An Enzyme Made Donated Skin From People Over 70 Resemble 31-Year-Old Skin
Built with AI, it clears the sugary buildup the body cannot remove on its own.
All-In with Chamath, Jason, Sacks & Friedberg · Can the AI Industry Regulate Itself? Stripe Wants PayPal, China Catches Up, NY Bans DatacentersScientists used a lab-designed enzyme to break down the sugar and fat buildup between cells in donated skin from people over 70, making the skin chemically resemble that of a 31-year-old.
With age, sugars and fats stick to proteins in the material between cells. The buildup, called glycation, makes tissue sticky. Collagen fibers that should stay apart clump together, causing wrinkles, stiff joints, and inflammation.
The team from Calico, Google's age-reversal startup, working with Revel Pharma, focused on a compound called CML, which they describe as the predominant form of this buildup driving aging. They started with an AlphaFold-designed candidate aimed at CML, then created and screened growing batches of altered versions through five rounds to sharpen its activity.
The enzyme removed 52 to 97 percent of CML from body proteins in lab tests. On donated elderly skin, it cleared 55 percent, which the researchers said left CML at a level resembling 31-year-old skin.
Questions people ask
Has this been tried on living people?
No. The work described was on donated skin from elderly patients, not treatment of living volunteers. The researchers still have to work out how the enzyme might be delivered and whether it is safe and lasting in a living body.
What does it mean for skin to resemble 31-year-old skin?
The comparison is about how much CML is stuck to skin proteins. After the enzyme cleared 55 percent, the researchers described the remaining level as resembling that of 31-year-old skin.
What is AlphaFold and why did it matter here?
It is a Google DeepMind program that predicts how proteins fold into three-dimensional shapes. Knowing the shape helps scientists find a starting protein that can latch onto a target like CML, and then improve it by testing variants.