CMLase (CrGO-897)

TL;DR

CMLase (CrGO-897) is an engineered bacterial flavin adenine dinucleotide (FAD)-dependent glycine oxidase that removes protein-bound Nε-carboxymethyl-lysine (CML) by oxidative dealkylation, regenerating lysine. Trabosh et al. reported substantial CML reduction on model proteins and in ex-vivo human lens, aortic, and skin samples at 5 µM overnight.¹ This is the first engineered-biologic example in the age-adduct-deglycation mechanism class, but it is not a demonstrated AGE crosslink breaker: CML is a mature, stable, monovalent lysine adduct, and the study did not test cleavage of glucosepane, pentosidine, or any other true covalent crosslink.

CMLase remains early preclinical research material. It has not been administered to a living animal or human; delivery, tissue penetration, pharmacokinetics, immunogenicity, off-target proteome activity, functional benefit, and safety are unknown. gap/needs-human-replication gap/needs-replication

Identity

Field	Status
Agent	Engineered glycine-oxidase variant CrGO-897
Starting scaffold	Glycine oxidase from Calidithermus roseus (CrGO)
Modality	Recombinant enzyme / biologic
Canonical identifiers	None reported in the paper. Exact-name PubChem searches for CMLase and CrGO-897 returned no CID, and an exact-name ChEMBL search for CMLase returned no record on 2026-07-14; DrugBank was not independently established. Fields remain null.
WHO international nonproprietary name	None reported or independently established; left null
Molecular weight / final therapeutic sequence	The paper does not provide a final sequence or canonical biologic record from which an unambiguous molecular weight can be assigned; left null
Development stage	Preclinical biochemical and ex-vivo human-tissue proof of concept

The paper describes CrGO-897 as the product of five rounds of directed evolution after structural screening. More than ten libraries, cumulatively exceeding 5 × 10⁸ variants, were evaluated. Its mutation reporting is internally inconsistent: the main text says the final enzyme contains 15 substitutions plus a two-residue deletion relative to starting CrGO, whereas Supplementary Table 2 lists 7 + 1 + 3 + 2 + 3 = 16 cumulative substitutions plus the deletion. The main text and supplement also number the three CrGO-794 substitutions as A61G/A198S/Q214K versus A59G/A196S/Q212K, respectively, apparently using different numbering frames after the two-residue deletion without labeling them. Because no final sequence is supplied, the discrepancy cannot be reconciled from the publication.¹

Mechanism of action

Direct repair of an established AGE adduct

CMLase oxidizes the N-carboxymethyl substituent attached to a protein lysine, producing restored lysine together with hydrogen peroxide and glyoxylate/glyoxylic acid. The net action is repair of an already formed AGE adduct rather than inhibition of glycation, scavenging of an upstream dicarbonyl, or accelerated turnover of the damaged protein.¹

This is why the canonical mechanism value is age-adduct-deglycation. It is distinct from:

AGE-formation inhibitors such as carbonyl traps, which reduce future AGE formation but do not restore existing modified residues.
Early Amadori-product deglycation by fn3k or fructosyl-amino acid oxidases, which generally acts earlier in the Maillard cascade.
age-crosslink-cleavage, which would require severing a bifunctional bridge such as glucosepane between two residues.

Substrate specificity

CrGO-897 had Kₘ = 0.51 ± 0.10 mM and kcat = 5.3 ± 0.4 × 10⁻³ s⁻¹ on the AA-CML-AA peptide, corresponding to catalytic efficiency of approximately 1.0 × 10⁻² s⁻¹ mM⁻¹ and an approximately ninefold gain over the starting CrGO scaffold. With 5 µM CMLase and 1 µM CML-modified bovine serum albumin (CML-BSA), hydrogen peroxide accumulated at 0.90 µM/h.¹

The reported specificity panel found no detectable activity on unmodified BSA, carboxymethyl-arginine-modified BSA, other tested canonical free amino acids besides glycine, or glycine/arginine positioned at the termini of tested tripeptides. Activity on CML at protein N-terminal α-amines was not directly tested. Site-resolved proteomics found incomplete coverage, but the Methods identify the active enzyme in that experiment as the immediate predecessor CrGO-865, not final CrGO-897; Figure 3 and the Results call it CMLase. Incomplete site coverage is therefore demonstrated for the engineering lineage, but exact CrGO-897 site coverage is unresolved.¹

Evidence

Model proteins

At 5 µM overnight, CMLase reduced anti-CML enzyme-linked immunosorbent assay (ELISA) signal by 52–97% across chemically CML-modified BSA, casein, collagen, hemoglobin, and sheep retinal total-protein extract (Figure 3A, n = 4 assay replicates). Sodium dodecyl sulfate–polyacrylamide gel electrophoresis under the site-proteomics reaction conditions found no CML-BSA fragmentation; however, the Methods identify the active enzyme for those conditions as CrGO-865. A separate duplicate capillary western explicitly using 1 µM CrGO-897 also showed reduced anti-CML signal relative to inactive CeGO-763.¹

Site-resolved liquid chromatography–tandem mass spectrometry (LC–MS/MS) on CML-BSA found lower CML occupancy at 30 of 33 mapped lysines after treatment; 21 sites decreased by more than 50%, 7 by more than 90%, and 3 heavily modified sites by less than 5% (n = 3 assay replicates). The Methods specify CrGO-865 for this experiment despite the Figure 3/Results label of CMLase. Repair was not determined by surface exposure alone; a secondary site-level analysis found a weak correlation with local flexibility (B-factor; Spearman r = 0.3695, approximate two-sided p = 0.0343, n = 33 sites), and sequence-context associations were descriptive.¹

Ex-vivo human tissue

Tissue	Exposure and assay	Result	Biological N
Lens, age 64	5 µM overnight; hydrolysis + LC–MS/MS	Paper states 45% lower total CML; deposited triplicate means imply ~47%	one donor homogenate; three assay measurements
Lens, age 64	5 µM overnight; anti-CML ELISA	Paper states 78% lower signal; deposited triplicate means imply ~77%	same donor; three assay measurements
Abdominal aorta, age 75	5 µM overnight; anti-CML immunohistochemistry	Paper states >70% lower DAB staining; deposited means imply ~66% versus inactive enzyme (~62% versus no-enzyme/antibody-only)	four aligned spatial regions from one donor
Abdominal skin, age 74	5 µM overnight; anti-CML immunohistochemistry	Paper states >55% lower staining; deposited means imply ~58% in epidermis and ~63% in dermis versus inactive enzyme	four spatial regions from one donor; separately treated age-31 tissue

The figures document seven donors total: one lens donor, four aorta donors (three young contextual specimens and the 75-year-old efficacy specimen), and two skin donors. The lens difference between LC–MS/MS (45% stated) and ELISA (78% stated) is consistent with preferential enzyme access to surface-exposed, antibody-detectable CML while hydrolysis/mass spectrometry captures buried CML as well. Each headline older-tissue demonstration nevertheless uses one donor; assay measurements and image regions are not independent human subjects. No treatment-comparison inferential statistics or confidence intervals were reported, and the aortic “>70%” statement is not reproduced by the deposited group means.¹

What was not tested

No living mammalian cell culture, animal, or human was treated to test efficacy. The directed-evolution selection did use living E. coli expressing enzyme in the periplasm, but it was not a therapeutic model.
No RAGE binding/signaling, nuclear factor-κB activation, inflammatory cytokine, fibrosis, tissue-elasticity, arterial-compliance, lens-transparency, or other functional endpoint was measured.
No glucosepane, pentosidine, or other crosslink substrate was tested.
No administration route, formulation, biodistribution, pharmacokinetics, or pharmacodynamic duration was established.
No dose-response in intact tissue was reported; the human ex-vivo experiments used 5 µM overnight (18 h), and no therapeutically achievable living-tissue exposure was established.

Critical boundary: CML repair is not crosslink cleavage

CML modifies one lysine residue. Its removal can restore local charge and may reduce ligand burden for RAGE, but it does not sever a bridge between collagen chains or demonstrate restoration of extracellular-matrix mechanics. glucosepane is a bifunctional lysine–arginine crosslink and the dominant mature AGE crosslink in adult human extracellular matrix; the Trabosh paper identifies it as a possible target for future enzyme engineering and describes it as still resistant to reversal.¹

Therefore, CMLase belongs to age-adduct-deglycation, not age-crosslink-cleavage. It is relevant to the broader age-crosslink-breakers landscape as proof that mature AGE chemistry can be enzymatically engineered, but it does not validate that intervention class’s defining crosslink-cleavage claim.

Hallmark mapping

Hallmark	Rationale	Evidence level
loss-of-proteostasis	Direct chemical repair of damaged protein lysines	Target engagement only; restored protein function was not tested
altered-intercellular-communication	CML-modified proteins can act as aberrant extracellular signals through RAGE	Indirect mapping; RAGE signaling was not measured after treatment
chronic-inflammation	Reduced CML ligand burden could lower downstream inflammatory signaling	Hypothesized consequence only; no cytokine or inflammation endpoint

The hallmark links support mechanism-matrix discoverability; they do not mean that CMLase reversed any formal hallmark phenotype in vivo.¹

Delivery, safety, and translational constraints

Tissue penetration. Lens homogenate and 4-µm fixed tissue sections minimize diffusion barriers. An intact artery or dermis has dense, crosslinked extracellular matrix that a large enzyme may not penetrate. gap/dose-response-unclear
Immunogenicity. CMLase is derived from a bacterial oxidase. Repeat administration could provoke neutralizing antibodies, hypersensitivity, or cellular immune responses; de-immunization has not been reported.
Reaction byproducts. Each repair event produces hydrogen peroxide and glyoxylate/glyoxylic acid. The authors argue that low CML abundance should make this load small relative to endogenous clearance, but local production and toxicity were not measured.¹
Incomplete site access. Three heavily modified BSA lysines showed less than 5% repair in the site-proteomics experiment. Because the Methods assign that experiment to CrGO-865 while the Results call the enzyme CMLase, the exact resistant-site profile of CrGO-897 remains unresolved.
Unknown off-target scope. The tested substrate panel was limited. Proteome-wide modification, oxidation, and long-duration exposure studies are absent.
No functional efficacy or durability. It is unknown whether CML returns after enzyme clearance, whether protein function is restored, or whether repeated dosing is required. gap/long-term-unknown
Commercial conflict. Revel Pharmaceuticals filed U.S. Provisional Patent Application 64/039,597 on the work, naming first author Narisa Trabosh and senior author Aaron Cravens as inventors; both were Revel-affiliated. Other authors declared no conflict.¹

Human evidence and clinical trials

Human lens, aorta, and skin were used ex vivo, but no participant was dosed and no clinical efficacy or safety outcome exists. This supports human-evidence-level: preclinical-only, not limited human evidence.

As of 2026-07-14, separate official ClinicalTrials.gov v2 exact-intervention searches for CMLase and CrGO-897 each returned zero recruiting or active-not-recruiting trials. clinical-trials-active: 0 is therefore current to the literature-check date.

Literature recency check

The required 2026-07-14 searches found:

PubMed high-priority search for CMLase/CrGO-897 plus randomized trials, meta-analyses, or systematic reviews (2021-07-14 through 2026-07-14): 0 hits.
PubMed broad recent search for CMLase, CrGO-897, or “enzymatic deglycation” (2024-07-14 through 2026-07-14): 0 hits.
Europe PMC named search (2021-07-14 through 2026-07-14): 2 hits, comprising the 2023 FN3K/FAOD skin-deglycation study and a 2026 review false-positive; neither independently evaluated CMLase.
The founding Trabosh 2026 DOI was published on the search date and was not yet indexed by PubMed or Europe PMC.

No independent CMLase replication, in-vivo efficacy paper, clinical study, or contradictory result was found. literature-checked-through: 2026-07-14 records this search. gap/needs-replication

Development priority

The decisive next experiment is a blinded, dose-ranging in-vivo delivery study in aged large-animal skin or artery that jointly measures:

tissue exposure and penetration;
LC–MS/MS-confirmed CML target engagement;
RAGE/nuclear factor-κB and inflammatory readouts;
tissue biomechanics or vascular function;
pharmacokinetics, anti-drug antibodies, proteome-wide off-targets, and local hydrogen-peroxide/glyoxylate burden.

Without those data, the main uncertainty is not whether purified CMLase can catalyze CML removal, but whether a safe exposure can reach enough naturally modified sites in living tissue to produce a durable functional benefit.

Footnotes

trabosh-2026-cmlase-deglycation · doi:10.1038/s41467-026-75141-2 · in-vitro + ex-vivo human tissue · enzyme kinetics and site proteomics n = 3 assay replicates, model-protein ELISA n = 4, CrGO-897 capillary western n = 2; seven human donors documented, but headline older lens/aorta/skin outcomes each use one donor and aorta/skin image quantification uses n = 4 regions · model: living E. coli selection, purified CML-modified proteins, and human lens, aorta, and skin samples · no treatment-comparison p-values; secondary B-factor correlation p = 0.0343 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹²

🧬 Aging Wiki

Explorer

cmlase