The switch, not the gene: a week spent on what a cell is allowed to read

Every cell in your body carries the same DNA. What makes a liver cell a liver cell is not a different set of genes. It is which genes that cell is allowed to read. This week, an unusual number of results and one large funding round landed on that same idea.

A drug that switches a gene off without touching the DNA

Epicrispr Biotechnologies raised another $90 million on Tuesday [1]. The money backs EPI-321, an experimental treatment for facioscapulohumeral muscular dystrophy, or FSHD, a progressive muscle-wasting disease [1].

In FSHD, a gene called DUX4 is switched on too hard, and muscle cells degenerate [1]. Conventional genetic medicine would cut the gene. EPI-321 does not cut anything [1]. It binds to a specific stretch of DUX4 and adds a chemical mark that suppresses the protein it makes [1].

The technique has a name: epigenetic editing [1]. It uses CRISPR tools to switch genes on or off rather than alter DNA [1]. Backers argue that dialling a gene up or down, without breaking or rewriting the sequence, is less risky than cutting [1].

Keep the caveats. Epicrispr has finished enrolling an early-stage study, and six-month follow-up data are due in early October [1]. What exists now is company-reported data from a small number of patients, described by the chief executive as “important initial evidence” [1]. At least four other drugmakers are chasing DUX4 by other routes [1].

What an ageing brain stops keeping quiet

DNA does not float loose. It is wound around proteins into a package called chromatin, and how tightly it is wound decides what can be read. A paper in Nature Neuroscience on Monday reports that this packaging loosens with age in mouse and human tissue [2].

When it loosens, stretches that were held shut start being read again. That includes ancient viral sequences parked in our own genome long ago. In microglia, the brain’s resident immune cells, the paper links that reawakening to inflammation and to cells locking into a worn-out state [2]. The full paper sits behind a paywall, so treat the mechanism as reported rather than examined here.

Stanford reported a matching surprise the same week. Large numbers of immune cells from the bone marrow enter the human brain from middle age, and can turn into microglia [22]. That overturns a long-held assumption that the brain’s immune system stays separate from the body’s [22]. A bone-marrow cell that becomes a brain cell has changed nothing about its DNA. It has changed what it reads.

The ageing-clock work moved too. A Nature Medicine study built “tissue clocks” from 25,712 tissue slide images across 40 tissue types in 983 people, and found organs age at different speeds [9]. A separate study tested whether accelerated epigenetic age and metabolic syndrome push each other in both directions [6].

Cells that get stuck in the wrong state

Three papers this week describe the same failure from different angles: a cell settling into a state it should not hold.

In Crohn’s disease, researchers found that the supporting cells of gut tissue adopt an inflamed state with a distinctively open chromatin pattern [3]. They then treated those cells with drugs that remodel that packaging, and checked at least two weeks later [3]. The authors are careful about the implication: such drugs are already used in cancer, but their known toxicity may be harder to justify in a chronic inflammatory disease [3].

In cancer, a Nature Cell Biology paper ties EZH2, an enzyme that silences genes, to cholesterol production inside tumours, and calls it an unexpected weak point [7]. In triple-negative breast cancer, a genome-wide scan of DNA methylation marks turned up patterns that track tumour identity and the immune cells around it [5]. And a single-cell atlas of the human pancreas found more capacity for cells to shift identity than expected, across development, health and disease [8].

The tools arrived at the same time

None of this is readable without instruments. A Nature Biotechnology paper describes OneCell CUT&Tag, which reads chromatin marks, gene activity and surface proteins in the same single cell [4].

Separately, UK scientists announced a 20-million-pound Cambridge project to grow organoids - tiny clumps of tissue grown from a patient’s own cells - from NHS patients, for drug testing [17]. The aim is fewer animal studies and results that reflect how disease varies between people [17]. Researchers were blunt that animal use will not vanish, only fall [17].

Elsewhere: a big miss in lung cancer, and a safety reckoning

AstraZeneca stopped a phase 3 trial of volrustomig on Monday [10]. A phase 3 trial is the large final test before approval. The drug is a bispecific antibody, one built to grab two targets at once [10]. It was tested against Merck’s Keytruda plus chemotherapy in 895 people with advanced lung cancer [10]. An independent monitoring committee judged that neither survival goal was likely to be met in patients whose tumours lacked PD-L1, a marker that predicts response [10]. AstraZeneca released no data and is continuing the drug’s trials in cervical, head and neck, and mesothelioma cancers [10][11].

The same field delivered a win. Taiho and Cullinan said their pill zipalertinib, added to chemotherapy, delayed cancer growth better than chemotherapy alone in 285 previously untreated patients [12]. The numbers have not been released; the partners called the gain “clinically meaningful” and the monitoring committee recommended unblinding the study [12]. Cullinan is in line for a $100 million payment, and a US decision in a later-line setting is due by 27 February [12]. The niche is narrow: EGFR mutations appear in roughly 10-15% of US lung tumours, and this particular alteration in a fraction of those [13].

On safety, Fierce Biotech asked whether China’s fast track for early gene-therapy studies can still be trusted [14]. Two deaths tied to that pathway went undisclosed [14]. Science and Retraction Watch reported that a six-year-old girl with a rare, non-fatal disorder died last year, days after an experimental gene-editing treatment part-funded by her parents [14]. STAT uncovered a second case, a boy in a Duchenne trial who died of acute respiratory failure [14]. Both preceded new Chinese rules tightening the pathway [14].

Also this week: the FDA approved Bristol Myers Squibb’s iberdomide for multiple myeloma, the debut of a new class of cancer drug [21]. Nature covered the first narcolepsy drug aimed at the disorder’s root cause rather than its symptoms, approved earlier this month [18]. And Fulcrum Therapeutics agreed an all-stock merger with Slate Medicines, carrying $245 million to fund migraine trials, after dropping its own lead drug [19][20].

The number that keeps moving the wrong way

US kindergarten vaccination coverage fell again in the 2025-26 school year, by 0.1% [15]. Exemptions from at least one vaccine rose to 4.2%, from 3.6% [15].

The CDC counted 2,566 confirmed measles cases so far in 2026, the highest US annual total in more than 35 years [16]. There have been 169 hospitalisations and no deaths, against 243 hospitalisations and three deaths in 2025 [16]. Some 93% of cases were in people who were unvaccinated or whose status was unknown [16]. Cases have appeared in 45 states plus New York City and Washington DC [16].