How a misread signal turns muscle into bone

In Short. In a rare disease that turns muscle into bone, cells obey a chemical message they should ignore, and new drugs that block that message stop most stray bone in small trials, without yet proving they keep people moving for life.

MOST CHILDREN with the disease are born with oddly formed big toes. Otherwise they seem healthy until, usually in early childhood, hot and painful swellings rise in the muscles of the neck and shoulders. Doctors call these episodes flare-ups. When one subsides it can leave something permanent behind, a plate or strut of new bone where muscle, tendon or ligament used to be. Over the years it works down the body into the limbs, and joints lock one by one. Patients can be fairly stable, a doctor at one of the drugmakers told Fierce Pharma, a trade publication, and then wake up to find that a jaw or a shoulder has locked overnight.

The cruelest feature is what sets flare-ups off. Almost any injury to muscle can do it, a fall, say, or an invasive medical procedure. So can a viral illness such as flu. The damaged tissue swells, then turns to bone faster than usual. That rules out the obvious remedy. The extra bone is ordinary bone in the wrong place, and a surgeon could in principle cut it out. But the operation is itself an injury, exactly what provokes new bone.

The disease is called fibrodysplasia ossificans progressiva (FOP). It affects about 900 people worldwide. By the age of 30 most patients need a wheelchair and full-time help. Median life expectancy is 56, and death often comes when bone around the ribcage stops the lungs from expanding. Until recently medicine could only ease the symptoms. That is now changing, and the change began with a single letter.

To see why, it helps to know how cells talk. They send each other chemical messages, often small proteins released by one cell and picked up by another. The picking up is done by receptors, proteins that sit in a cell's outer membrane with one end outside, to catch a particular message, and one end inside, to pass the news on. Each receptor is tuned to certain messages. When the right one docks, the receptor sets off a chain of relay proteins that carries instructions to the cell's DNA. One family of messages, the bone morphogenetic proteins (BMPs), tells cells to become cartilage and bone. That is how the skeleton is built in the first place.

The instructions for each receptor are written in a gene, a stretch of DNA spelled out in a four-letter chemical alphabet (A, C, G and T). The cell reads the letters three at a time, and each triplet specifies one of the building blocks, called amino acids, that are strung together to make a protein. Change a single letter and the cell may put the wrong building block in one spot. Occasionally that changes everything.

In 2006 Eileen Shore, Frederick Kaplan and their colleagues at the University of Pennsylvania traced FOP to a gene called ACVR1 (also known as ALK2), which makes one of the receptors for BMPs. In every affected person they examined, whether the disease ran in the family or not, they found the same misspelling. At letter 617 of the gene's code a G had become an A. That swaps building block number 206 in the receptor from arginine to histidine, a change geneticists write as R206H. The swap sits in a region that controls when the receptor switches on. Computer modeling suggested it loosened that control, leaving the receptor stuck "on" and telling cells to make bone when they should not.

Researchers then argued over whether the mutant receptor fired by itself or over-reacted to its normal BMP messages. Either way, the assumed fault was in how loudly it spoke.

In September 2015 a team at an American biotech firm, led by Sarah Hatsell and Aris Economides, reported something nobody had expected. The mutant receptor had started answering a different message altogether, called activin A. Activin A can reach the same receptor, but in healthy cells it does not tell it to make bone. If anything it damps BMP signaling down. The mutation had turned a brake into an accelerator.

Mice that carry the human mutation from conception die around birth, so to test the idea the team built mice in which it could be switched on later. Once it was, the animals grew stray bone much like people with FOP. Injecting activin A triggered bone in these mice but not in normal ones. And an antibody against activin A (an immune protein built to latch onto one target) blocked the stray bone. Activin A, the authors concluded, was both necessary and sufficient to drive the disease in their mice. Months later a group at Kyoto University reached the same conclusion using stem cells made from patients' own cells. Activin A pushed those cells to form cartilage, which then became bone, the same cartilage-first route by which the normal skeleton forms.

The lesson reaches beyond FOP. A mutation need not make a protein louder or quieter. It can change what the protein listens to. For drug hunters that offered several places to intervene. One could remove the message, block the receptor, or make the cells downstream deaf to the order.

Cutting the line

In just over a month America's drug regulator, the Food and Drug Administration (FDA), has approved two drugs aimed at the misread signal. They join a third, approved in 2023, that works further down the chain.

That older drug is palovarotene, sold by Ipsen, a French drugmaker, as Sohonos. It is a retinoid, a chemical cousin of vitamin A, and it acts on a receptor inside cells that helps regulate skeletal development. In 2011 researchers at Thomas Jefferson University showed that such drugs essentially prevented stray bone in mice. Treated cells stopped responding to BMP and made less of the relay proteins that carry the receptor's message inward. In effect the drug makes would-be bone cells deaf to the order, whoever gives it. It is approved for girls from eight and boys from ten, and its cost falls on growth. In growing children it can close the growth plates early and for good. These are the zones near the ends of bones where lengthening happens. Its label warns of that alongside the risk of harm to an unborn child.

The second drug goes after the message itself. The firm behind the 2015 mouse study was Regeneron, and it spent the following decade turning that finding into a medicine. Garetosmab, sold as Pasatru, is an antibody that mops up activin A before it can reach the receptor. Patients receive it by drip every four weeks. The FDA approved it on August 19th for adults only. According to Fierce Pharma, five of the 44 patients in the drug's earlier trial died. There was no clear link to the drug, but it could not be ruled out, and plans to seek approval were dropped for a time. The later trial reported no deaths. Pasatru's label warns of harm to an unborn child, of skin and soft-tissue infections and of nosebleeds.

The third drug, approved on September 25th, blocks the receptor. Zilurgisertib, developed by Incyte and sold by Mirum Pharmaceuticals as Atebrioz, is a daily pill designed to inhibit ALK2 itself, which should stop the signal inside the cell whichever message arrives. It is approved for people aged 12 and over. According to the FDA, it too can harm a fetus. That all three carry such warnings is no great surprise. Pasatru's label ties the risk partly to how the drug works, and the signals these drugs interfere with also govern how the skeleton grows.

Bones of contention

Knowing which drug works best is hard, because each was tested differently, and in very few people.

The fairest test splits patients by chance, the equivalent of a coin toss, into those who get the drug and those who do not, so that the two groups differ only in the treatment. Trialists call this a randomized trial. The comparison group gets an identical-looking dummy, a placebo. Both new drugs were tested this way. Sohonos was not. In its main trial 107 patients all took the drug and knew it, a design called open-label. Their new bone was compared with that of untreated people followed in an earlier study that simply watched how the disease unfolds, a natural-history comparison. That is weaker, because the groups were not formed by chance and may differ in ways nobody measured. Against that yardstick Sohonos cut the yearly volume of new bone by 54%, as estimated with a weighted statistical model.

A trial must also declare in advance the single measure that will decide whether the drug worked, its primary endpoint. Other measures, secondary endpoints, can support the case but carry less weight. And drugs pass through numbered stages of testing in people, of which the third, larger stage is usually the decisive one.

Pasatru's trial was one of those, a phase 3 study. It randomly assigned 63 adults to one of two doses or to placebo for 56 weeks. Its primary endpoint was the number of new bone lesions seen on whole-body CT scans. The 21 people on placebo grew 19 new lesions between them. The 23 on the higher dose grew two, and the 19 on the lower dose grew one, cuts of 90% and 94%. Flare-ups counted by doctors fell from 66 on placebo to nine on the higher dose (the lower dose, with 53, made no clear difference). But flare-ups reported by patients themselves did not differ significantly, and nor, on the trial's pre-set statistical test, did the volume of new bone.

Atebrioz's trial was a mirror image. It was a phase 2 study, a stage earlier. It assigned 63 people aged 12 and over to the pill or placebo for 24 weeks. Its primary endpoint was the share of patients who grew any new lesion. One of 32 on the drug did, against five of 30 on placebo. That looks like a big gap, but in groups this small it could plausibly have been luck (p=0.0986, where the conventional bar is below 0.05). So the trial missed its primary endpoint, as Fierce Pharma also noted. The FDA based its approval instead on a secondary endpoint, the change in the total volume of new bone. After 24 weeks that fell by an average of 3.2 cubic centimeters on the drug and rose by 24.6 on placebo. Because the main test failed, the companies call the statistics on such later measures "nominal", meaning suggestive rather than proven. After week 24 everyone got the drug, and by week 48 none of the 61 patients with scans had grown a new lesion.

Side by side, the two trials succeeded on opposite yardsticks. Pasatru clearly reduced the number of new lesions but not, on its pre-set test, their volume. Atebrioz showed a large drop in volume but missed on the share of people with any new lesion. Part of the reason is arithmetic. In each trial only five people on placebo grew any new lesion, so one patient either way could shift the verdict. The two drugs have never been compared directly, and the trials ran for different lengths in different age groups.

The real test is children. The drugs aim to stop bone forming, so they should matter most for those who have lost least. Mirum is already testing its pill in children as young as two, and Regeneron told Fierce Pharma it planned a trial in children. "The time to treat is as early as possible," said Susan Rhee, who leads the Pasatru program. But what matters to a family is not cubic centimeters on a scan after six months or a year. It is whether a child can still lift a hand to eat at 30, or breathe easily at 50, and no trial has yet run long enough to show that. FOP has become a textbook case of medicine moving from a misspelt gene to a mechanism to a treatment. Whether those treatments buy a freer life is something only today's children, growing up on them, will be able to show.