Where Advanced Glycation End Products Come From

In 1981 Vincent Monnier and Anthony Cerami incubated lens proteins with reducing sugars and watched them turn yellow. The pigments were fluorescent, the proteins became cross-linked, and both features resembled what pathologists had described in aged and cataractous human lenses. The reaction had a name already, borrowed from food chemistry: nonenzymatic browning, the Maillard reaction. Their argument in Science was that the same chemistry that browns stored food also runs, slowly, inside a living body.

That is the short answer to where advanced glycation end products come from. They come from sugars reacting with proteins without an enzyme, in a pan and in a capillary, by the same mechanism. The long answer is more interesting, because the two routes deposit different molecules at different rates, and because the instruments used to count them disagree with each other.

The Maillard reaction as it runs inside the body

The opening step is condensation of a reducing sugar with a free amino group on a protein – most often the epsilon-amino group of a lysine residue. The adduct rearranges, dehydrates, and proceeds to a heterogeneous family of end products: some coloured, some fluorescent, some able to bridge two protein chains.

Monnier and Cerami made a point that still structures the field. Most proteins in living systems turn over fast enough to be replaced before any of this matters. A few do not. Lens crystallins and skin collagen are exceptionally long-lived, and those are the proteins in which Maillard chemistry accumulates.

Glucose is a sluggish glycating agent. Much of the damage attributed to it is done by dicarbonyl intermediates that form downstream, above all methylglyoxal (PubChem CID 880; C3H4O2; 72.06 g/mol; CAS 78-98-8), which reacts with arginine and lysine orders of magnitude faster than glucose does. The best-characterised stable adduct is N-epsilon-(carboxymethyl)lysine, universally abbreviated CML (PubChem CID 123800; C8H16N2O4; 204.22 g/mol; CAS 5746-04-3). CML is stable and relatively inert, which is exactly why it is used as a marker: it survives long enough to be counted.

Where advanced glycation end products come from in food

Foods contain these products before anyone eats them, and heating creates more. Two research groups built databases of food AGE content, using different instruments, and the comparison is instructive.

The first, published by Uribarri and colleagues in 2010, covers 549 food items. Samples were homogenised and the supernatants assayed by enzyme-linked immunosorbent assay based on a monoclonal anti-CML antibody, with results expressed as AGE kilounits per 100 g. Dry heat promoted new AGE formation by more than 10- to 100-fold above the uncooked state across food categories. Animal-derived foods high in fat and protein were generally AGE-rich and prone to further formation during cooking; carbohydrate-rich foods such as vegetables, fruits, whole grains and milk contained relatively few, even after cooking. Formation was reduced by moist heat, shorter cooking times, lower temperatures and acidic ingredients such as lemon juice or vinegar, and was prevented in their assays by the inhibitor aminoguanidine.

The second, published by Scheijen and colleagues in 2016, quantified three named compounds – CML, N-epsilon-(1-carboxyethyl)lysine (CEL) and the methylglyoxal-derived MG-H1 – in the protein fraction of 190 food items by ultra-performance liquid chromatography tandem mass spectrometry. Highest levels sat in high-heat processed nut or grain products and canned meats; lowest in fruits, vegetables, butter and coffee.

The two databases broadly agree on direction and disagree on almost everything else. One reports an antibody-defined activity in arbitrary kilounits; the other reports molar quantities of three specific molecules. A number from one cannot be converted into a number from the other. This is a statement about analytical chemistry, not a dietary recommendation.

Cross-linking and the case of glucosepane

Browning is cosmetic. Cross-linking is mechanical. When a Maillard product bridges two collagen chains, the matrix stiffens and resists proteolytic digestion, and that is the change associated with ageing tissue and accelerated in diabetes. For years the known cross-links were present in quantities far too low to explain the effect.

Sell and colleagues addressed this in 2005 by measuring lysine-arginine cross-links derived from glucose, methylglyoxal, glyoxal and 3-deoxyglucosone – glucosepane, MODIC, GODIC and DOGDIC – in enzymatically digested human skin collagen (n = 110) and glomerular basement membrane (n = 28), using isotope dilution liquid chromatography/mass spectrometry. In non-diabetic controls at age 90, levels reached 2000, 30 and 15 pmol/mg for glucosepane, MODIC and GODIC respectively. Diabetes raised glucosepane to 5000 pmol/mg (p < 0.0001) while the others stayed below 60 pmol/mg. Glucosepane (PubChem CID 135565852; C18H32N6O6; 428.5 g/mol; CAS 257290-23-6) was, the authors concluded, the single major cross-link of the senescent extracellular matrix discovered so far, accounting for up to more than 120 mole% of triple-helical collagen modification in diabetes.

Two orders of magnitude separate glucosepane from the next cross-link in that dataset. Any account of glycation that stops at CML is measuring the marker that is easiest to detect rather than the modification that does the structural work.

How advanced glycation end products are measured

Four instruments are in common use, and they answer different questions.

Method	What it reports	Typical units	Principal limitation
ELISA with monoclonal anti-CML antibody	Antibody-reactive material in a sample	AGE kilounits per 100 g	Defines the analyte by what the antibody binds; not a molar quantity of a named compound
UPLC-MS/MS with stable isotopic dilution	Named adducts: CML, CEL, MG-H1 and others	mg or nmol per 100 g, or per mg protein	Requires exhaustive enzymatic hydrolysis for protein-bound residues; measures only what is targeted
Isotope dilution LC/MS of tissue hydrolysates	Cross-links such as glucosepane, MODIC, GODIC	pmol per mg collagen	Needs a tissue biopsy; not available in living population studies
Skin autofluorescence reader	Bulk fluorescence of skin, 420-600 nm emission	Arbitrary autofluorescence units	Non-specific; many AGEs are not fluorescent and other fluorophores are

Rabbani and Thornalley describe liquid chromatography-tandem mass spectrometry combined with stable isotopic dilution analysis as the reference method for quantifying glycation, oxidation and nitration adducts, giving high sensitivity and high specificity and allowing concurrent measurement of many adducts in one run. Everything else is a proxy.

The non-invasive option deserves a specific caution. Meerwaldt and colleagues validated a skin autofluorescence reader against biopsies from 46 patients with diabetes and 46 matched controls. Autofluorescence correlated with collagen-linked fluorescence, pentosidine, CML and CEL, but the correlation coefficients ran from 0.47 to 0.62. In 32 of the 46 patients with diabetes, values sat above the 95% confidence interval of the control mean. The method was validated in non-pigmented skin. A correlation in that range is useful for ranking groups and weak for characterising an individual.

What dietary intake predicts in plasma, and what it does not

The CODAM study measured plasma and urinary CML, CEL and MG-H1 by UPLC-MS/MS in 450 participants and estimated intake with a food frequency questionnaire combined with the mass-spectrometry food database. Higher estimated intake of each compound was associated with higher levels of the corresponding free adduct in plasma and urine. For plasma the standardised coefficients were 0.253 (95% CI 0.086 to 0.415) for CML, 0.194 (0.040 to 0.339) for CEL and 0.223 (0.069 to 0.373) for MG-H1.

The same paper reports that associations between dietary AGEs and the corresponding protein-bound plasma adducts were not significant. That distinction carries the weight. Free adducts are small molecules in transit toward renal clearance. The protein-bound pool is the one that sits on long-lived proteins and cross-links them. An observed association with the first is not evidence about the second, and the gap between them is where most confident statements about dietary glycation quietly fail.

Frequently asked questions

What is the Maillard reaction in the human body?

It is the same nonenzymatic browning chemistry that occurs in heated or stored food: a reducing sugar condenses with a free amino group on a protein, rearranges and dehydrates to yield pigments and cross-linked products. Monnier and Cerami proposed in 1981 that long-lived proteins such as lens crystallins and skin collagen are vulnerable because they are not replaced quickly enough to escape it.

What is carboxymethyllysine and where does it form?

N-epsilon-(carboxymethyl)lysine, or CML, is a stable glycation adduct of lysine (PubChem CID 123800, C8H16N2O4, 204.22 g/mol, CAS 5746-04-3). It forms both in the body and in heated foods, from glucose and from oxidative fragmentation of sugars and lipids. Its chemical stability is why it is the most frequently quantified AGE marker, in tissue, plasma, urine and food.

Does cooking method change advanced glycation end products in food?

Measurements say yes, substantially. In the 549-item immunoassay database, dry heat increased AGE content by more than 10- to 100-fold above the uncooked state across food categories, while moist heat, shorter cooking times, lower temperatures and acidic ingredients reduced new formation. The mass-spectrometry database independently found the highest levels in high-heat processed nut and grain products and canned meats.

Why do two dietary AGE databases disagree?

Because they measure different things. The 2010 database reports antibody-reactive material in arbitrary kilounits from an ELISA built on a monoclonal anti-CML antibody. The 2016 database reports molar quantities of three named compounds by UPLC-MS/MS. Neither is a rescaled version of the other, and values from one cannot be converted into values from the other.

Is skin autofluorescence a reliable measure of AGEs?

It is a useful non-invasive proxy with clear limits. Validated against skin biopsies, autofluorescence correlated with collagen-linked fluorescence, pentosidine, CML and CEL at coefficients of 0.47 to 0.62, and was validated in non-pigmented skin. Many AGEs are not fluorescent and other skin fluorophores are, so the signal is not specific to glycation.

Do dietary AGEs raise AGE levels in blood?

In 450 CODAM participants, higher estimated dietary intake of CML, CEL and MG-H1 was associated with higher free adduct levels in plasma and urine. Associations with the corresponding protein-bound plasma adducts were not statistically significant. Since the protein-bound pool is the one implicated in tissue cross-linking, that null result is as important as the positive one.

What is glucosepane and why does it matter?

Glucosepane is a lysine-arginine cross-link derived from glucose (PubChem CID 135565852, C18H32N6O6, 428.5 g/mol). Measured by isotope dilution mass spectrometry in human skin collagen, it reached roughly 2000 pmol/mg at age 90 in non-diabetic controls and 5000 pmol/mg in diabetes, two orders of magnitude above other quantified cross-links in the same samples.

References

Research use only. Nonsensia Lab supplies analytical reference standards for laboratory and research applications. This article is published for scientific and educational purposes. It is not medical advice, it does not describe any use in humans, and nothing in it should be read as a recommendation to administer any substance to a person or animal.

Filed under: Longevity Science

This article is part of our guide to The Biology of Ageing: Hallmarks, Chemistry and What the Evidence Supports.

Glycation and Advanced Glycation End Products