In the early 2000s, a company called Alteon Inc. thought it had solved a problem that had stumped biochemists for two decades: how to reverse the specific kind of chemical damage that builds up in old proteins. Its compound, alagebrium (ALT-711), showed real signs of working in a small open-label pilot — patients with stiffened hearts saw measurable improvement in how well the heart’s main chamber filled with blood between beats. A larger, placebo-controlled trial later found no significant benefit on exercise tolerance or cardiac function, muddying that early signal. A follow-on study was terminated before completion when the company’s finances collapsed, and the drug never reached approval — leaving open exactly how much of alagebrium’s story was the chemistry underdelivering and how much was the runway running out.
A new paper in Nature Communications, published 14 July 2026, is the most serious attempt at this same problem since alagebrium’s collapse. It’s worth reading this new result against that history, because the two failure modes this field has run into — the biology not working, and the business not surviving contact with the clinic — are very different problems, and only one of them has actually been solved here.
The damage nobody could undo
The target in both cases is a class of molecule called advanced glycation end-products, or AGEs — formed when sugar reacts with long-lived proteins in a slow process chemists sometimes compare to bread browning in a toaster, except it happens inside the body over decades instead of minutes in an oven. Once these cross-links form, there’s no known biological process that removes them. That assumption — permanent, one-way, unfixable — is what alagebrium tried to challenge chemically, and what the new paper is trying to challenge with an engineered enzyme instead.
The new work, led by Aaron Cravens at Revel Pharmaceuticals with collaborators reported at Calico Life Sciences and the University of Colorado Anschutz Medical Campus, built an enzyme called CMLase, targeting a specific AGE called Nε-carboxymethyl-lysine, or CML. Starting from a bacterial glycine oxidase, the team ran more than 500 million variants through directed evolution before arriving at a version that could recognize and dismantle CML specifically, tested on tissue donated by people who had died: skin, an artery, and lens material. In donated skin, the treatment cut CML levels by more than fifty-five per cent — enough to bring the tissue below levels typically seen in a 31-year-old. In the abdominal aorta of a 75-year-old donor, overnight treatment cut CML by more than seventy per cent. In lens tissue from a 64-year-old donor, reductions ran 45 to 78 per cent depending on the measurement method.
What’s actually different this time
The paper’s authors cite alagebrium’s history directly, which is notable — it would have been easy to skip past it. What they’re implicitly arguing is that CMLase fails differently than alagebrium did, if it fails at all.
Alagebrium was a small molecule, chemically breaking cross-links wherever it found the right bond — a blunt instrument that worked, to the extent it worked, somewhat indiscriminately. CMLase is an engineered enzyme with a single, specific target: it doesn’t touch glucosepane, a different and possibly more consequential AGE thought to be the dominant cross-linker in aged collagen, at all. That specificity is a genuine improvement in precision. It does not, on its own, protect against the failure mode that actually killed alagebrium — money, not chemistry. Revel is a private company that has previously disclosed a $3.8 million NIH grant and a $12 million seed round; there is no clinical timeline in the paper, and the distance between “this works in a dish” and “this survives the years and funding a clinical program requires” is exactly the distance alagebrium never crossed.
Alagebrium never had to face a second, newer failure mode, because it was a small molecule rather than a protein: CMLase originates from bacteria, and the paper flags — without resolving — the question of whether a person’s immune system would react against it under repeated dosing. That’s a biological risk, not a business one — one alagebrium’s chemistry never had to face.
The base rate this sits against
None of this means CMLase is heading for the same ending. It means the relevant comparison isn’t “will the chemistry work” — early signs on that front are more specific and mechanistically cleaner than what alagebrium offered. The comparison is: this is a small field with essentially one prior serious attempt, and that attempt is remembered less for whether the science held up and more for the fact that it never got the runway to find out. One pharmacologist, reacting to the new paper, called the approach “pretty bold” — language that reads as uncertain rather than as a hedge covering for enthusiasm.
This is laboratory work on donated tissue, not a human trial, and it does not overturn forty-five years of literature treating this damage as permanent. It gives one research team a result, on donated tissue, suggesting that assumption is worth testing further. Whether CMLase gets the chance alagebrium didn’t — the years of funding, the immunogenicity data, the trial in a living organism rather than a tissue sample — is a question this paper can’t answer and doesn’t try to.