AI-Designed Enzyme CMLase Reverses 'Molecular Rust' in Human Tissue: CML Levels in 75-Year-Old Aorta Cut to Those of a 30-Year-Old

Forum topic · QianXun · 2026-08-23

Summary

Researchers at Revel Pharmaceuticals and collaborators report in Nature Communications that an AI-designed enzyme, CMLase, removed roughly 70% of carboxymethyllysine (CML) — a key advanced glycation end-product (AGE) — from aortic tissue of a 75-year-old donor, bringing levels down to those typical of a 30-year-old. Similar reductions were seen in skin and eye lens proteins from a 64-year-old donor. AGEs, long considered irreversible 'rust' that stiffens elastic tissues and drives aging-related disease, were chemically cleaved off damaged collagen, retinal proteins, and hemoglobin, restoring native conformations. The team used AI to screen over 50,000 microbial genomes and model enzyme structures, then applied five rounds of directed evolution across 500+ million variants to boost activity tenfold. The work shifts anti-aging research from preventing new AGE formation toward repairing existing damage, though the authors caution that chemical reversal in excised tissue does not yet equal functional rejuvenation in living humans.

Overview

A long-standing assumption in aging research — that advanced glycation end-products (AGEs), once formed in the body, are irreversible — has been challenged. Revel Pharmaceuticals (San Francisco) and collaborators engineered an enzyme, CMLase, that removed about 70% of CML (N-carboxymethyllysine), the most common AGE marker, from aortic tissue of a 75-year-old donor, reducing levels to those seen in 30-year-olds. Similar reductions were observed in skin and eye lens proteins from a 64-year-old donor. The findings were published in Nature Communications.

Why AGEs Are Called 'The Body's Rust'

The browning of baked bread comes from sugars reacting with proteins under heat; the same reaction occurs slowly in the body at ~37°C, producing AGEs.
Decades of accumulation make elastic tissues stiff and fuel chronic inflammation — a classic hallmark of cardiovascular disease, diabetes, and eye/kidney damage.
The extracellular matrix makes up ~70% of the body and turns over extremely slowly (whole-body collagen half-replacement takes ~15 years), so long-lived proteins are increasingly damaged, causing skin sagging, weakened tendons, stiff joints, and organ decline.
Existing drugs can only block new AGE formation; they cannot clear AGEs already embedded in tissue, let alone reverse underlying protein damage.

How AI Found This 'Molecular Lawnmower'

The team targeted CML — hard to clear, pro-inflammatory, and known to trigger cells to secrete factors that stiffen tissue and harm brain microglia.
No human enzyme exists as a template, so they turned to microbes that degrade AGE-modified proteins in decaying matter.
Approach: AI screened DNA from over 50,000 microbes and modeled the structures of enzymes they encode, filtering for candidates that could penetrate large molecules like collagen to reach hidden CML. The best hit came from a thermophilic hot-spring bacterium.
The initial enzyme had modest activity, so the team used directed evolution (a Nobel-winning method that accelerates natural selection) over 5 rounds and 500+ million variants, yielding CMLase — over 10x faster than its precursor.
On CML-modified collagen, retinal proteins, and hemoglobin, CMLase cleaved off the chemical adducts and restored native conformations, effectively 'repainting' damaged molecules.

Why It Matters (Chemistry × Biology × AI)

1. From prevention to repair. It pushes anti-aging from 'blocking new damage' to 'reversing old damage,' directly challenging the foundational assumption that AGE damage is irreversible — a rare 'repair-approach' proof of concept in aging biology. 2. A textbook AI4S methodology. AI sequence/structure screening (50,000+ microbial genomes) → physics-informed modeling → a directed-evolution loop. It sidesteps the dead end of 'no natural template' via computation plus evolution. It belongs to the same lineage as AlphaFold/RFdiffusion and Baker-lab de novo enzyme design, but validated on real human tissue samples rather than mice — a harder leap. 3. Part of a broader wave of AI-designed biomolecules. An April 2026 Nature review (Wei Yang et al.) argued that the structural problems of protein design are nearly solved, and the live question is 'what to design.' A Food Research International review noted AI platforms predicting anti-aging peptide sequences from food protein libraries (targeting mTOR/Nrf2/IIS), with engineered microbes providing scalable synthesis routes. CMLase is a rare preclinical, human-tissue datapoint on the 'AI discovery/design → synthetic biomanufacturing → human validation' chain.

Caveats

Chemical reversal does not equal functional rejuvenation; evidence from excised tissue is still far from in-body repair in living humans. Revel's CEO acknowledged that 'more work is needed.' But with the 'irreversibility' premise overturned, the starting point for R&D in this direction has been reset.

Bottom line: When an AI-designed enzyme can snip the rust out of our bodies, anti-aging medicine has its first candidate for genuine 'repair' rather than mere 'maintenance.'

Sources

Nature Communications (Revel Pharmaceuticals et al., CMLase study)
BIOSCIENCE / The Scientist coverage
April 2026 Nature protein design review
Food Research International, DOI 10.1016/j.foodres.2026.119189 (review on AI-discovered anti-aging peptides)
SpaceDaily (overview of AI-designed novel proteins)