Proteomic Analysis of Arginine Adducts on Glyoxal-modified Ribonuclease*

Accumulation of advanced glycation end-products (AGEs) on proteins is associated with the development of diabetic complications. Although the overall extent of modification of protein by AGEs is limited, localization of these modifications at a few critical sites might have a significant effect on protein structure and function. In the present study, we describe the sites of modification of RNase by glyoxal under physiological conditions. Arg39 and Arg85, which are closest to the active site of the enzyme, were identified as the primary sites of formation of the glyoxal-derived dihydroxyimidazolidine and hydroimidazolone adducts. Lower amounts of modification were detected at Arg10, while Arg33 appeared to be unmodified. We conclude that dihydroxyimidazolidine adducts are the primary products of modification of protein by glyoxal, that Arg39 and Arg85 are the primary sites of modification of RNase by glyoxal, and that modification of arginine residues during Maillard reactions of proteins is a highly selective process.

proteins by glucose is implicated in the pathogenesis of longterm diabetic complications, including vascular and renal disease and blindness.
Glyoxal and glycolaldehyde are products of autoxidation of glucose or glucose adducts to proteins. Other carbohydrates, such as fructose, arabinose, and ascorbate, may also degrade to glyoxal, possibly through intermediate adducts to protein. Glyoxal may also be formed directly during oxidative degradation of polyunsaturated fatty acids (4) and during myeloperoxidase-mediated degradation of serine at sites of inflammation (5). Plasma glyoxal levels are much lower than those of glucose, but glyoxal is a far more reactive carbonyl compound. Normal plasma levels of glyoxal are reported to be 215-230 nM (6 -8) but increase to 350 -470 nM in diabetic subjects (7,8), ϳ400 nM in uraemia (6), and ϳ760 nM in end-stage renal disease (6). Because of its high reactivity, the fraction of glyoxal bound to proteins may significantly exceed the measured glyoxal concentration in plasma.
In previous work (11) on chemical modification of the model protein RNase by glucose, we concluded that CML was formed primarily by oxidation of Amadori adducts of glucose to protein and that free glyoxal, which was also formed in the reaction system, was not a significant precursor of CML. In the present study, we extend this work to analysis of the sites and products of modification of RNase by glyoxal. We show that, like the carboxymethylation of RNase, formation of G-DH and G-H in the RNase-glyoxal incubations is a site-specific process and that Arg 39 and Arg 85 , which are the arginine residues closest to the active site of RNase, are the primary sites of modification of the enzyme. Our results indicate that dicarbonyl compounds react primarily with arginine residues in protein and that there is a high degree of specificity to the modification of both arginine and lysine residues during the Maillard reaction.
Modification of Protein by Glucose or Glyoxal and Preparation of Tryptic Digests-RNase (13.7 mg, 1 mol) was dissolved in 1 ml of a solution of glucose (0.4 M) or glyoxal (1 or 5 mM) in phosphate buffer (0.2 M, pH 7.4) and incubated under air at 37°C for 3, 7, and 14 days (glucose) or 1, 3, and 7 days (glyoxal). The recovered protein was reduced with DTT, derivatized with 4-vinylpyridine, and digested with trypsin (enzyme:substrate ratio of 5:100 (w/w)) at 37°C for 5 h. All samples were prepared in triplicate. The procedures have been published in detail elsewhere (11).
Reaction of RNase-Glyoxal Incubations with HA-Aliquots of the 7-day RNase-1 mM glyoxal incubation (40 l, 0.04 mol protein) were mixed with a 10-M excess of HA (carbonyl groups:amino groups ϭ 1:5) in 0.2 M, pH 7.4 phosphate buffer (40 l) containing diethylene-triaminepentaacetic acid (DTPA, final concentration 0.1 mM) and incubated under nitrogen for 1, 3, and 7 days at 37°C. Control incubations used phosphate buffer in place of HA solution. Further aliquots of the 7-day RNase-1 mM glyoxal incubation (40 l) were ultrafiltrated to remove any unreacted glyoxal (11), resuspended in an equivalent volume of 0.4 M, pH 7.4 phosphate buffer (40 l) containing DTPA (final concentration 0.1 mM) and incubated for 1 and 7 days.
Tryptic Digestion of HA Incubations-Incubations were digested using a modification of the procedure reported by Brock et al. (11). Protein (0.5 mg) was diluted into 50 l of 0.1 M MOPS buffer containing 6 M urea and 1 mM EDTA. DTT (0.1 mol) dissolved in MOPS buffer:water (1:3, v/v, 5 l) was added to the protein solution, which was flushed with nitrogen for 60 s prior to incubation at 37°C for 3 h. 4-Vinylpyridine (2.5 mol) mixed in methanol:water (5:2 v/v, 5 l) was added and the protein was derivatized in the dark at room temperature for 1 h. DTT solution (35 l, 3.5 mol) was added to quench the reaction. The sample was diluted in water (355 l). Sequencing grade trypsin (100 g) was dissolved in 100 l of 100 mM HCl, and 25 l was added to the protein solution (enzyme:substrate ratio of 5:100 m/m), which was flushed with nitrogen and incubated at 37°C for 5 h. Digestion was terminated by freezing at Ϫ20°C.
Incubations of CMA with OPD and Amino Acid Analysis-CMA and OPD were each dissolved in 0.2 M, pH 7.4 phosphate buffer to give a final CMA concentration of 5 mM, and a CMA:OPD molar ratio of 1:5. DTPA was added to give a final concentration 0.1 mM. Aliquots were incubated under nitrogen at 37°C for 1, 3, and 7 days. Amino acid analysis was conducted on a divinylbenzene cation-exchange column (3 ϫ 250 mm) (Pickering Labs, Mountain View, CA) with a sodium citrate gradient. Amino acids were quantified by post-column fluorescence using o-phthaldehyde. CMA was quantified with reference to a standard calibration curve (0 -10 nmol CMA).
ESI-LC-MS-Samples were fractionated on an Agilent (Palo Alto, CA) series 1100 liquid chromatograph, coupled to a Micromass (Manchester, United Kingdom) Quattro mass spectrometer (for full scan experiments) or a Micromass Q-TOF mass spectrometer (for peptide sequencing experiments). Separations were conducted using a C 18 column (250 ϫ 2 mm) and a gradient running from 0.1% aqueous acid to ACN with a flow rate of 0.2 ml/min. TFA was used as the ion-pairing reagent on the Quattro mass spectrometer and formic acid on the Q-TOF instrument. The conditions have been described in detail previously (11), and any modifications are given below.
Full-scan Experiments-Major differences in the peptide complement of samples were located by treating the entire total ion chromatogram (TIC) of each sample as a single "peak" and generating a single combined spectrum for the entire chromatogram. In a separate analysis of the data, each predicted peptide (unmodified and modified) was located within the TIC by calculating the masses of the different charged forms, extracting ion chromatograms and confirming the same retention time for each charged form, as described previously (11).
For the RNase-carbonyl incubations, estimates of the fractional modification of each peptide were determined as a percentage of the sum of all charged species of that peptide, divided by the sum of all of the charged species for the modified and unmodified peptides. This involved extracting ion chromatograms for the different charged forms of each peptide from the full-scan data and summing the peak areas for each charged form. The N-terminal or C-terminal unmodified peptide that eluted closest to the corresponding modified peptide was selected for calculations to minimize differences in mass spectrometer response between unmodified and corresponding modified peptides. For example, when Arg 10 is modified, there is no cleavage by trypsin between Arg 10 and Gln 11 . The relevant unmodified peptides are 8 FER 10 ( 8 F-R 10 ) and 11 QHMDSSTSAASSSNYCNQMMK 31 ( 11 Q-K 31 ). Peptides 8 F-R 10 and 11 Q-K 31 elute at 14.5 and 21.0 min, respectively, compared with a retention time of 22.5 min for the uncleaved peptide containing modified Arg 10 . Therefore, peptide 11 Q-K 31 , which eluted closest to the modified peptide, was selected as the unmodified peptide when estimating the extent of modification at Arg 10 . The amount of the unmodified and each monitored modified peptide was then expressed as a percentage of the peak area for the Arg 10 peptide group. The retention times of the members of each arginine peptide group that were used for calculations are shown in Table I.
For the HA reactions, peptides were expressed relative to the yield of the unmodified C-terminal peptide, which served as an internal reference, correcting for differences in amount of protein injected (11).
MS/MS for Peptide Sequencing-Tryptic digests were first ana-lyzed by LC-MS with the Q-TOF operating in survey mode. CID data were collected by data-dependent switching to MS/MS mode and collection of TOF data over the mass range 100 -2,000 amu. Fragmentation spectra were interpreted manually.

RESULTS
Lys 41 in the active site of RNase is the primary site of glycation and carboxymethylation of RNase (11). Although the extent of modification of this peptide by 0.4 M glucose and 1 mM glyoxal was similar, CML was identified as the primary product derived from glucose. Although the glyoxal-derived modification was not characterized, we established that glyoxal was not a significant intermediate in the formation of CML. In the present study, we analyzed the products of modification of RNase by glyoxal.
Location of Modified Peptides in the Glyoxal Reactions-As one approach to identifying modified peptides in the RNaseglyoxal reactions, the entire TIC, obtained using the triple quadrupole mass spectrometer for each glucose and glyoxal incubation, was treated as a single chromatographic "peak," yielding a single mass spectrum for each sample (Fig. 2, A and B). These single mass spectra were visually compared in 100 mass unit increments to locate ions that were unique to one or the other sample. Comparison of the spectra of the glucose and 5 mM glyoxal incubations at 7 days, revealed a prominent ion at m/z 982 in the glyoxal spectrum that was absent from the glucose spectrum ( Fig. 2

, A, B, E, and J).
Extraction of the ion chromatogram at m/z 982 from the full-scan data revealed one major peak, eluting at 27.21 min (Fig. 3A). The spectrum of this peak contained ions at m/z 1473, 982, and 736 (Fig. 3B), corresponding to the 2 ϩ , 3 ϩ , and 4 ϩ ions of the same unidentified peptide that was named "Unknown 1" (UK1). Q-TOF analysis established that the proposed 3 ϩ ion was triply charged.
Signals corresponding to the 2 ϩ and 4 ϩ ions, at m/z 1473 and 736, respectively, were also located in the single combined spectrum of the glyoxal incubation but not in that for the glucose incubation (Fig. 2, C, G, H, and L). The m/z ratios of these different charged forms imply a mass of ϳ2942 Da for UK1 that was detected only in the glyoxal, and not in the glucose, reactions.
Modification at an arginine residue in a protein would lead to the formation of a tryptic peptide with a missed cleavage, because trypsin cannot cleave C-terminal to a modified arginine residue. Glyoxal is known to react with the guanidino group of both free and peptide-bound arginine to give products including hydroimidazolones (HIs (12,13)), G-DHs (14), and CMA (10,15). The predicted mass of UK1 corresponded to the mass of the tryptic peptide with a missed cleavage C-terminal to Arg 39 , peptide 38 D-K 61 , where Arg 39 is modified to either G-DH or CMA, which are isobaric compounds. Confirmation of the Amino Acid Sequence of UK1 by MS/ MS-To confirm the amino acid sequence of UK1, the tryptic digest of the 7-day 5 mM glyoxal incubation was analyzed by ESI-LC-MS/MS using the Q-TOF mass spectrometer. Initially, experiments were performed on the 2 ϩ and 3 ϩ ions of peptide 38 D-K 61 with Arg 39 modified to G-DH or CMA, but spectra were much stronger for the 3 ϩ ion and therefore the fragmentation pattern obtained for it was analyzed in detail (Fig. 4). The data were of sufficiently high resolution for the determination of the charge state of each fragment ion and the spectrum was interpreted manually.
As shown in Fig. 4, a series of singly charged y ions (y 3 -y 15 ) and the doubly charged y 20 ion were observed. Some b ions were also observed but were generally less intense than the y ions. Most were doubly charged (b 9 , b 11 -b 19 ) but some singly charged ions (b 6 , b 9 , b 10 , b 12 ) were also seen. The b ions provide evidence that UK1 incorporates Arg 39 and a modification accounting for 58 amu. The y ions establish that the precursor ion contains the sequence of amino acids VHESLADVQAVCS, confirming that UK1 incorporates tryptic peptide 40 C-K 61 . All the MS data presented in Figs. 3 and 4 establish the identity of peptide UK1 to be 38 D-K 61 with either Arg 39 modified to G-DH or CMA, or Lys 41 modified to CML.
We performed numerous additional MS experiments with the aim of generating additional data to confirm the identity of UK1. Both precursor ion experiments for parents of the immonium ion of G-DH or CMA at m/z 187 and multiple reaction monitoring experiments to search for the immonium ion of G-DH or CMA and the b 2 ion at m/z 330, generated from peptide 38 D-K 61 with Arg 39 modified to G-DH or CMA, gave signals at the pertinent nominal masses but the identities were ambiguous due to the limited resolution of the triple quadrupole instrument. For example, the fragment at m/z 187 may also represent an a-type internal ion for DV or a b-type internal ion for AD, both of which were present in the sequence of the modified peptide. Analysis by capillary LC-MS/MS on a newer generation Q-TOF API US instrument using a limited scan range (to maximize sensitivity) and different collision energies and sample concentrations also did not reveal either ion of interest. The peptide was less amenable to analysis by MALDI-TOF/TOF than ESI-MS. The immonium ion of arginine is known to be of very low intensity (16) and it is likely that the immonium ion of the G-DH or CMA adduct of Arg 39 will also be weak.
Discrimination Between G-DH, CMA, and CML-The formation of G-DH is reversible, while CMA and CML are formed irreversibly (14). To investigate whether the modified peptide in the glyoxal incubations was G-DH or CMA or both adducts on Arg 39 or CML on Lys 41 , protein from the 7-day 1 mM glyoxal reaction was incubated with a 5-fold molar excess (based on carbonyl groups) of HA. Parallel incubations were conducted without HA, both with and without prior removal of any unreacted glyoxal. As shown in Fig. 5, the amount of UK1 decreased by 75% over a 7-day period, with a half-life of ϳ2 days.
In contrast, authentic CMA was completely stable in a parallel incubation with OPD for up to 7 days. Because the majority (ϳ75%) of the Arg 39 adduct in the 7-day 1 mM glyoxal incubations was reversed by incubation with HA, we conclude that the peptide contained G-DH rather than the isobaric CMA.
The possibility that UK1 might be peptide 38 D-K 61 with Lys 41 modified to CML was ruled out because CML is extremely stable, e.g. to 6 N HCl at 95°C for 18 h (9). The identification of UK1 as peptide 38 D-K 61 with Lys 41 modified to CML would also require that the peptide contain a missed cleavage site C-terminal to Arg 39 but this unmodified peptide was not detected in tryptic digests of either the glyoxalmodified RNase incubations or of native RNase, so that cleavage of Arg 39 appears to be efficient. We did detect trace levels of peptide 40 C-K 61 with Lys 41 modified to CML but UK1 was present at Ͼ100-fold higher intensity. Thus, UK1 must be peptide 38 D-K 61 with Arg 39 modified to G-DH, in line with earlier reports that dicarbonyls, such as methylglyoxal, modify arginine in preference to lysine residues on protein (13). Fig. 5 also shows that the amount of the unmodified peptide 40 C-K 61 increases with time of incubation with HA. Peptide 38 D-K 61 with Arg 39 modified to G-H was also located within the MS data by extracting ion chromatograms at m/z 1463, 975, and 732 for the 2 ϩ , 3 ϩ , and 4 ϩ charged forms, respectively. This peptide also decreased on incubation with HA, but at a much slower rate than the loss of G-DH, with ϳ80% remaining after 7 days (compared with ϳ25% for G-DH). In control experiments, incubating the 1 mM glyoxal incubations, both with and without prior removal of unreacted glyoxal and in the absence of HA, resulted in no change in the amount of peptide 38 D-K 61 in which Arg 39 was modified to G-DH. Thus, the discharge of G-DH adducts was dependent on reaction with HA.

Site Specificity and Kinetics of Formation of G-DH and G-H Adducts-
We have demonstrated that although CML forms at every lysine when RNase is incubated with glucose, the distribution of CML is not uniform and Lys 41 is the major site of modification (11). Therefore, we anticipated that G-DH and G-H might also form at other arginine residues in RNase, i.e. Arg 10 , Arg 33 , and Arg 85 , in addition to Arg 39 , and that their distribution among the arginine residues might be uneven. We therefore searched for ions with calculated m/z values of the different charge states of the predicted peptides, 8 FERQH-MDSSTSAASSSNYCNQMMK 31 ( 8 F-K 31 ), 32 SRNLTK 37 ( 32 S-K 37 ) and 67 NGETNCYQSYSTMSITDCRETGSSKYPN-CAYK 98 ( 67 N-K 98 ), where Arg 10 , Arg 33 , and Arg 85 were modified, respectively, to G-DH or G-H. In previous studies, we observed that trypsin produced almost no cleavage between Lys 91 and Tyr 92 of RNase. Therefore, modifications of peptide 67 N-K 98 , with two missed cleavage sites, were monitored. Peptides 8 F-K 31 and 67 N-K 98 , containing, respectively, Arg 10 and Arg 85 modified to G-DH and G-H, were located by extracting ion chromatograms from the full-scan data, thus establishing the formation of both of these adducts at Arg 10 and Arg 85 . When the 7-day 1 mM glyoxal reaction was incu- bated with HA, the kinetics of loss of peptides containing Arg 10 and Arg 85 modified to G-DH/CMA were similar to those of peptide 38 D-K 61 with Arg 39 modified to G-DH, indicating that G-DH and not CMA was also the major or sole adduct on these arginine residues. Peptide 32 S-K 37 with Arg 33 modified to either G-DH or G-H was not detected in the RNase-glyoxal reactions, despite a strong signal for the native peptide 34 NLTK 37 , indicating that Arg 33 was not converted to these adducts in detectable amounts in the glyoxal incubations, even after 7 days in 5 mM glyoxal.
Although G-H exists as three structural isomers, we could detect only one peak within the LC-MS data corresponding to G-H on Arg 10 , Arg 39 , and Arg 85 , possibly because the isomers were not resolved on the LC column or because some of them were present below the limit of detection. The 3 ϩ ions of the G-DH adducts of peptides 8 F-K 31 and 67 N-K 98 , and the G-H adduct of peptide 38 D-K 61 were also detected in the single combined spectrum that was extracted from the fullscan data of the glyoxal incubations, but not from the glucose incubations (Fig. 2, D, F, I, and K), and their retention times match those expected for these adducts. None of these mod-ified peptides was present in sufficiently high concentrations for peptide sequencing experiments.
The kinetics of formation of the peptides containing arginine residues, together with the kinetics of loss of the relevant unmodified peptides in the 1 mM glyoxal incubations, are shown in Fig. 6. Similar graphs were obtained for the 5 mM glyoxal incubations, but with a 2-to 3-fold increased response. Arg 39 and Arg 85 were identified as the major sites of adduct formation.

DISCUSSION
Location of Unpredicted Peptides-The application of MS to the analysis of proteins is a powerful technique for identification and localization of chemical modifications of proteins. However, the technique works best when the structure of the modified amino acid is known in advance. For our studies on modification of RNase by glyoxal, we were faced with the loss of a peptide containing Lys 41 , but the mass of the modification was uncertain. We therefore treated the TIC as a single peak to yield a single mass spectrum for the entire run. A visual comparison of the spectra for glucose-and glyoxal- modified protein facilitated the detection of ions that were present in only one of the samples. This allowed us to locate an unknown peptide within our RNase-glyoxal incubations that was absent from our glucose incubations, and that was subsequently identified as 38 D-K 61 in which Arg 39 was modified to G-DH. Sequencing the peptide, together with the data from the HA incubations, confirmed its identity and the site of modification. Analysis of the TIC is a simple and rapid procedure but, in general, is limited to detection of quantitatively major modifications because less abundant modified peptides are obscured by background ions. Using this approach, we identified Arg 39 as the major site of modification of RNase by glyoxal. Weak signals for the 3 ϩ ions of peptides 8 F-K 31 and 67 N-K 98 , containing Arg 10 or Arg 85 , respectively, modified to G-DH, as well as the 3 ϩ ion of peptide 38 D-K 61 containing Arg 39 modified to G-H, were also observed in these analyses.
Estimating Amounts of Modifications-Quantitation of absolute and relative modification of proteins by LC-MS/MS is a major challenge. We have previously estimated the extent of modification of different lysine residues in RNase by calculating relative amount (RA) values (11), i.e. the area units for charged species of the modified peptide, compared with the area units of the C-terminal peptide of RNase. One advantage of this technique is that values are normalized between samples, but a disadvantage is that the mass spectrometer response to different tryptic peptides of native RNase is variable, limiting the validity of the technique for comparing amounts of different tryptic peptides. The use of RA values to estimate modifications is more meaningful when the number of amino acid residues in the unmodified and modified peptides is the same or similar; in this case the mass spectrometer response is likely to be similar. Biemel and Lederer (17) established that lysine, fructoselysine, and N 6 -(2,3-dihydroxy-4-quinoxalin-2-ylbutyl)-L-lysinate gave comparable responses when analyzed by ESI-LC-MS. They also provided evidence that these modifications on an identical peptide resulted in an almost equivalent mass spectrometer response. Therefore, this approach was applied in this study to monitor the loss of the G-DH and G-H adducts at Arg 39 within peptide 38 D-K 61 and regeneration of peptide 40 C-K 61 in the HA reactions. These peptides differed in length by only two amino acid residues and also possessed very similar retention times on the HPLC column (Table I).
To follow the kinetics of formation of G-DH and G-H adducts at the different arginine residues within RNase, we expressed the data as a percentage of the unmodified and modified peptides within each arginine peptide group. This method eliminated variations in RA values due to inter-sample and inter-batch variations in the derivatization/digestion procedure and LC-MS response.
Identification of Peptide 38 D-K 61 Containing Arg 39 Modified to G-DH-Based on reversibility of modification by treatment with HA, UK1 was identified as primarily (Ͼ75%) peptide 38 D-K 61 containing the G-DH adduct at Arg 39 . Glomb and Lang (14) monitored the degradation of authentic G-DH in the presence of OPD or aminoguanidine (AG) at 37°C and pH 7. These experiments yielded half-lives of 1-2 days for G-DH, consistent with the half-life we obtained following incubation of our 7-day 1 mM glyoxal incubations with HA. CMA has been identified in collagen incubated in 1 M glucose (10), and it has been demonstrated, using ESI-LC-MS, that levels of CMA on serum proteins from diabetic subjects are significantly higher than those from age-matched controls (15). G-DH was not measured in either of these studies but, based on our data and those of Glomb and Lang (14), it may be present at higher concentrations than CMA.
Site Specificity of G-DH and G-H Formation-As shown previously (11), the formation of CML in RNase-glucose systems is site-specific (11), occurring primarily at Lys 41 in the active site and derived primarily from the Amadori adduct. Based on the present work, modification of protein by glyoxal is also site-specific, occurring primarily on Arg 39 and Arg 85 in RNase, with small amounts of modifications forming on Arg 10 , but no detectable adducts on Arg 33 . In previous work, Arg 39 and Arg 85 were also identified as the main sites of modification of RNase by ␣-dicarbonyls (18). The earliest work involved modification of bovine RNase A by phenylglyoxal, which also yielded a reversible adduct (19). Later studies, involving the incubation of RNase with 1,2-cyclohexanedione (20, 21) and 3-ethoxy-2-ketobutanal (22), concluded that Arg 39 and Arg 85 were the most modified sites, followed by Arg 10 , with Arg 33 being unreactive. The adduct formed from 1,2-cyclohexanedione, i.e. N 7 ,N 8 -(1,2-dihydroxycyclohex-1,2-ylene)-L-arginine, was characterized (20) and, like the glyoxal adduct, G-DH, its formation was reversible on incubation with HA. The lack of reactivity at Arg 33 was attributed to hydrogen-bonding to Asp 14 (19,22).
Relative Amounts of AGEs on Proteins-The data presented here show that, in the glyoxal incubations, the order of abundance of the adducts was G-DH Ͼ G-H Ͼ CML. Only low levels of CML peptides were formed from glyoxal and, in the glucose incubations, no G-DH or G-H adducts could be de-tected. We have previously reported (11) that, in the glucose incubations, the main route to CML is by oxidation of the ARP. This is confirmed by the absence of G-DH or G-H adducts on the glycated protein, indicating that little glyoxal was formed, and the higher abundance of peptide 40 C-K 61 carboxylated at Lys 41 in the glucose incubation compared with the glyoxal system.
AGEs are potential biomarkers of various disease states, including diabetes, atherosclerosis, and neurodegenerative disease. Relative amounts of different AGEs in physiological proteins and fluids may differ according to the severity of disease or particular tissue, reflecting local differences in carbonyl formation and oxidative stress. In general, HI adducts are reported to be present at higher concentrations than CML on plasma and tissue proteins (23,24). For example, G-H1 has been reported to be present at about twice the level of CML on plasma proteins of healthy human subjects. Levels of G-H1 on retina, nerve, and plasma proteins of streptozotocininduced diabetic rats also increased significantly compared with those of normal controls (24), reflecting elevated levels of plasma glyoxal in diabetes. Little attention has been given to G-DH adducts on protein (14), but our studies suggest that they may be present at higher concentrations than either CMA or G-H.

CONCLUSIONS
In conclusion, this study has demonstrated that G-DH is the primary product formed on reaction of glyoxal with RNase under physiological conditions. Formation of both G-DH and G-H is site-specific and Arg 39 and Arg 85 are the favored sites of modification of the protein. Neither G-DH nor G-H is formed in detectable amounts when RNase is incubated with glucose. The absence of G-DH and G-H in the glucose incubations supports a conclusion from our earlier work on CML that glyoxal formation is not an important route to CML in RNaseglucose incubations. Our findings suggest that arginine residues are the favored sites of modification of proteins by dicarbonyls and that inhibition of dicarbonyl formation and entrapment of these highly reactive species may be an important means of inhibiting AGE formation in vivo.