Production of N^ε-(Carboxymethyl)Lysine Is Impaired in Mice Deficient in NADPH Oxidase: A Role for Phagocyte-Derived Oxidants in the Formation of Advanced Glycation End Products During Inflammation

Crystalline catalase (from bovine liver, thymol-free), ribonuclease A (RNase A; lyophilized from bovine pancreas, code RAF), and glycolaldehyde were purchased from Boehringer Mannheim Biochemicals (Indianapolis, IN), Worthington Biochemical (Freehold, NJ), and Fluka Chemical (Rononkoma, NY), respectively. [d₄]CML and CML were provided by Dr. S. Thorpe (University of South Carolina) (8). CGD (Cybb^−/−) mice (29) and wild-type (Cybb^+/+) littermates in a mixed C57BL/6J and 129-SV genetic background were provided by Dr. R. LeBoeuf (University of Washington). CGD mice (28) and wild-type mice in a C57BL/6J genetic background were provided by Jackson Laboratories (Bar Harbor, MA). All other materials were purchased from Sigma Chemical (St. Louis, MO) unless otherwise indicated.

Mice were injected intraperitoneally with 1 ml 4% sterile thioglycollate broth to recruit neutrophils into the peritoneal cavity (28). Peritoneal inflammatory cells were harvested 16–20 h later by lavage with 10 ml ice-cold medium A (magnesium-, calcium-, phenol-, and bicarbonate-free Hank’s balanced salt solution [Gibco-BRL] supplemented with 100 μmol/l diethylenetriamine pentaacetic acid [DTPA], pH 7.2). DTPA was included in medium A to inhibit metal-catalyzed oxidation reactions. Cells were collected by low-speed centrifugation (400g for 5 min), resuspended in ice-cold medium A, and immediately used for experiments.

Neutrophils were isolated from CGD (Cybb^−/−) and wild-type (Cybb^+/+) mice in the C57BL/6J background (28). Freshly harvested cells (2 × 10⁶/ml) were incubated at 37°C in medium A supplemented with 1 mmol/l l-serine. The cells were stimulated with 200 nmol/l phorbol myristate acetate (PMA) and maintained in suspension by intermittent inversion. The reactions were stopped at the indicated times by pelleting the neutrophils by centrifugation (10 min × 8,000g at 4°C). Supernatants were immediately derivatized and analyzed for glycolaldehyde by high-performance liquid chromatography (HPLC).

Glycolaldehyde was derivatized with 3-methyl-2-benzothiozolinone hydrazone hydrochloride (MBTH) (30) by incubating 200 μl supernatant for 20 min at room temperature with 568 μl sodium phosphate buffer (5 mmol/l, pH 7.0), 25 μl HCl (6.0 N), and 132 μl MBTH (155 mmol/l). The azine derivative was reacted with 75 μl FeCl₃ (370 mmol/l) and incubated for >10 min at room temperature. The resulting blue-colored derivative was stable for at least 5 h. Aldehyde production was quantified by HPLC using a standard curve generated with MBTH derivatives of authentic glycolaldehyde.

Reverse-phase HPLC was performed on a C₁₈ column (5 μm resin, 4.6 × 250 mm, Beckman uPorasil) at a flow rate of 1 ml/min. The column was equilibrated with 70% solvent A (0% methanol and 0.1% trifluoroacetic acid, pH 2.5) and 30% solvent B (100% methanol and 0.1% trifluoroacetic acid, pH 2.5). The column was eluted with a discontinuous gradient of solvent B: 30% solvent B for 3 min, 30–60% solvent B over 2 min, then 60–100% solvent B over 20 min. The eluted MBTH derivatives were detected by absorbance at 598 nm (30).

Neutrophils were isolated from CGD (Cybb^−/−) and wild-type (Cybb^+/+) mice in the C57BL/6J background (28). To modify RNase A, freshly harvested cells (2 × 10⁶ cells/ml) were added to medium B (50 mmol/l sodium phosphate, 100 mmol/l NaCl, 4 mmol/l KCl, and 100 μmol/l DTPA, pH 7.2) supplemented with RNase A (1 mg/ml) and 260 μmol/l l-gluatamate, 210 μmol/l l-alanine, 200 μmol/l l-serine, 175 μmol/l glycine, 165 μmol/l l-valine, 100 μmol/l l-proline, and 100 μmol/l l-lysine (31). Neutrophils were activated with 200 nmol/l PMA and incubated for 60 min at 37°C, with intermittent inversion. After removing the neutrophils by centrifugation (10 min × 8,000g), the mixture was incubated for the indicated amount of time at 37°C. Reactions were terminated by freezing at −20°C. For CML analysis, the samples were dried under vacuum by centrifugal evaporation, 32 pmol [d₄]CML and 45 nmol l-[d₈]lysine were added, and the protein pellet was hydrolyzed with 6 N HCl for 24 h at 110°C under N₂ (32).

Peritoneal inflammatory cells were isolated 24 h after thioglycollate injections from age-matched wild-type (Cybb^+/+) and CGD (Cybb^−/−) mice in a homogeneous C57BL/6J or mixed C57 BL6/6J and 129-SV genetic background and immediately exposed to 500 mmol/l NaBH₄ in 200 mmol/l borate buffer (pH 9) to reduce Schiff bases (32). After an overnight incubation at 4°C, the samples were dialyzed against distilled water for 2 days at 4°C and then delipidated three times using water:methanol:water-washed ether (1:3:7 vol/vol/vol). After the samples were dried by centrifugal evaporation, internal standards were added, and the protein pellet was hydrolyzed with acid as described above.

Amino acid hydrolysates were dried in vacuo, resuspended in 1 ml of 1% trifluoroacetic acid, and immediately passed over a solid-phase C₁₈ extraction column (1-ml Supelclean LC-18 SPE tubes; Supelco, Bellefonte, PA) that had been equilibrated with methanol and 1% trifluoroacetic acid. The column was washed with two 1-ml aliquots of 1% trifluoroacetic acid, and the flow-through and two washes were pooled and dried by centrifugal evaporation (32). The n-propyl pentafluoropropionyl derivatives of CML and lysine were prepared and subjected to isotope-dilution gas chromatography/mass spectrometry (GC/MS) as described (25,33). CML, [d₄]CML, l-lysine, and l-[d₈]lysine were quantified using selected ion monitoring of ions of charge-to-mass ratios (m/z) 560, 564, 460, and 468, respectively, as described (25,32). Selected ion monitoring GC/MS was performed on a Finnigan SSQ equipped with a 12-m DB-1 capillary column (PJ Cobert, St. Louis, MO) (0.2 mm ID, 0.33 μm film thickness). Samples were injected with a 20:1 split with methane as the reagent gas. The injector port and detector were kept at 250°C, with the source at 200°C. To analyze CML and lysine, the initial GC column temperature of 120°C was held for 3 min and then increased from 120 to 300°C at 20°C/min. The temperature was then held at 300°C for 2 min.

Previous in vitro studies indicate that one mechanism for CML formation in model systems involves the conversion of l-serine into glycolaldehyde by a reaction that is inhibited by catalase, suggesting a requirement for H₂O₂ (15,21). Mouse neutrophils isolated from wild-type mice and activated with PMA in a balanced salt solution converted l-serine into a compound that reacted with MBTH, a derivatizing agent that forms a stable-colored adduct with aldehydes (30). The derivative comigrated with derivatized authentic glycolaldehyde on reverse-phase HPLC (Fig. 1). Other closely related carbonyl compounds (glyceraldehyde, glyoxal, acrolein, formaldehyde, and acetaldehyde) exhibited HPLC retention times that were distinctly different from that of glycolaldehyde (data not shown). Aldehyde synthesis by the cells required l-serine. In contrast, neutrophils isolated from CGD mice and activated with PMA in the presence of l-serine failed to generate glycolaldehyde (Fig. 1). These observations suggest that conversion of l-serine to glycolaldehyde requires oxidants derived from NADPH oxidase.

When we monitored glycolaldehyde formation by neutrophils isolated from wild-type mice, we observed a short lag phase followed by a linear progress curve that reached a plateau by 90 min (Fig. 2A). Aldehyde production increased with rising concentrations of l-serine (Fig. 2B) and was optimal around neutral pH (Fig. 2C). In contrast, production of free HOCl by myeloperoxidase is optimal under acidic conditions (24), suggesting that other factors affect glycolaldehyde production. Glycolaldehyde formation required activation of the cells with PMA (Fig. 3) and was inhibited by cyanide, implicating a heme protein such as myeloperoxidase in the reaction. Superoxide dismutase (which enhances dismutation of superoxide to H₂O₂ 500-fold at neutral pH) (34) increased the product yield (Fig. 3), perhaps by increasing the availability of H₂O₂ (35). Alternatively, superoxide dismutase may have prevented superoxide from inactivating myeloperoxidase (36).

In contrast to wild-type neutrophils, neutrophils isolated from CGD mice and activated with PMA failed to convert l-serine into glycolaldehyde under any of the conditions tested (Fig. 3). Moreover, catalase, a peroxide scavenger, inhibited glycolaldehyde production by neutrophils isolated from wild-type mice (Fig. 3). Collectively, these observations indicate that neutrophils use the H₂O₂ produced by the NADPH oxidase when they generate glycolaldehyde, a reactive α-hydroxyaldehyde that promotes AGE formation in vitro (15).

To determine whether activated phagocytes use intermediates derived from NADPH oxidase to produce CML on model proteins, we isolated neutrophils from wild-type mice and then stimulated them with PMA in a physiological salt solution supplemented with RNase A. To mimic a more physiological milieu, plasma concentrations of the seven most common amino acids (260 μmol/l l-gluatamate, 210 μmol/l l-alanine, 200 μmol/l l-serine, 175 μmol/l glycine, 165 μmol/l l-valine, 100 μmol/l l-proline, and 100 μmol/l l-lysine) were included in the medium (31). After a 1-h incubation, the cells were removed by centrifugation and the medium was incubated for 72 h at 37°C. The RNase A was then hydrolyzed with acid, the resulting amino acids were derivatized, and then the derivatives were subjected to analysis by negative-ion electron-capture GC/MS.

The n-propyl pentafluoropropionyl derivative of CML, monitored as ions at m/z 540 (M^·− –2 HF) and 560 (M^·− –HF), was readily detected in the amino acid mixture by selected ion monitoring (Fig. 4). The relative abundance and retention time of the two most abundant ions derived from the compound generated by the activated neutrophils were identical to those of authentic CML (Fig. 4). During selected ion monitoring, the ions derived from d₄-labeled CML eluted slightly earlier than did those of CML, as has been reported for other deuterated internal standards (33). These results indicate that CML is present in acid hydrolysates of a model protein exposed to activated neutrophils.

We next determined the relative importance of oxidants produced by the phagocyte NADPH oxidase to CML production in vitro. CML was quantified in RNase A, which was exposed to PMA-stimulated cells in medium supplemented with plasma concentrations of amino acids using isotope-dilution GC/MS. Neutrophils isolated from wild-type mice markedly increased CML levels (Fig. 5A). Generation of this AGE required stimulation of the neutrophils with PMA, implicating cellular activation and oxidant generation in the reaction (Fig. 5B). Omitting l-serine from the reaction mixture inhibited CML formation by ∼80% (Fig. 5B). Neutrophils isolated from CGD mice produced only low levels of CML (Fig. 5). These results indicate that activated human neutrophils use oxidants generated by their NADPH oxidase to generate CML in a physiological mixture of amino acids. They also indicate that low levels of CML are formed under these conditions by a different pathway independent of NADPH oxidase.

To determine whether oxidants derived from the NADPH oxidase might generate AGEs in vivo, we measured CML levels in inflammatory cells isolated from mice injected intraperitoneally with thioglycollate (Fig. 6). This treatment triggers an inflammatory response that initially recruits neutrophils into the peritoneum (28). Subsequently, the inflammatory cell population evolves into a mixture that contains mostly monocytes and macrophages. This model system therefore mimics many of the cellular events of a classic acute inflammatory response.

Peritoneal inflammatory cells were harvested 24 h after the thioglycollate injections, using wild-type and CGD mice in two genetic backgrounds (age-matched littermates in a mixed 129-SV and C57BL/6J background or age-matched mice in the homogeneous C57BL/6J background). The phagocytes from the C57BL/6J wild-type mice were 78 ± 14% neutrophils, 16 ± 11% macrophages, and 0 ± 1% eosinophils (n = 8). Those from the CGD mice were 76 ± 26% neutrophils, 18 ± 21% macrophages, and 2 ± 2% eosinophils (n = 7). After reduction and delipidation, cellular proteins were hydrolyzed with acid and derivatized. Using isotope-dilution, negative-ion electron-capture GC/MS, we then quantified the n-propyl pentafluoropropionyl derivative of CML. Compared with the samples from the wild-type mice, there was 40% less protein-bound CML in peritoneal inflammatory cells isolated from the CGD mice in both the mixed genetic background (187 ± 53 vs. 109 ± 28 μmol/mol lysine; n = 8 and 5; P < 0.02) and the homogeneous C57BL/J6 background (100 ± 24 vs. 60 ± 8 μmol/mol lysine; n = 7 and 6; P < 0.004). These observations strongly suggest that oxidants derived from phagocyte NADPH oxidase generate CML in vivo.

In the current studies, we used animals deficient in NADPH oxidase (28), the major pathway by which phagocytes produce H₂O₂ (19), to show that oxidants derived from NADPH oxidase contribute to CML formation in vivo. First, we showed that neutrophils isolated from wild-type mice, but not those from mice deficient in NADPH oxidase (CGD mice), converted l-serine into glycolaldehyde. The peroxide scavenger catalase inhibited glycolaldehyde production by wild-type neutrophils. These results indicate that phagocytes generate glycolaldehyde by a reaction dependent on H₂O₂ produced by the phagocyte NADPH oxidase. Second, CML was produced on RNase A when the protein was exposed to activated wild-type neutrophils in a physiological salt solution, which contained the seven most common plasma amino acids (including l-serine). In contrast, only low levels of CML were formed when we exposed RNase A to neutrophils isolated from the CGD mice. This indicates that oxidants generated by NADPH oxidase are necessary for CML formation under physiologically plausible conditions. Finally, we used isotope-dilution GC/MS analysis, a sensitive and specific analytical method (32,33), to quantify CML in neutrophils isolated from animals with chemically induced peritoneal inflammation. There was markedly less CML in neutrophils isolated from CGD mice than in those isolated from wild-type animals. Collectively, these observations indicate that glucose is not the only pathway for the generation of AGEs and that oxidants derived from phagocyte NADPH oxidase represent one pathway for generating CML in vivo.

Previous studies (8,14,37) have shown that CML increases in skin collagen with aging and that the rate of increase is greater in diabetic subjects than in age- and sex-matched control subjects. In euglycemic humans, CML increases at a rate of ∼4 μmol per mole of lysine per year (8). In mice with chemically induced peritoneal inflammation, we found that wild-type neutrophils had acquired ∼50 μmol more CML per mole of lysine than CGD neutrophils. Because the neutrophils were harvested from the peritoneum 24 h after stimulation, these observations suggest that phagocyte-derived oxidants produce CML during acute inflammation at a rate that is ∼4,000-fold greater than that observed in human skin collagen.

Immunohistochemical studies (13,14,26) have shown that high levels of CML and AGEs are present in the cytoplasm of macrophage foam cells of atherosclerotic lesions in nondiabetic animals and humans. These observations suggest that pathways independent of glucose contribute to CML formation in atherosclerosis, a chronic inflammatory disease. Consistent with this proposal, immunohistochemical studies (14) found little difference between the amount of CML in the macrophages of diabetic and nondiabetic human atherosclerotic lesions. A monoclonal antibody to AGEs that was subsequently shown to react predominantly with CML gave similar results (38). Therefore, an AGE-generating pathway that does not involve glucose might operate in the artery wall. One mechanism may involve myeloperoxidase, which colocalizes in part with lipid-laden macrophages in atherosclerotic lesions (27). Elevated levels of protein and lipid oxidation products characteristic of myeloperoxidase, including an adduct derived from protein-bound glycolaldehyde, have been detected in human atherosclerotic tissue and in LDL isolated from lesions (29,39,40). Another mechanism for accumulating AGEs inside macrophages may involve the uptake of AGE-modified matrix proteins or lipoproteins by cell-surface proteins, such as receptors of advanced glycation end products (RAGEs) or scavenger receptors (11). Importantly, high concentrations of oxidants are secreted into the phagolysosome of macrophages (19), and protein-bound lipid oxidation products are found in lysosomal-like structures in macrophage foam cells (41). These observations indicate that NADPH oxidase might contribute to the formation of CML and other AGEs in macrophage foam cells. In future studies, it will be important to determine whether oxidants produced by phagocytes contribute to AGE formation in humans.

It has been proposed (12) that increased levels of reactive carbonyls—“carbonyl stress”—contribute to AGE formation and the long-term development of complications in diabetes and renal failure. One pathophysiologically relevant mechanism involves glucose, which is a carbonyl in its open chain form. Wolf et al. (42) first proposed that reducing sugars participates in metal ion–dependent oxidation reactions that generate superoxide, H₂O₂, hydroxyl radicals, and reactive aldehydes. Moreover, oxidation is critical for the formation of CML (1,6). Thus, glycoxidation reactions appear necessary for generating this AGE in vitro. Another potential pathway involves mitochondria, which generate reactive species. Many lines of evidence suggest that glucose increases the production of reactive species inside cultured vascular cells and that inhibitors of mitochondrial electron transport block this increase (43). Lipid oxidation also produces reactive carbonyl intermediates, such as glyoxal, that promote CML formation (16,17). Moreover, epitopes for protein-bound lipid oxidation products and AGEs colocalize in part in atherosclerotic tissue, suggesting that lipid oxidation contributes to AGE formation during atherogenesis (14,17,38). Glucose and lipid oxidation products have also been implicated in the formation of AGEs in renal failure and diabetic nephropathy (1–4,12).

The H₂O₂ produced by NADPH oxidase might also generate carbonyls and AGEs by other mechanisms. For example, the myeloperoxidase-H₂O₂ system of phagocytes converts tyrosine to oxidizing intermediates (35). Tyrosyl radical generated by this system peroxidizes lipids by reactions that do not require free metal ions (35). This pathway also generates dityrosine cross-links in proteins, and elevated levels of dityrosine have been detected in human atherosclerotic lesions (44), raising the possibility that carbonyls derived from oxidized lipids promote CML formation in vivo (17). Another potential pathway involves the H₂O₂-dependent oxidation of protein-bound Amadori and post-Amadori products (45). Reagent H₂O₂ promotes the formation of CML and pentosidine on glycated collagen in vitro. This pathway may be important physiologically because catalase inhibits the formation of AGEs on collagen in vivo (45). Thus, CML formation at sites of inflammation is likely to involve a complex interplay among oxidants, carbohydrates, lipids, and amino acids.

Our results indicate that oxidants derived from phagocyte NADPH oxidase promote CML formation during acute inflammation in mice. These observations suggest that the generation of reactive intermediates by phagocyte NADPH oxidase contributes to CML formation during atherogenesis. They also support the hypothesis that generation of reactive carbonyls by the oxidation of glucose, lipids, and amino acids is a common mechanism for AGE formation in aging and disease.

This work was supported by National Institutes of Health Grants DK02456 and AG021191.

We thank Dr. Susan Thorpe (University of South Carolina) for providing natural-abundance and isotope-labeled CML, Dr. Rene LeBoeuf (University of Washington) for providing CGD mice, and Jennifer Moreno for excellent technical assistance.