Conversion of arginine into ornithine by advanced glycation in senescent human collagen and lens crystallins.

Long lived proteins undergo age-related postsynthetic modifications that destabilize them by altering their conformation, charge, and helicity, thereby enhancing their resistance toward proteolysis and propensity to aggregate. The unexpected finding of substantial amounts of ornithine, the nonprotein amino acid, and decarbamidation product of arginine in acid hydrolysates of lens crystallins and skin collagen led us to investigate its source and mechanism of formation. In order to exclude ornithine formation as an artifact of acid hydrolysis, proteins were reductively alkylated with formaldehyde to convert ornithine to dimethyl-ornithine. The proteins were assayed for carboxymethyl-ornithine and glycated ornithine ("furornithine") by liquid chromatography coupled to electrospray ionization mass spectrometry. Ornithine in acid hydrolysates of human lens and skin proteins increased from 1 to 15 nmol/mg protein from ages 10 to 90 years, whereas dimethyl-ornithine increased from 0.5 to 15 and from 0 to 5 nmol/mg protein, respectively. Carboxymethyl-ornithine and furornithine increased with age in lens and skin from approximately 0 to 60 and 0 to 180 pmol/mg protein, respectively. In collagen, ornithine was elevated above levels of nondiabetic controls only when both diabetes and end stage renal disease were present. The age-related increase of these modifications provides evidence for substantial in vivo formation of ornithine in aging human tissue proteins. The mechanism of ornithine formation is not known, but data suggest that arginine-derived advanced glycation end products might serve as precursors for the in vivo conversion of ornithine from arginine.

Reducing sugars react with long lived proteins in vivo, resulting in molecular adducts and cross-links collectively referred to as advanced glycation end products (AGEs), 1 which are thought to be important in diabetic complications and the aging phenomenon (1)(2)(3)(4)(5)(6). Glycation begins with the reducing sugar reacting nonenzymatically with the free amino group of the protein to form an Amadori product (1). The latter can undergo a series of complex reactions involving both oxidative and nonoxidative chemistry resulting in AGE formation (7). Although the ⑀-amino group of lysine bears primary importance in its reactivity with reducing sugars in forming Amadori products and AGEs, evidence suggests that the guanidino group of arginine is also actively involved in these reactions. Arginine can react with the oxoaldehydes methylglyoxal (MG), glyoxal (G), and 3-deoxyglucosone (3DG), to produce the hydroimidazolones MG-H1, G-H1, and 3DG-H1, respectively ( Fig. 1) (8,9). Likewise, MG can also react directly with arginine to form argpyrimidine ( Fig. 1) (10). Additionally, arginine is also able to undergo reaction with the lysine-derived Amadori products of pentoses and glucose to form the cross-links pentosidine (11) and glucosepane (12). Similarly, MG, G, and 3DG can react with lysine and arginine to form the imidazoline cross-links MODIC, GODIC, and DOGDIC, respectively ( Fig. 1) (12)(13)(14).
In the present research, the unexpected and consistent finding of ornithine in our chromatographic profiles of acid-hydrolyzed insoluble human lens and collagen in association with chronological age led us to investigate its source and mechanism of formation. In vivo, ornithine is formed in the liver by the urea cycle via enzymatic cleavage of arginine into ornithine and urea by arginase (15,16). Similar cleavage occurs in extrahepatic tissues, where ornithine is a precursor for the biosynthesis of proline, spermidine, spermine, and nitric oxide (17). However, ornithine, to our knowledge, has never been described in the primary structure of proteins, since there is neither any codon nor any aminoacyl-tRNA and amino acid synthetase specific for ornithine for transcription, translation, and protein synthesis (16). Thus, ornithine is not found in native proteins (16). Because of these issues, a number of interesting questions arose concerning the origin of ornithine in our biological specimens (i.e. as to whether it is truly present in proteins in vivo or whether it is merely an artifact of acid hydrolysis of proteins rich in hydroimidazolones as suggested by Paul et al. (18)).
Thus, the purpose of the present research is 2-fold: (i) to investigate whether ornithine truly exists in the primary structure of long lived proteins and (ii) to investigate its biological significance as to its age-related increase and mechanism of formation in insoluble skin collagen and lens proteins. The major outcome of this study is that it provides the first and unequivocal demonstration that ornithine forms in substantial quantities during aging of collagen and lens crystallins and that it itself becomes glycated and glycoxidized with age.
Overview of Analytical Techniques Used in This Work-For both historical reasons, such as the introduction of new instrumentation during the course of this work, and scientific reasons, such as the stability, sensitivity, and derivatibility of the products assayed, various techniques have been used. To summarize, for most of the experiments described below, high pressure liquid chromatography (HPLC) coupled to electrospray ionization (ESI) mass spectrometry (scan mode n ϭ 2; i.e. MS/MS) with selected reaction monitoring (SRM) has been used for qualitative and quantitative analyses as follows: (i) isobutyl ester derivatives of ornithine, CMO, and DMO in acid hydrolysates of lens and collagen; (ii) isobutyl ester derivative of ornithine in lens digest; and (iii) underivatized furornithine and ornithine levels in acid hydrolysates of the same samples. In several experiments, HPLC ESI-MS (scan mode n ϭ 1) has been used to quantitate ornithine as follows: (i) isopropyl derivative of acid-hydrolyzed lens; (ii) isobutyl derivative of acid-hydrolyzed standards of arginine and argpyrimidine (Fig. 1); (iii) various acid-hydrolyzed standards of AGEs listed in Fig. 1 without derivatization. Except for pentosidine (m/z ϭ 379), all ions measured by MS were in the (M ϩ H) ϩ form.
Tissue Procurement, Donor Information, and Classification of Lenses-For the most part, skin samples used in this study are identical to those described in an earlier study (20). Totals of n ϭ 113 skin samples were from the University Hospitals of Cleveland (Cleveland, OH), n ϭ 6 from the National Disease Research Interchange (Philadelphia, PA), and n ϭ 1 (age 19 years) from the Cuyahoga County Coroner's Office (Cleveland, OH). Totals of n ϭ 46 human lenses were obtained from the National Disease Research Interchange. Categorization as to gender and race, diagnoses of diabetes and renal disease, and primary cause of death were evaluated from each donor's post-mortem records and clinical charts (skin) or donor information and medical history review forms (lens). Additionally, lenses were classified according to the degree of pigmentation into Pirie grade I (n ϭ 19), II (n ϭ 11), III (n ϭ 6), and IV (n ϭ 0) (21). For some lenses, an arbitrary grade 1.5 (n ϭ 7), 2.5 (n ϭ 3) was used when classification between I and II or between II and III, etc., was equivocal.
Preparation of Human Insoluble Skin Collagen-Tissues were stored at Ϫ80°C until use. Briefly, skin samples were defatted (2:1 chloroform-methanol), extracted (1 M sodium chloride and 0.5 M acetic acid), digested (0.1% pepsin in 0.5 M acetic acid at 4°C), and freeze-dried as previously detailed (20). Samples of ϳ5 mg were weighed and processed as described below.
Preparation and Enzymatic Digestion of Human Lens Crystallins-Human lenses were stored frozen at Ϫ20°C until processing. Upon thawing, each lens was immediately homogenized in Chelex-treated phosphate-buffered saline, and the insoluble pellet was obtained by modification of procedures of Nagaraj et al. (22) (see Supplemental Material). The insoluble pellet of lens crystallins obtained was freezedried and stored at Ϫ20°C.
In order to avoid artifactual generation of ornithine during acid hydrolysis, lens crystallins were proteolytically digested. The enzymatic hydrolysis protocol described by Glomb and Pfahler (23) was used with some modification. Freeze-dried samples of ϳ5.5 mg were weighed and sequentially digested at 37°C at pH 7.4 for 24-h consecutive intervals by proteinase K, peptidase, Pronase, and aminopeptidase M (see Supplemental Material).
The overall percentage digestion for each sample was determined by the ninhydrin reaction (24) by dividing total leucine equivalents assayed for each digest by the same upon acid hydrolysis. For the results below, protein concentration was approximated by assuming that every 10 mol of leucine equivalents is equal to 1 mg of protein.
Sample Preparation of Skin Collagen for HPLC-Skin collagen samples were injected onto the HPLC either as isobutyl derivatives or nonderivatized forms (see Supplemental Material). For derivatization, aliquots of acid-hydrolyzed collagen, the equivalent of 50 g of hydroxyproline (ϳ357 g of collagen) were pipetted into borosilicate tubes together with the following deuterated internal standards: 7 nmol of d 6 -ornithine and 98 pmol of d 4 -CML. The content of each tube was dried by a SpeedVac evaporator (Thermo Savant, Holbrook, NY) followed by conversion of amino acids into their isobutyl esters by heating the tubes at 73°C for 1.5 h in the presence of anhydrous 2-butanol acidified with 0.4 M thionyl chloride. The latter reagents were removed by the Speed-Vac set to a high drying rate. The dried residue was reconstituted with 25 l of an injection solvent consisting of 11.5 mM formic acid and 2.75 mM tridecafluoroheptanoic acid (TDFHA) and transferred to a microinjection vial.
In cases where samples were injected in the underivatized form, aliquots of acid-hydrolyzed collagen or lens hydrolysates were pipetted directly into the microvials and spiked with the internal standards as previously described. After drying the samples, the content of each vial was reconstituted with 20 l of the injection solvent followed by injection into the HPLC using the autosampler as described below.
Ornithine Determination in Enzymatic Digests of Insoluble Lens Crystallins-In order to determine the effects of acid hydrolysis on ornithine yield, the enzymatic lens digest was analyzed with and without acid hydrolysis (see Supplemental Material). Briefly, aliquots of digests were acid-hydrolyzed followed by esterification with 2-propanol for HPLC ESI-MS analysis. Ornithine concentration was determined by its parent mass at m/z 175 with correction for the response of the internal standard, d 6 -ornithine, at m/z 181. Additionally, enzymatic digests of lens crystallins were prepared without acid hydrolysis for ornithine assay by HPLC ESI-MS/MS SRM analyses. An aliquot of each digest was spiked with internal standards and derivatized with 2-butanol.
Reductive Alkylation of Insoluble Collagen and Lens Crystallins-In order to prove that ornithine is present in proteins prior to any manipulation, samples were reductively alkylated in order to generate acidstable DMO using the procedure of Jentoft and Dearborn (25). Totals of 3 mg of insoluble lens pellet or 6 -7 mg of insoluble skin collagen were reductively alkylated by continuous rotation overnight at 4°C in tubes containing 83 mM each of sodium cyanoborohydride and formaldehyde in 0.1 M sodium phosphate buffer, pH 7 (see Supplemental Material). The recovered pellet was freeze-dried and acid-hydrolyzed according to previously published procedures (26). Additionally, in order to study the effect of protein unfolding on the yield of DMO, some samples were reductively alkylated in the presence of urea. The procedure was done simultaneously with that described for reductive alkylation, except that 7 M urea was added to the buffer prior to adjusting the pH to 7.
Reduction of Collagen with Borohydride and Cyanoborohydride-Collagen was reduced according to procedures stated above for reductive alkylation, except no formaldehyde was used in the buffer. The purpose of this experiment was to investigate whether precursors of ornithine exist that may be composed of acid-labile reducible chemical bonds. Collagen was prepared by weighing insoluble skin samples from n ϭ 11 different donors of age range 15-90 years. Each sample was weighed into three portions consisting of 6 mg each followed by treating each portion in one of three ways as follows: (i) control, no reduction; (ii) reduction with 0.1 M sodium borohydride; or (iii) reduction with 0.1 M sodium cyanoborohydride.
Quantitative Analyses of Ornithine, CMO, Furornithine, and DMO by HPLC ESI-MS/MS SRM-Liquid chromatography was carried out using a Michrom Magic 2002 HPLC (Michrom BioResources, Auburn, CA). The pump flow at 120 l/min was split 1:12 (column/waste) by the Magic Precolumn Variable Splitter (Michrom), which resulted in a column flow at 10 l/min. Samples of 1-2 l were injected onto a 15 cm ϫ 0.32 mm, 5-m Discovery BIO Wide Pore C18 capillary type column (Supelco; Sigma) using an Alcott model 718AL autosampler (Alcott Chromatography, Norcross, GA). The eluate from the column was transferred to the Finnigan LCQDuo Mass Spectrometer (Thermo Electron, San Jose, CA) as detailed elsewhere (see Supplemental Material).
Solvents used for HPLC were as follows: solvent A, water; solvent B, 70% acetonitrile in water (Burdick & Jackson Certified Solvents, Honeywell, Inc., Muskegon, MI). For ornithine, CMO, and DMO assays, both solvents contained 11.5 mM formic acid and 0.25 mM TDFHA. For the furornithine assay where samples were not derivatized, the TDFHA concentration in solvent A was increased to 1.5 mM to maintain peak sharpness and resolution. The use of TDFHA as an ion pairing agent for HPLC separations was adapted from the work of Chaimbault et al. (27).
Programming of the MS for assays of ornithine, CMO, furornithine, and DMO, including parent and product ions, can be found in the figure legends and the Supplemental Material. A detailed description of the system controls for the LC/MS can also be found in the Supplemental Material.
Preparation of AGE-BSA from the Incubation of BSA with Reducing Sugar-Bovine serum albumin (BSA, Fraction V, heat shock, fatty acid-free; Roche Applied Science) was incubated at 15 mg/ml in 0.05 M sodium phosphate buffer, pH 7.4, for 7 days at 37°C in 10-ml volumes with 20 mM each of the following sugars: D-ribose, DL-glyceraldehyde, glyoxal, and methylglyoxal. In addition, BSA was incubated similarly with 200 mM D-glucose for either 3 or 82 days (see Supplemental Material).
Reductive Alkylation, Acid Hydrolysis, and Assay for Ornithine and DMO in AGE-BSA-AGE-BSA preparations were reductively alkylated as described above with minor modifications. Samples of ϳ7 mg each of AGE-BSA preparation, but also unmodified BSA as a control, were reductively alkylated in duplicates. Further procedures for acid hydrolysis and assays of these preparations are detailed in the Supplemental Material.
Acid Hydrolysis of Arginine and a Series of AGE Standards for Screening of Precursors for Ornithine Formation-A series of experiments were designed to determine whether ornithine could be generated from arginine and arginine-derived AGEs shown in Fig. 1. In the first experiment, aliquots of L-arginine (SigmaUltra) consisting of 0.25, 0.5, 1.0, and 1.5 nmol were each acid-hydrolyzed in a total of 300 l of 6 N HCl (see Supplemental Material). In a second experiment, aliquots of standards (Table I)  Statistical Methods-Statistical analyses were done using procedures previously described by us (20,26,28) using SPSS software (Graduate Pack version 11.5, SPSS Inc., Chicago, IL). In these analyses, race, gender, diabetes, and renal failure were coded with indicator variables as follows: race, 1 ϭ Caucasian, 2 ϭ African-American; gender, 1 ϭ male, 2 ϭ female; diabetes, 0 ϭ absence of diabetes, 1 ϭ type 1, 2 ϭ type 2; renal failure, 0 ϭ absence of chronic renal disease or end stage renal disease (ESRD), 1 ϭ chronic renal disease, 2 ϭ ESRD, as previously defined (20). For skin, there were n ϭ 68 Caucasians and n ϭ 45 African-Americans. All lens samples were from Caucasians.

RESULTS
The products assayed in this study and their presumed chemical relationships are summarized in Fig. 2, whereby any of the arginine-AGEs in Fig. 1 is regarded as a potential precursor for ornithine formation in vivo or during acid hydrolysis.
Ornithine, CMO, and Furornithine Levels in Acid-hydrolyzed Human Insoluble Skin Collagen-Ornithine, glycated ornithine (furornithine), and CMO were measured in acid hydrolysates of insoluble skin collagen by HPLC ESI-MS/MS SRM. Amino acids were converted into their isobutyl ester derivatives, as detailed under "Experimental Procedures," whereupon ornithine eluted at ϳ25-26 min in the chromatogram (Fig. 3). Skin levels followed as a function of age, diabetes, and renal failure are shown in Fig. 4A. Ornithine increased linearly (p Ͻ 0.0001) with age in nondiabetic subjects without renal failure and reached ϳ 4 nmol/mg collagen at age 90 years (Fig.  4A). Stepwise multivariate linear regression analyses revealed that age (p Ͻ 0.0001) and renal failure (p ϭ 0.035) were selected as predictors for ornithine formation in skin with the equation, ornithine ϭ 1.4 ϩ 0.03(age) ϩ 0.458(renal failure), whereby renal failure included both chronic renal disease and ESRD as defined under "Statistical Methods." For further investigation of the effects of renal failure on ornithine levels in skin collagen, levels were age-adjusted to 50 years and categorized into nine different groups according to the presence or absence of diabetes and renal failure as previously published (20); i.e. type 1 versus type 2 diabetes; chronic renal failure versus end stage renal disease (ESRD) (data not shown). Statistical comparisons using the Mann-Whitney test showed that ornithine was elevated above levels for nondiabetic control subjects without renal failure, but only when both diabetes and ESRD were present (p ϭ 0.022).
Since ornithine is reportedly not found in native proteins (16), further evidence for its presence in the primary structure of collagen was sought. Based on the structural homology between lysine and ornithine, we hypothesized that if ornithine was actually present in intact protein and not an artifact of acid hydrolysis, then the Amadori product, ␦-fructosyl-ornithine, and the AGE, CMO, should be present in senescent proteins. In analogy to the conversion of fructosyl-lysine to furosine and CML, we expected to detect "furornithine" and CMO (Fig. 2) in the acid hydrolysate, but in lower quantities compared with furosine and CML, since furornithine and CMO are secondary modifications. Thus, quantitative analysis of these ornithinederived AGE modifications, furornithine and CMO (Fig. 2), would be considerably more challenging and require assays of higher sensitivity than furosine and CML. On the other hand, the finding that ornithine is a major age-related modification with levels up to 7 nmol/mg collagen in some skin samples (Fig.  4A) suggested that even at an expected yield at 1% of total ornithine, sufficient sensitivity for the detection of ornithine modifications would be achieved.
Furornithine (i.e. the acid-hydrolyzed product of glycated ornithine) (Fig. 2) was measured in the acid hydrolysate of insoluble skin collagen samples by HPLC ESI-MS/MS SRM without prior derivatization. Furornithine eluted at ϳ21.9 min (Fig. 5) (i.e. after ornithine (RT ϳ19.5 min) but before furosine (RT ϳ22.5 min)). Results showed that levels significantly (p Ͻ 0.0001) increased from 0 to 175 pmol/mg collagen between the ages of 8 and 90 years (Fig. 4C). Levels were significantly (p ϭ 0.022) elevated by diabetes but not renal failure (p ϭ 0.46). These three factors, including gender and race, were further examined by stepwise multivariate linear regression analyses. Both age (p Ͻ 0.0001) and diabetes (p ϭ 0.013) were selected as predicting factors for furornithine formation in skin with the equation, ornithine ϭ 15 ϩ 0.8(age) ϩ 11(diabetes). Levels were age-adjusted to 50 years, and the effect of diabetes was further investigated by analysis of variance and the Student-Newman-Keuls post hoc test. The results showed that diabetes was still a significant (p ϭ 0.007) factor even after adjustment for age. Levels were significantly (p Ͻ 0.05) elevated for individuals with type 2 diabetes versus nondiabetic controls by the Student-Newman-Keuls test (data not shown).
Ornithine, CMO, and Furornithine Levels in Acid-hydrolyzed Insoluble Human Lens Crystallins-Ornithine significantly (p ϭ 0.0012) increased with age in acid-hydrolyzed lens crystallins from nondiabetic subjects without renal disease (Fig.  6A). Univariate regression analysis showed significant effects due to age (p ϭ 0.013) and diabetes (p ϭ 0.044), but not gender (p ϭ 0.92) and Pirie grade (p ϭ 0.11). Multivariate analysis consisting of ornithine levels versus age and diabetes showed that the bivariate model was significant (p ϭ 0.025), but each contributing factor within this model was not (i.e. age, p ϭ 0.068; diabetes, p ϭ 0.26). Furthermore, stepwise regression analysis selected age (p ϭ 0.013), but not diabetes (p ϭ 0.82), Pirie grade (p ϭ 0.63), and gender (p ϭ 0.99). Last and surprisingly, levels were positively (r ϭ 0.63) and significantly (p ϭ 0.004) correlated with age for Pirie grade I (i.e. the least pig-mented lenses) (Fig. 6A). No such correlation was found for Pirie grades ϾI (r ϭ 0.21, p ϭ 0.31), possibly due to a larger variability in these groups.
CMO in lens crystallins (Fig. 6B) was not associated with any factor. There was a nonsignificant (p ϭ 0.091) trend for levels to increase with age in nondiabetic subjects, as shown in  Furornithine in lens crystallins increased significantly (p ϭ 0.04) with age in nondiabetic subjects (Fig. 6C). When all data were collectively analyzed, there were significant effects due to age (p ϭ 0.029) and Pirie grade (p ϭ 0.028) but not gender (p ϭ 0.67). The effect of diabetes approached statistical significance (p ϭ 0.066). Stepwise regression analysis failed to select Pirie grade as a significant (p ϭ 0.77) factor, suggesting that Pirie and age were intercorrelated. This latter association is well known and was confirmed by Spearman's correlation analysis, which showed that Pirie grade is strongly associated with age (r ϭ 0.63, p Ͻ 0.0001).
Overall, these results suggest that more samples would be needed to increase the statistical power of the tests as reflected by the borderline significance of diabetes and the lack of association of ornithine with lens samples for Pirie grade ϾI.
Ornithine Levels in Enzymatic Digests of Insoluble Human Lens Crystallins-Similar studies (data not shown) were also carried out with insoluble human lens crystallins that were sequentially digested by four different enzymes. This digestion protocol resulted in an average of 65% of the lens protein being digested into amino acids as determined by ninhydrin. Ornithine measurements in these digests showed levels nonsignificantly (p ϭ 0.17) increased with age by univariate regression analysis. All other factors, including diabetes (p ϭ 0.69), Pirie grade (p ϭ 0.37), and gender (p ϭ 0.24), were nonsignificant. However, since it was noted that more variation occurred in ornithine levels for lens samples of Pirie grade I, these data (n ϭ 19) were subsequently removed from the analysis. Reanalysis of the remaining levels (i.e. Pirie grade ϾI, n ϭ 27) showed that ornithine significantly (p ϭ 0.003) increased with age (r ϭ 0.55) (data not shown). These results suggest incomplete cleavage of ornithine from proteins in digests of insoluble lens crystallins and perhaps more resistance to enzymatic digestion, particularly in the diabetic lens samples. Surprisingly, more ornithine was recovered from these digests (i.e. up to 30 nmol/mg protein) in comparison with acid hydrolysis (i.e. about 15 nmol/mg protein), suggesting that the digestion procedure partly converted arginine or arginine-AGEs into ornithine.

The Effect of Chemical Reduction by Borohydride on Measured Ornithine Levels in Collagen-
The experiments so far have unequivocally demonstrated the presence of fructosyl ornithine and CMO in collagen and lens crystallins. However, uncertainty persisted as to how much ornithine was being artifactually generated either by enzymatic or acid hydrolysis. In particular, one question was whether a borohydride-or cyanoborohydride-reducible labile precursor served as ornithine source during processing. To investigate this question, insoluble skin collagen samples were reduced with either sodium borohydride or cyanoborohydride. Ornithine levels in the acid hydrolysates were compared with those from unreduced samples (n ϭ 11 donors/group). The results showed that ornithine levels increased significantly (p ϭ 0.008) with age in both reduced and nonreduced samples. However, the effect of reduction by either agent was insignificant (p ϭ 0.76), suggesting that the ornithine formation pathway does not involve reducible chemical bonds (data not shown).

The Effect of Reductive Alkylation of Collagen and Lens Crystallins on Levels of DMO-
In all studies so far, regression lines for ornithine did not go through the zero point at age "zero," in contrast to those for furornithine and CMO (Figs. 4 and 6), again suggesting that some ornithine was being generated during processing of the protein. In order to unequivocally address the question of whether and how much ornithine was actually present in lens crystallins and collagen prior to any treatment, the intact protein was reductively alkylated with formaldehyde in order to convert the ornithine ␦-amino group into the acid-stable dimethyl adduct (i.e. DMO). An HPLC ESI-MS/MS SRM assay was developed for detection of DMO derivatized as isobutyl esters. As shown in Fig. 7, DMO eluted as a single peak at ϳ0.4 min after ornithine (RT ϳ25.1-25.4 min) in skin collagen samples. The peak area of DMO assayed in a reductively alkylated skin sample from an 89-year-old donor is larger than that from a 15-year-old. DMO was not detected in protein samples without reductive alkylation at any time during these analyses. DMO levels were modeled as linear functions for both collagen (Fig. 8A) and lens (Fig. 8B). Levels significantly increased in both insoluble human skin collagen (p Ͻ 0.0001) and lens crystallins (p ϭ 0.004), reaching up to ϳ5 and 8 nmol/mg protein, respectively. In addition, both lines intercepted approximately at the point of origin as evidenced by the y intercept at age 0: collagen (controls), Ϫ0.16 Ϯ 0.31; lens, 0.042 Ϯ 1.7 (Fig. 8).
Finally, arguing that the reductive alkylation reaction may not have hit all existing ornithine residues due to protein folding and hiding of residues, selected samples of collagen were denaturated using 7 M urea prior to the reaction. A small but significant (p ϭ 0.037) increase of about 10% in recovery of DMO was noted in urea-treated skin samples versus untreated controls (Fig. 8A).
The overall conclusion from the comparison of the data above ( Fig. 8) with that of Figs. 4 and 6 is that acid hydrolysis and protein manipulation apparently contributed little to total ornithine levels.
Mechanism of Ornithine Formation: The Role of Spontaneous versus Acid-catalyzed Conversion of AGEs into Ornithine-Further studies were undertaken to address the role of argininederived AGEs in ornithine formation. Various AGE-protein preparations were prepared by incubation of BSA at 37°C for 7 days with 20 mM each of ribose, glyceraldehyde, glyoxal, and methylglyoxal. In addition, the effects of both short and long Fig. 3). Parent ion for furornithine (Furorn) at m/z ϭ 241 (M ϩ H) ϩ was fragmented into product ion at m/z ϭ 126 for monitoring. Samples in 1-l volumes containing ϳ18 g of underivatized acid-hydrolyzed collagen spiked with 283 pmol d 6 -ornithine (Orn) and 4 pmol of d 4 -CML were injected onto the column as described in the legend to Fig. 3. term glycation on ornithine formation were investigated by incubating BSA with 200 mM glucose for either 3 or 82 days. In order to verify ornithine formation in these preparations, aliquots were reductively alkylated with formaldehyde before acid hydrolysis to convert putative ornithine into DMO. Subsequently, the preparations were acid-hydrolyzed, and DMO levels were measured as isobutyl esters. Fig. 9A indeed shows that ornithine is present in substantial quantities in acid hydrolysates of AGEs, whereby methylglyoxal was the most active precursor (p Ͻ 0.0001 versus control) followed by glyceraldehyde (p Ͻ 0.001) Ϸ glyoxal (p Ͻ 0.001) Ͼ ribose (p Ͻ 0.05) Ϸ 82-day glucose (p Ͻ 0.05) Ͼ 3-day glucose (p Ͼ 0.05, not significant).

FIG. 5. Reprocessed ion chromatograms (RIC) for assay of furornithine by HPLC ESI-MS/MS SRM (see
Most interestingly, however, a strong indication of ornithine's presence in these AGE-BSA samples prior to acid hydrolysis was found (Fig. 9B). The highest amount of DMO was measured in 82-day glucose (p Ͻ 0.0001 versus control), where levels reached ϳ2 nmol/mg protein followed by glyoxal (p Ͻ 0.0001) Ͼ ribose (p Ͻ 0.001) Ͼ glyceraldehyde Ϸ methylglyoxal (p Ͻ 0.01) Ͼ 3-day glucose (p Ͼ 0.05, not significant). Thus, these results clearly indicate that ornithine can form spontaneously during glycation reactions but that acid hydrolysis of AGEs can generate substantial amounts of ornithine as judged by the 10-fold difference in scale between Fig. 9, A and B.
Finally, several experiments were designed to determine the possible origin of ornithine formation during acid hydrolysis. The results showed that the commercial source of ultrapure arginine was contaminated with ornithine. However, after correction for this factor, there was no net conversion of arginine into ornithine over concentrations examined (data not shown). We then sought to address the question of which AGE contributes most ornithine in an acid hydrolysate. Each of the arginine-based AGEs listed in Fig. 1 was acid-hydrolyzed, and recovery analysis was performed. As shown in Table I, pentosidine, argpyrimidine, MG-H1, and GH-1 generated the most ornithine, whereby we estimated based on the data from Ahmed et al. (29) that MG-H1 and G-H1 would generate most of the artifactual ornithine in a hydrolysate from a 90-yearold lens. Since MG-H1 and G-H1 levels in collagen are not known, this estimation cannot be made. Most interestingly, however, the sum of artifactual ornithine formed in hydrolysate from the lens is still less than 1 nmol/mg (i.e. almost 8 times lower than total ornithine or ornithine that can be measured as DMO).
From these data, it can be concluded that a rough estimate of ornithine in biological proteins with about 15% error can be made from acid hydrolysates without prior derivatization into DMO. The other surprise was that 40 and 75% of MG-H1 and GH-1, respectively, can be recovered from a 6 N HCl hydrolysate after 16 h at 110°C. This result, however, did not hold true for glucosepane, which was completely destroyed (Table I). DISCUSSION This study provides, to our knowledge, the first evidence for the presence of ornithine and its glycation products fructosylornithine and CMO in biological proteins. The data strongly suggest that ornithine formation in senescent proteins is the result of chemical conversion of arginine into an AGE product that subsequently spontaneously degrades into ornithine. As a lysine analogue, ornithine is then available to undergo the same modifications as those that affect lysine residues. Amid these conclusions, a number of observations need to be taken into consideration.
The first and most important question was to precisely understand the contribution of acid hydrolysis on ornithine formation. Paul et al. (18) reported the presence of ornithine in the amino acid analysis chromatograms of acid hydrolysates of MG-modified proteins. They attributed this observation to the breakdown by acid hydrolysis of the hydroimidazolones (e.g. MG-H1 and G-H1) (Fig. 1), characterized as the major AGE adducts in the in vitro reaction between ribose and collagen (18). This proposition led us to perform extensive studies on the origin of ornithine in our analytical samples.
Several data unequivocally support the notion that most ornithine measured forms in situ during protein aging and is not simply an artifact of acid hydrolysis. First, furornithine, the acid-catalyzed rearrangement product of glycated ornithine (Fig. 2), significantly increased with age in both insoluble collagen (p Ͻ 0.0001) and lens (p ϭ 0.029), whereas CMO signif- icantly (p Ͻ 0.0001) increased in insoluble collagen, suggesting that ornithine must be present in intact protein before acid hydrolysis. Second, the derivatization of the intact proteins into DMO prior to hydrolysis revealed that the latter significantly increased with age in both collagen (p Ͻ 0.0001) and lens (p ϭ 0.005). Third, DMO was present in reductively alkylated AGE-modified BSA samples, implying that incubation of the proteins under mild conditions with glycating agents suffices to convert substantial quantities of arginine into ornithine. Expressed as a percentage of arginine residues, ornithine level in Figs. 4A and 6A represents an estimated 1.1% modification (mol/mol) in collagen and 1.5% modification in lens crystallins in a 90-year-old person. Finally, ornithine was also detected in enzymatic digests of insoluble lens crystallins and significantly increased with age (p ϭ 0.009) for all data excluding those of nondiabetic Pirie I (not shown).
Thus, whereas our data both unequivocally demonstrate that ornithine is present in senescent proteins and can form during acid hydrolysis of AGEs, the next question was what percentage of ornithine in the acid hydrolysates of biological samples was generated during the procedure. A strong significant (p Ͻ 0.0001) correlation between ornithine and DMO levels within FIG. 7. Chromatograms for assay of DMO in insoluble human skin collagen. DMO was assayed by HPLC ESI-MS/MS SRM using d 6 -ornithine (Orn) as the internal standard (see Fig. 3). DMO was converted into isobutyl esters by derivatization with 2-butanol. The parent ion of DMO at m/z 217 was fragmented and monitored for product (daughter) ions at m/z ϭ 172 and m/z ϭ 116. Volumes consisting of 2 l containing ϳ28.6 g of collagen were injected as described in Fig. 3. the same donor (Fig. 10) revealed that for skin donors Ͼ50 years of age, more than 90 Ϯ 20% (S.D.) of measured ornithine levels is present before acid hydrolysis. For lens, more variation was noted in this relationship (Fig. 10), which approached significance (p ϭ 0.093). DMO recovery was 87 Ϯ 32% (S.D.) of that for ornithine.
Years ago, Sletten et al. (30) reported the presence of ornithine in the amino acid chromatograms from acid hydrolysates of urate-binding ␣ 1 -␣ 2 globulin purified from human plasma (31). However, a very large level was measured in this protein (i.e. Ͼ700 nmol/mg protein), which was considerably greater than levels measured for collagen and lens crystallins in this study (i.e. 0 -20 nmol/mg) (Figs. 4A and 6A) and within the same range as those typically measured for the standard physiological amino acids commonly found in the extracellular matrix (ECM) (30). The protein is reportedly of high carbohydrate content (12.1%), and thus artifactual generation of ornithine due to decomposition of glycoprotein during acid hydrolysis was not addressed.
Several hypothetical mechanisms may explain the origin of ornithine in biological samples, whereby decreased protein turnover with age (32) undoubtedly plays a major role in its age-related accumulation. The first and most attractive hypothesis involves enzymatic cleavage of the arginine moiety of specific AGEs. This would represent a physiological repair mechanism to maintain homeostasis by regeneration of the positive charge loss during AGE formation, thus retaining both protein structure and function. This hypothesis was proposed by Sletten et al. (30) as a possible explanation of ornithine in the urate-binding ␣ 1 -␣ 2 globulin as previously discussed. Sletten et al. (30) hypothesized the splitting off of urea from arginine residues in this protein by catalysis involving a specific arginase. Indeed, widespread expression of arginase I in mouse tissues including skin has been reported (17). However, in skin, localization was to the hair follicle, not the ECM (17). Furthermore, arginase expression appears to be localized to tissues with rapid growth and thus high turnover rates (17). In addition, this enzyme has activity toward free arginine in tissues, and it seems rather doubtful that it would also act on intact proteins. Thus, an enzyme-catalyzed mechanism of ornithine formation appears highly unlikely.
A second mechanism would involve the covalent attachment of free ornithine to glycated proteins. This hypothesis is supported by studies that show that free lysine and soluble proteins like albumin can covalently attach to glycated calf skin collagen (33). The strongest argument against covalent incorporation of ornithine as an explanation for our findings is that ornithine would have to be chemically bound via an N-␣-amino group in order to leave the ␦-amino group free for derivatization into DMO and yet allow intact release upon acid hydrolysis. Whereas sugar amide bonds have been described (23), this type of linkage is yet unheard of.
The third and most likely mechanism involves in situ decomposition of arginine-derived AGEs via a nonenzymatic, spontaneous mechanism, similar to the age-dependent deamidation of asparagine and glutamine residues in proteins (34). Some of the AGEs known to accumulate with age in collagen and lens crystallins are shown in Fig. 1. Interestingly, a recent study revealed large amounts of glyoxal and methylglyoxal hydroimi- dazolones in human lens crystallins (29) that could serve as precursors for ornithine (Table I). However, it has recently been reported that the hydroimidazolones are short lived under physiological conditions (29,35). Whether the same phenomenon of decomposition into ornithine occurs spontaneously with age in vivo needs to be shown. In this regard, a puzzling finding is that diseases that favor AGE formation, such as diabetes and ESRD, catalyzed ornithine formation only when present in combination (Fig. 4A). The contribution of dietary sources of ornithine, which might be particularly important in ESRD (36,37), will need to be investigated.
Finally, the appearance of ornithine residues in aging ECM and lens crystallins would have numerous structural and biological consequences. Arginine residues play a critical role in the so-called arginine-glycine-aspartic acid (RGD) sequence of extracellular matrix molecules that is recognized by integrins (38 -41) for biological anchoring of cells. The formation of hydroimidazolones would effectively remove the positive charge and thus impair cell attachment, perhaps resulting in apoptosis (42). Restoration of the charge might prevent this detachment.
In summary, the discovery of substantial quantities of ornithine in the aging lens and ECM raises a number of exciting questions concerning its origin, mechanism of formation, and biological role. The fact that short term incubation of oxoaldehydes with protein fairly rapidly generates protein-bound ornithine warrants investigation on cellular presence and the role of these ornithine-rich proteins, especially the argininerich histones. Most importantly, the role of ornithine formation on protein structure and turnover should be pursued with high priority.