The CMLase enzyme has removed a form of age-related chemical damage from old human tissue samples in the laboratory. That is a real and potentially important result. It is not, however, evidence that scientists can now reverse aging in living people.
The distinction matters because the study has nearly everything a viral longevity story needs: artificial intelligence, an enzyme designed to repair molecular damage, human tissue, and dramatic-sounding reductions of more than 70 percent in arterial samples and more than 55 percent in skin. Yet the most accurate description is also the most interesting one: researchers have shown that a specific protein modification long regarded as effectively irreversible can be enzymatically repaired ex vivo—outside a living body.
The peer-reviewed paper, “Reversal of protein chemical aging by enzymatic deglycation”, was published in Nature Communications on 14 July 2026 by researchers affiliated with Revel Pharmaceuticals, Calico Life Sciences and the University of Colorado Anschutz Medical Campus. Its subject is CMLase, an engineered enzyme that targets Nε-carboxymethyl-lysine, usually shortened to CML.
The short verdict
| Claim | What the evidence shows |
|---|---|
| “An enzyme repaired age-related damage in human tissue.” | Yes, in laboratory-treated tissue samples. CML staining fell substantially in arterial and skin samples from older donors. |
| “The enzyme was created with AI.” | Broadly yes, but that phrase is incomplete. AlphaFold-derived structures and computational screening helped researchers identify a starting scaffold; extensive laboratory selection and five rounds of directed evolution followed. |
| “Scientists reversed human aging.” | No. No living person was treated, and the study did not test lifespan, physical function, disease outcomes, whole-body rejuvenation or long-term safety. |
This is therefore neither a miracle cure nor an empty headline. It is a persuasive proof of concept for molecular repair—one that may open a new therapeutic direction if the difficult steps between a tissue experiment and a safe medicine can eventually be completed.
What is CML, and why does it accumulate with age?
Proteins are not permanent, pristine structures. During normal metabolism they are continually exposed to sugars, reactive carbonyl compounds, oxidation and other chemical stresses. Some proteins are rapidly replaced, but others—particularly components of the extracellular matrix—can remain in the body for years. The longer they persist, the more opportunities they have to acquire chemical modifications.
One important family of modifications is known as advanced glycation end products, or AGEs. Glycation begins when sugars or sugar-derived reactive molecules attach to proteins without the controlled help of an enzyme. A chain of reactions can then produce stable adducts and cross-links. CML is one of the most abundant and widely studied AGE modifications found in aging tissue.
CML matters for at least two reasons. First, attaching it to lysine changes the chemistry of a protein at that location and may interfere with the protein’s structure, interactions or mechanical properties. Second, CML can participate in biological signalling associated with inflammation and oxidative stress, including through the receptor for advanced glycation end products, RAGE. Its presence is not a simple clock that explains aging on its own, but it is part of the molecular wear that accumulates over time and is also associated with metabolic and vascular disease.
Until now, most strategies aimed at glycation have focused on prevention: lowering the production of reactive precursors, trapping them before they modify proteins, or reducing downstream inflammation. Those approaches may limit new damage, but they do not necessarily restore proteins that have already carried an AGE for years. The unusual promise of CMLase is that it is designed to do precisely that—to remove an existing CML modification and regenerate the native lysine residue.
How researchers built CMLase
The enzyme did not appear fully formed from a single AI prompt. The development process combined structural databases, computational filtering, biochemical insight and large-scale experimental evolution.
The team began with glycine oxidases, enzymes whose chemistry offered a possible route to oxidising CML. They searched UniProt sequences that had structures available through the AlphaFold Protein Structure Database. According to the methods, the search produced 71,426 non-redundant sequences; 56,499 had associated AlphaFold structures. After filtering by protein length, tens of thousands of candidates were compared structurally to find scaffolds with physical access to the sort of peptide-bound CML found in proteins.
This is where AI-assisted protein prediction was genuinely useful. AlphaFold did not demonstrate rejuvenation, and it did not replace experiments. It gave the researchers predicted three-dimensional structures at a scale that would have been impractical to obtain through traditional structural biology alone. Those models helped narrow the search for an enzyme architecture with a suitably open active site.
The selected starting enzyme still needed major improvement. The researchers created and tested enormous libraries of variants—more than 500 million in total—and used directed evolution to enrich versions with better activity. Across five evolutionary rounds, the team arrived at the variant called CrGO-897, or CMLase. The paper reports more than a tenfold improvement in catalytic efficiency on peptide substrates and activity against CML attached to full-length proteins.
That sequence of events is a useful corrective to the familiar phrase “AI-designed enzyme.” A more accurate formulation is: AI-predicted structures enabled a computational search, human researchers designed the screening strategy, and laboratory evolution selected an enzyme that worked. Computation accelerated discovery; biochemical experiments supplied the evidence.
What the experiments actually found
The researchers first tested the engineered enzyme on defined molecules and model proteins. They examined whether CMLase could act on CML while sparing ordinary lysine and other amino-acid modifications. They also studied the chemical reaction products, supporting the conclusion that the enzyme oxidises CML and restores the lysine side chain.
The work then moved to biologically richer samples, including human lens proteins and sections of aged human tissue. In abdominal aorta tissue from a 75-year-old donor, consecutive sections were incubated overnight either without enzyme, with an inactive control or with active CMLase. Immunohistochemical staining was used to visualise and quantify CML. Across spatially aligned regions, active CMLase reduced the measured CML signal by more than 70 percent.
In skin from older human donors, overnight CMLase treatment produced a reduction of more than 55 percent in CML staining across epidermal and dermal layers. The authors report that staining after treatment fell below the level observed in a sample of 31-year-old skin.
That last comparison can easily generate an exaggerated headline: “old skin made younger than 31-year-old skin.” But staining intensity for one chemical adduct is not the biological age of skin. The experiment did not show that the treated tissue regained youthful elasticity, wound healing, cellular composition, immune function or appearance. It showed that the enzyme removed much of one detectable molecular modification under controlled laboratory conditions.
Why the controls and measurement method matter
A percentage is only as informative as the experiment behind it. In this case, the researchers used immunohistochemistry: an antibody recognises CML in a tissue section, and a coloured signal reveals where the modification is present. Image analysis can then compare staining intensity between aligned regions. This is a standard way to visualise a molecular target in its tissue context, but it is not a direct measurement of youth, health or organ performance.
The comparison with an inactivated enzyme is particularly useful. If both active and inactive preparations had reduced the signal equally, washing, incubation or some non-specific property of the protein could have explained the result. The stronger effect with active CMLase supports an enzyme-dependent reaction. The authors also used antibody controls and competitive antigen experiments to test whether the staining was recognising CML rather than an unrelated feature of the tissue.
There are still limits. Tissue staining is semi-quantitative and depends on sampling, antibody performance, imaging and analysis choices. Donor tissues also differ in age, health history and baseline damage. The study demonstrates biochemical activity across several substrates, but the human-tissue experiments do not amount to a large population study. Replication with more donors, orthogonal chemical measurements and blinded analysis would strengthen confidence in the size and generality of the effect.
Another subtle point is that “more than 70 percent” refers to the measured CML signal after an overnight treatment under the reported conditions. It does not mean that 70 percent of arterial aging disappeared. The denominator is one molecular readout, not all age-related damage in the artery. Keeping the denominator clear is one of the simplest ways to separate an exciting scientific result from an inflated health claim.
The result remains scientifically meaningful. CML on long-lived proteins was often treated as damage that could be prevented but not directly undone. Demonstrating enzymatic removal in naturally aged human tissue changes that assumption. It suggests that at least some molecular scars of aging may be repairable, even after they have formed.
Why “reversing protein chemical aging” is not the same as reversing aging
The title of the paper is careful: it refers to protein chemical aging. Online summaries often drop those two middle words, transforming a specific molecular finding into a claim about an entire organism.
Biological aging is not one reaction. The influential updated “Hallmarks of Aging” framework describes interacting processes that include genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, disabled autophagy, mitochondrial dysfunction, cellular senescence, stem-cell exhaustion, altered intercellular communication, chronic inflammation and dysbiosis, among others. Extracellular protein damage is relevant, but it is only one part of that network.
Even a perfect CML-removing drug would not automatically repair DNA mutations, replace lost stem cells, restore mitochondrial quality control or eliminate senescent cells. Nor does removing a molecular marker prove that the intervention improves a symptom or prevents a disease. In medicine, a biochemical change is most valuable when it leads to a meaningful benefit: stronger tissue, better vascular function, fewer complications, improved quality of life or longer healthy survival.
The present study did not measure those outcomes. Its tissue sections were no longer part of functioning organs. There was no circulation, immune system, metabolism, repeated dosing or opportunity to observe side effects. “Ex vivo human tissue” is more physiologically relevant than a simple test tube, but it is not a clinical trial.
The biggest unanswered questions
1. Can the enzyme reach the right tissues?
An enzyme that works when applied directly to a thin tissue section may behave differently when injected, inhaled, swallowed or applied to intact skin. Therapeutic proteins can be degraded, cleared rapidly, blocked by tissue barriers or prevented from entering the extracellular structures where long-lived glycated proteins reside. A viable treatment will require a delivery strategy that produces adequate exposure without affecting unintended sites.
2. Is it selective enough in a living system?
CMLase is intended to distinguish CML-modified lysine from normal lysine and other modifications. The paper includes specificity experiments, but a whole organism contains an immense range of proteins, metabolites and cellular environments. Broader profiling will be needed to identify rare off-target reactions and the consequences of the enzyme’s reaction products.
3. Could the immune system react to it?
Engineered enzymes can be recognised as foreign. An immune response might neutralise the treatment, cause inflammation or make repeat dosing unsafe. Researchers would need to characterise immunogenicity and, if necessary, redesign the protein or delivery schedule.
4. Does removing CML improve tissue function?
A fall in staining is evidence of target engagement, not automatically of benefit. Future studies must test whether treatment changes mechanical properties, vascular compliance, inflammatory signalling, organ performance or disease-relevant outcomes. The most compelling results would link the molecular repair to a reproducible functional improvement.
5. How durable is the effect?
Glycation continues as metabolism continues. Even if accumulated CML is removed, new CML can form. Researchers will need to learn how fast damage returns, whether repeated treatment is practical and whether prevention plus repair works better than either strategy alone.
6. What happens in animals and, eventually, people?
Before responsible human testing, researchers generally need pharmacology, dosing, toxicology and efficacy data in appropriate models. If those stages succeed, early clinical trials would first evaluate safety and biological activity—not promise rejuvenation. Larger controlled trials would then be required to show a clinically meaningful effect for a defined condition.
Where might a CML-repair strategy matter most?
The most credible near-term path may not be a universal “anti-aging” treatment. It may be a therapy for a specific tissue or disease in which CML burden contributes to pathology and can be measured alongside function.
Long-lived extracellular proteins in arteries, skin, tendons, cartilage and the eye are logical areas of interest because damage can accumulate where protein turnover is slow. Vascular stiffness and complications of diabetes are especially relevant to AGE biology, although relevance is not proof that CMLase will treat them. A targeted indication gives researchers clearer endpoints, a defined patient group and a more realistic way to weigh benefit against risk.
Local delivery could also be simpler than whole-body exposure. For example, an intervention applied to a restricted tissue might avoid some distribution and safety problems. These are development possibilities, not results from the current study, and the eventual direction will depend on preclinical evidence.
How to read viral longevity headlines about this study
Longevity news often compresses a chain of evidence into a single emotional sentence. A molecular effect becomes “rejuvenation”; an experiment on excised tissue becomes “tested in humans”; a predicted structure becomes “AI discovered the cure.” Readers can protect themselves with five questions:
- What exactly was treated? A purified protein, cultured cells, tissue sections, an animal or a living person?
- What was measured? A molecular marker, tissue function, symptoms, disease events or lifespan?
- Was there a control? In this study, inactive-enzyme and untreated comparisons helped attribute the reduction in staining to active CMLase.
- Has the result been independently replicated? A peer-reviewed first report is a beginning, not a final verdict.
- Do the authors have commercial interests? Several authors are affiliated with Revel Pharmaceuticals or Calico. That does not invalidate the work, but it makes transparent disclosure and independent confirmation important.
Video footage and AI-generated illustrations can make a scientific story easier to understand, but they are not evidence. When a television segment credits a YouTube channel for visuals, that credit may identify only the source of the footage. The evidence must still be traced to the paper, its methods, figures and data.
A genuinely promising result—without the miracle language
The CMLase study deserves attention because it attacks accumulated molecular damage directly. Instead of merely slowing the creation of new AGEs, the enzyme removed an existing modification and restored native lysine in experimental systems. It worked on model proteins and reduced CML staining in naturally aged human arterial and skin samples. The combination of computational structural screening, AlphaFold resources, massive variant libraries and directed evolution also illustrates how modern protein engineering can explore therapeutic chemistry that biology did not conveniently provide.
At the same time, the leap from “one AGE can be removed from excised tissue” to “human aging is reversible” crosses several missing bridges. We do not yet know whether CMLase can be delivered safely, whether it improves tissue function, whether benefits persist, whether it works in a whole organism or whether it will ever become a medicine.
The responsible conclusion is not disappointing. A form of protein damage previously considered essentially permanent appears amenable to targeted enzymatic repair. That is a powerful proof of principle. If future studies establish safety, delivery and functional benefit, it could become one component of therapies aimed at age-related disease. Today, however, it is an early platform result—not human rejuvenation.
Frequently asked questions
Has aging been reversed in humans?
No. The study treated human tissue samples outside the body. It did not administer CMLase to living people or measure biological age, healthspan or lifespan.
What did CMLase reverse?
It removed Nε-carboxymethyl-lysine, an advanced glycation end product attached to proteins, and regenerated native lysine. The authors describe this as reversal of protein chemical aging.
How large was the effect?
Immunohistochemical measurements showed more than a 70 percent reduction in CML content in elderly arterial tissue and more than a 55 percent reduction in elderly human skin after laboratory treatment.
Was CMLase made with artificial intelligence?
AI-assisted structure prediction played a meaningful role. The team used structures available in AlphaFoldDB for large-scale computational screening. They then relied on experimental screening and directed evolution of more than 500 million variants to produce the final enzyme.
Can I buy or receive CMLase treatment?
No approved anti-aging treatment based on this study is described. The work is preclinical and does not establish a safe dose, route of administration or benefit in people.
Does lowering AGEs through diet have the same effect?
No. Diet and metabolic health may influence the formation and exposure to some glycation products, but the study tested a specific engineered enzyme applied directly to existing CML on proteins. It should not be used to justify supplements or consumer products without separate clinical evidence.
What result would move the field forward next?
Convincing demonstrations of safe delivery and improved tissue function in living models would be important. Independent replication, detailed off-target testing and eventually carefully controlled human trials would be required before any clinical claim.
Sources and further reading
- Trabosh N, Smith J, Hsu MY-H, et al. Reversal of protein chemical aging by enzymatic deglycation. Nature Communications. 2026;17. DOI: 10.1038/s41467-026-75141-2.
- PubMed record for the CMLase study, including bibliographic information and abstract.
- López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer G. Hallmarks of Aging: An Expanding Universe. Cell. 2023;186(2):243–278.
- Jumper J, Evans R, Pritzel A, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596:583–589.
- Schalkwijk CG, Stehouwer CDA. Methylglyoxal, a highly reactive dicarbonyl compound, in diabetes, its vascular complications, and other age-related diseases. Physiological Reviews. 2020;100:407–461.
Medical note: This article is for education and science communication. It does not provide medical advice and should not be used to select a treatment or supplement.
