double-strand-break liability is the limiting design constraint. The Nature paper and its PubMed record are the original reading paths.

Miniature base editors use zinc fingers to fit inside AAV

A July 20 Nature Communications study addresses a different problem: conventional base editors can exceed AAV packaging limits. The authors attach programmable zinc-finger proteins to miniature editors based on Un1Cas12f1 and OgeuIscB, then use protein-language-model optimization to improve the zinc-finger components. A single AAV delivered the resulting editor in a mouse model of spinal muscular atrophy, where it converted the SMN2 exon 7 target and restored exon inclusion. The abstract does not report the in vivo editing percentage or a numerical functional effect, so those figures should not be inferred from the paper's positive wording. 4

The paper is most relevant to in vivo delivery diligence. It narrows the packaging problem in mice; it does not answer biodistribution, immune recognition, repeat dosing, or human efficacy. The Nature Communications article is the primary source.

StaCas9 trades PAM breadth for compactness and activity

A July 21 bioRxiv preprint reports computation-assisted engineering of the compact StaCas9 nuclease. In the authors' assays, the editor reached as high as 95% at 12 endogenous sites carrying NNGG PAMs, while GUIDE-seq found no detected off-target at one tested site and one at another. At the corresponding sites, the comparison SpCas9-NG produced six and five off-target sites. The preprint also reports an engineered enStaCas9 variant that improved activity at several sites, although the gains remained target-dependent. 5

This is a preprint, and the limitations are explicit: StaCas9 still prefers NNG, especially NNGG, while activity falls at other PAMs and varies by cell type and locus. The comparison is useful for AAV-constrained editor design, but it is not evidence of therapeutic safety.

Capital and delivery: money is moving before human proof

Scribe Therapeutics priced an upsized IPO on July 23 at $15 per share for 8.58 million shares, or expected gross proceeds of $128.7 million. The company also agreed to sell 500,000 shares to Sanofi in a concurrent private placement. The shares were expected to begin trading on Nasdaq on July 24, with closing expected July 27. 6

Scribe's lead STX-1150 is an LNP-delivered, liver-targeted program intended to epigenetically silence PCSK9 without permanently changing DNA. Its S-1 describes STX-1200, targeting LPA, and STX-1400, targeting APOC3, as earlier programs. The filing also says the company has a limited operating history, no approved product, and substantial financing and translation risks. The capital event therefore expands Scribe's runway; it does not validate the pipeline in patients. 7

Apertura Gene Therapy said on July 23 that it had joined a Broad Institute-led consortium selected for ARPA-H's THRIVE program. The consortium plans a Pediatric Epilepsies and Rare CNS Gene Editing Platform, with shared manufacturing, regulatory precedent, and clinical infrastructure. Apertura's TfR1 CapX capsid is intended for intravenous delivery across the blood-brain barrier and will be paired with base- and prime-editing programs for rare pediatric central nervous system disease. This is a delivery and development-system announcement, not human data. 8

Another personnel move points in the same direction. Formation Bio hired Michael Ehlers, the co-founder and former CEO of exon-editing company Ascidian Therapeutics, to lead R&D. The move is relevant to the field's shift toward computationally supported genetic medicines, but no new program or clinical milestone was disclosed with the appointment. 9

FDA watch: no new decision, but the calendar still matters

This window produced no new FDA gene-therapy approval, complete response letter, advisory vote, IND clearance, or BLA/NDA acceptance to add to the ledger. The next CTGTAC meetings on the public calendar are July 29 for Capricor's deramiocel and July 30 for Replimune's vusolimogene oderparepvec plus nivolumab. They are scheduled advisory meetings, not FDA decisions, and neither product is a gene-editing therapy. 10 11

The practical implication is a quiet regulatory week rather than a quiet field. The clinical and capital signals arrived outside an FDA decision, so investors should not treat a press release, a trial launch, or an advisory-calendar entry as equivalent regulatory progress.

Governance: a fatal experiment that stayed out of the public record

A joint Science and Retraction Watch investigation published July 23 reported that a six-year-old girl with a CHD3 mutation and Snijders Blok-Campeau syndrome received an experimental base-editing treatment at Shanghai's Xinhua Hospital in 2025. The treatment used an AAV9-delivered editor administered through the cerebrospinal fluid, according to the investigation. The child died seven days later; documents reviewed by the reporters said an institutional ethics committee judged the death definitely related to treatment and identified thrombotic microangiopathy. 12

The same reporting said the death was not publicly disclosed, that the clinical record had not been updated for more than a year, and that the related Nature preclinical paper did not mention the clinical death. Those are findings and allegations in an investigative report, not a final regulatory adjudication. The hospital and researchers did not respond to the reporters' requests for comment, so the public record still lacks their account. 13

For the field, the immediate lesson is narrower than "stop the science." It is that first-in-human editing of the central nervous system requires an auditable chain from animal safety data to ethics review, trial registration, adverse-event reporting, and publication. A technically impressive editor cannot compensate for missing links in that chain.

What to carry forward

Treat RESTORE as a new clinical test of direct beta-globin correction, not as evidence that nula-cel works. Watch for the first safety and engraftment disclosures.
Separate insertion performance, editor size, and delivery. KNIT addresses large DNA insertion; the miniature-editor papers address packaging and targeting; none supplies human evidence.
Read Scribe's IPO as a financing signal with substantial execution risk, while the Broad-Apertura program is a bet on reusable delivery and trial infrastructure.
Track the China investigation for institutional responses, corrections, or regulator findings. Until then, keep the reported facts and the unresolved allegations distinct.