Reaction of acetaldehyde with hemoglobin.

Acetaldehyde reacted with hemoglobin at neutral pH and 37 degrees C to form adducts that were stable to dialysis and that were not reduced by sodium borohydride. Hemoglobin tetramers having 2, 3, and probably 4 molar eq of bound aldehyde were isolated by cation exchange chromatography. The sites of attachment of the aldehyde were the free amino groups of the N-terminal valine residues of the alpha and beta chains of hemoglobin. Derivatization of the beta chains caused a greater increase in the acidity of the hemoglobin than did derivatization of the alpha chains. Derivatization of the beta chains was also preferred over that of the alpha chains. Acetaldehyde derivatives of the N-terminal octapeptide of hemoglobin S (beta sT-1 peptide), Val-Gly-Gly, and tetraglycine were formed readily, contained 1 M eq of acetaldehyde/mol of peptide, and were not reduced by sodium borohydride. In contrast, Ala-Pro-Gly failed to form a 1:1 adduct with acetaldehyde. 13C NMR analysis of the peptide adducts formed with [1,2-13C]acetaldehyde indicated that tetrahedral diastereomeric derivatives were produced. The 13C chemical shifts of the adducts formed between hemoglobin and [1,2-13C]acetaldehyde were identical to those of the peptide adducts although resonances from the individual diastereomeric adducts at each hemoglobin site could not be resolved. The results cited above as well as other evidence indicate that acetaldehyde reacts with the amino termini of hemoglobin to form stable cyclic imidazolidinone derivatives. An exchange of acetaldehyde residues between peptides was also documented.

Modified hemoglobins resembling the known minor hemoglobins in their chromatographic and electrophoretic properties are formed when acetaldehyde is added to hemolysates, solutions of purified hemoglobin, or erythrocytes (1)(2)(3)(4). The possibility that metabolically produced acetaldehyde, i e . acetaldehyde generated by oxidation of ingested ethanol, might react directly with hemoglobin (1) or add indirectly to hemoglobin in the form of 5-deoxy-~-xylulose-l-phosphate, a product of the metabolism of acetaldehyde by erythrocytes (2,5), was supported by the observation that, in some alcoholic individuals, the proportion of minor hemoglobins was abnormally elevated (1,6). A modified nonglycosylated hemoglobin fraction having chromatographic properties similar to those of hemoglobin AI, appeared to account for this increase (7). Whether produced directly by addition of acetaldehyde or *This work was supported by Grant R01-HL-29754 from the National Institutes of Health and National Science Foundation Grant DMB-8413723. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked ''adUertisemnt" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom reprint requests should be sent: Albert Einstein College of Medicine, 1300 Morris Park Ave., Bronx, NY 10461.
indirectly by the addition of the deoxypentose phosphate, it was appreciated that the modified hemoglobin might serve to evaluate alcoholic consumption, much as levels of hemoglobin AI, are used to estimate the degree of chronic hyperglycemia in diabetes mellitus (1,2). The reaction of aldehydes with hemoglobin has received attention for other reasons, chiefly because of the effects of aldehydes on oxygen binding, gelation of hemoglobin S, and sickling (8)(9)(10). Sodium borohydride reduction of the Schiff base formed between deoxyhemoglobin and pyridoxal phosphate yielded a modified hemoglobin whose affinity for oxygen was permanently increased and unaffected by 2,3-diphosphoglycerate (8). Having stabilized the Schiff base derivative by reduction, it was possible to demonstrate by appIying degradative procedures that pyridoxal phosphate had become bound to the a-amino group of the N-terminal valine residue of , ! ? chains (8,ll).
Reductive alkylation, as exemplified above, has been the method of choice for localizing sites of binding of aldehydes to proteins (12). The procedure has been applied to elucidate sites of attachment to hemoglobin of glucose (13), glyceraldehyde (14), and glycolaldehyde (15). The reduced products are stable to denaturation and enzymatic and acid hydrolysis so that binding sites can be identified by chain analysis, peptide mapping, and amino acid analysis. A further advantage of reductive alkylation is that radioactive labeling of the sites can readily be accomplished with the use of either tritiated borohydrides as the reductant or labeled aldehyde.
The first step of reductive alkylation of proteins is formation of a Schiff base between the aldehyde and a free amino group of the protein. In the case of glucose and other carbonyl compounds having a neighboring hydroxyl (glyceraldehyde and glycolaldehyde), rearrangements of the Amadori type follow formation of the imine, stabilizing the linkage between added group and protein (16). Like their Schiff base precursors, the rearranged products are susceptible to borohydride reduction. It is noteworthy, however, that the reductive step influences the course of addition of the carbonyl compound. Incorporation of glyceraldehyde, for example, was more rapid in the presence of the reductant than in its absence. However, in the absence of cyanoborohydride, it formed rearrangementstabilized adducts to 6 chain termini 10 times more rapidly than to a chain termini whereas, in the presence of the reductant, addition occurred to both chains at approximately the same rate. Furthermore, in the absence of cyanoborohydride, addition of glyceraldehyde to the e-amino group of lysine residues was preferred over addition to a-amino groups (15).
Although reductive alkylation, keeping the foregoing limitations in mind, is a reliable technique for identifying reducible sites in proteins, it is not useful for localizing sites of addition of carbonyl compounds that form nonreducible adducts. Furthermore, it cannot be assumed that sites of aldehyde addition identified by reductive alkylation will necessar-681 1 ily be the same sites to which the aldehyde has become attached by a nonreducible linkage.
Although it is commonly held that the chief product of reaction between an aldehyde and a protein is a Schiff base, recent studies have shown that adducts of this kind are only one of several possible forms of aldehyde-modified proteins. Whereas Schiff base adducts are formed relatively quickly, readily dissociate by dilution or dialysis, and are susceptible to and stabilized by borohydride reduction, our concern here is with aldehyde-modified proteins that are formed more slowly than Schiff bases, are relatively stable to dialysis, and are not susceptible to reduction by borohydrides.
While the nature of the primary reaction between aldehydes and proteins is relatively simple, namely, nucleophilic attack of the free amino nitrogen on the aldehyde carbonyl, the ultimate effect of aldehydes on protein structure and function is complex. This is due not only to the intricacies of protein structure itself, but also to the influence of kinetic parameters on the structure, yield, and stability of the final product. Both the yield and nature of modified hemoglobins formed by reaction of acetaldehyde with hemoglobin are functions of the acetaldehyde concentration and the duration and temperature of reaction. At 37 "C and a reaction time of 30 min, Stevens et al. (1) observed that the yield of fast moving hemoglobin (FMH'), a fraction resembling the entire hemoglobin Al fraction in its chromatographic behavior, increased from 4.3% when the acetaldehyde concentration was 3 mM to 23.5% when the concentration was 30 mM. Virtually 100% of the hemoglobin in metabolizing human erythrocytes was transformed to FMH when the cells were exposed overnight at 37 "C to 15 mM acetaldehyde (3). Nguyen and Peterson (17), who incubated hemolysates with increasing concentrations of acetaldehyde at 37 "C for 24 h, resolved the FMH fraction into Ala+b and AI, fractions and found that, until the acetaldehyde concentration reached 1 mM or above, an increase in the Ala+b fraction was not observed whereas the AI, fraction increased above normal values with as low a concentration as The work described in this report was undertaken with the aims of elucidating the stoichiometry of addition and identifying sites of attachment of acetaldehyde to hemoglobin in stable adducts and to throw light on the mechanism of the addition reaction. Our findings disclose that in these adducts 0.5 mM. 'The abbreviations used are: FMH, fast moving hemoglobin; HPLC, high pressure liquid chromatography; @T-1, the octapeptide Val-His-Leu-Thr-Pro-Val-Glu-Lys; AcH, acetaldehyde; Cbz-, benzyloxycarbonyl-; TPCK, ~-l-tosyl-amido-2-phenylethyl chloromethyl ketone. acetaldehyde is covalently bound to the free amino groups of the N-terminal valine residues of a and p globin chains in tetramers having the compositions a$& a&, a'afi, and probably a$/%, where x represents an aldehyde-modified chain. 13C NMR spectrometry gave no evidence of the presence in these adducts of Schiff bases. Resonance peaks were observed, however, establishing the presence of diastereoisomeric tetrahedral addition products. Evidence will be presented suggesting that stable adducts of hemoglobin and acetaldehyde are imidazolidinone derivatives.

RESULTS
In order to define the conditions necessary to generate acetaldehyde-modified hemoglobins in good yield and having a range of ionic charge similar to that of the minor hemoglobins, 1 mM hemoglobin A. in 0.002 M bis-Tris, 0.15 M NaC1, pH 7.3, was reacted with 0, 0.2, 1.0, and 5.0 mM [1-3H] acetaldehyde for 20 h at 37 "C. The amount of FMH, estimated by chromatography on microcolumns of Bio-Rex 70 (22), was found to increase roughly in linear proportion to the aldehyde concentration. The yields were 3.8, 5.0, 17.4, and 55.1%, respectively. Selecting for further analysis the reaction mixture containing FMH in highest yield, preparative cation exchange chromatography was performed by the method of Abraham et al. (20), giving the results shown in Fig. 1. Seven acetaldehyde-modified fractions were isolated, accounting for 70% of the total hemoglobin and having ionic charges ranging from the most to the least basic of the known minor hemoglobins. The specific activity and amount of each fraction were determined and the stoichiometry of aldehyde addition was calculated from the specific activities of the hemoglobin fractions and acetaldehyde. The values obtained for Fractions II-VI1 are shown in Table I, column 4. Fraction I was omitted from these calculations since it contained some degraded material and labile radioactivity.
Hemoglobin having 2 or 3 molar eq of bound acetaldehyde/ mol of tetramer were the chief products. Yields of Fractions 111-VI1 were strikingly similar (Table I, column 2) while the order of elution from Bio-Rex 70 appeared to be determined by the number of bound aldehyde residues/M eq of hemoglobin. The total yield of modified hemoglobins based on the amounts of hemoglobin in the collected fractions was greater than given by the microcolumn procedure due to separation of Fraction VI1 from hemoglobin Ao, a degree of resolution not realized by the latter method. It is noteworthy that adducts having the same number of equivalents of bound aldehyde per tetramer nevertheless appeared to possess structural differences affecting the order of their elution from cationic exchangers. A somewhat higher degree of resolution of acetaldehyde-modified hemoglobins and a far more rapid analysis was obtained by cation exchange chromatography on a Synchropak CM-300 column, following the procedure of Huisman et al. (4). Fig. 2 shows the results of analyzing a dialyzed reaction mixture of 1 mM hemoglobin A. and 4.2 mM [1,2-3H]acetaldehyde (specific activity 5.3 x lo6 dpm/pmol) Portions of this paper (including "Materials and Methods," Figs. 2,3, 5, 7, and 9, and Tables 11-IV) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are available from the Journal of Biological Chemistry, 9650 Rockville Pike, Bethesda, MD 20814. Request Document No. 85M-3453, cite the authors, and include a check or money order for $5.60 per set of photocopies. Full size photocopies are also included in the microfilm edition of the Journal that is available from Waverly Press. after 20 h of incubation at 37 "C. Fractions IV and VI1 of Fig.   1 are each seen in Fig. 2 to have 2 components. As before, a total of 70% of the total hemoglobin had become modified. An aliquot taken at 4 h contained 33% modified hemoglobin with the most notable reductions in the yields of Fractions 11-IVa. The results in Table I, column 4, indicate that the relative acidity of acetaldehyde-modified hemoglobins, as evidenced by the order of their elution from cationic exchangers, increases with the number of chains to which the aldehyde is bound. This was confirmed by analysis of the fractions by isoelectric focusing, as shown in Fig. 3. The reaction mixture giving the results shown in Fig. 1 was found to focus as 4 discrete bands. The least acidic band was found in the same position as unmodified hemoglobin A,. Following in order of increasing acidity (net negative charge) were bands attributable to Fraction VII, Fractions V and VI, which were unresolved, and Fractions 11-IV, which were also unresolved. In general the order of elution from cationic exchangers correlated with their isoelectric points, although the chromatographic methods provided finer resolution.
To ascertain to which chains acetaldehyde was bound, chain analyses of the hemoglobin fractions were performed by reverse phase HPLC. The results of analyzing a known sample of hemoglobin A. and aliquots of Fractions 111, V, and VI1 are shown in Fig. 4. Chromatograms in column A include assays of radioactivity as well as UV absorbance. In column B are shown chromatograms of the fractions after adding hemoglobin & to provide unmodified a and p chains as markers.
Chain analyses of Fractions I1 and IV were indistinguishable from those of Fraction 111 while the results of analyzing Fraction VI were identical to those given by Fraction V. The number of modified a and /3 chains in each hemoglobin fraction are listed in columns 5 and 6 of Table I. Thus, from these results Fractions 11, 111, and IV are found to have the composition o(a"fi, Fractions V and VI, the composition a&, and Fraction VII, the composition a$pz, where x denotes modification due to acetaldehyde binding. These results show that modification of /3 chains brings about a greater increase in the relative acidity of the modified hemoglobin than does modification of a chains, and that multiple modifications, such as in Fraction 11-IV, result in the largest increase. The results also suggest that /3 chains were more reactive than a chains, in that the amount of hemoglobin having @ modifications was larger than that having a modifications. This was also indicated by the results obtained when 1.0 mM hemoglobin A0 was reacted with 0.  Table  I. Solid lines represent absorbance.

Time (min)
acidic fractions were much reduced in size. At the lower aldehyde concentration no significant formation of adducts having more than 2 modified chains was observed and hemoglobin tetramers having 2 modified ,8 chains (Fractions V and VI) were formed in larger amount than those having 2 modified a! chains (Fraction VII).
Reaction of Hemoglobin A1, with Acetaldehyde-Fractions V and VI, like hemoglobin Ale, have modified p chains and normal a chains and resemble hemoglobin Al, in their chromatographic and electrophoretic behavior. Having 1 molar eq of bound acetaldehyde per @ chain, the possible site of addition of the aldehyde is limited to the a-amino group of the Nterminal valine residue or to a free €-amino group of a lysine residue. In an earlier study hemoglobin A, was found to incorporate no radioactivity when erythrocytes were exposed to 0.180 mM [ l-3H]acetaldehyde suggesting that glycosylation of the 0-1 valine residue of hemoglobin blocked addition of acetaldehyde to hemoglobin. This inference was tested by reacting 55 PM hemoglobin A1,, which had been purified by chromatography on aminophenylboronated agarose (Glycogel B), with 5.0 mM [1,2-3H]acetaldehyde at pH 7.4 for 20 h at 37 "C. After dialysis to remove unbound acetaldehyde, chain analysis was performed, collecting 1-ml fractions each minute for counting. The results of this analysis, shown in Fig. 5, reveal modification and labeling of the a chains of hemoglobin AI, but not of the p chains. This finding clearly shows that the a-amino group of the N-terminal valine residue is the major if not the only site of stable attachment of acetaldehyde to the chains of hemoglobin. It should be emphasized that, despite the fact the the €-amino group of lysine residues in hemoglobin has been shown to form a Schiff base with acetaldehyde, glyceraldehyde, or glycolaldehyde, acetaldehyde does not stably bind to this site in the absence of reducing agents.
It was interesting to note that the binding of acetaldehyde by the a chains of hemoglobin A1, caused the glycosylated hemoglobin to elute earlier from Bio-Rex 70 than normal hemoglobin AI,, again illustrating the effect of multiple chain addition, even of an uncharged group, on the net negative charge of the glycosylated hemoglobin. This suggests that, when radioactive acetaldehyde is incubated with hemolysates or erythrocytes, radioactive fractions eluting from cationic exchangers before hemoglobin AI, may well be a chain modifications of the glycosylated hemoglobin. Like hemoglobin Al, these adducts would be borohydride-reducible.
Borohydride Reduction of Acetaldehyde-Hemoglobin Adduzts-Having established the stoichiometry of acetaldehyde binding by hemoglobin &, the question arose whether new reduction sites, equivalent in number to the added aldehyde residues, had been created. To answer this question acetaldehyde-hemoglobin & adducts were prepared as before but with the use of unlabeled rather than labeled acetaldehyde. After exhaustive dialysis to remove unbound acetaldehyde, the reaction mixture was reduced with standardized tritiated sodium borohydride, dialyzed to remove labile tritium, and after concentration by ultra-filtration, chromatographed on Bio-Rex 70 by the Abraham procedure to separate acetaldehyde-hemoglobin adducts. Using the data of Table I to assign the stoichiometry of aldehyde addition and the specific activities of the fractions and of the reduced standard to calculate the number of reduced sites, the results shown in Table I1 were obtained. Although each fraction took up a significant amount of tritium, the number of reduced sites falls far short of the number of added acetaldehyde residues. Except for the control reductions of hemoglobin Ale, which added even more than the expected 2 atom of tritium, and hemoglobin Ao, whose uptake of tritium is attributable to its content of glycosylated a chains (31), the uptake of radioactivity by the acetaldehyde-modified hemoglobins cannot be accounted for. This type of nonspecific reduction emphasizes the importance of using standardized radioactive borohydrides. The results which were obtained indicate in any case that carbonyl unsaturation of the aldehyde is lost on its addition to hemoglobin.
Tryptic Hydrolysis of Acetaldehyde-Hemoglobin Adducts-Since, other than those formed as Schiff bases, the bonds between acetaldehyde and hemoglobin are not reduced by borohydride and because the adducts are not stable to acid hydrolysis (l), a mixture of radioactive acetaldehyde-modified hemoglobins was subjected to tryptic hydrolysis with the intention of isolating and identifying labeled peptides and thereby localizing the site(s) of attachment of the aldehyde. An aliquot of the hemoglobin mixture having the composition shown in Fig. 2 was hydrolyzed with trypsin for 24 h and the resulting tryptic peptides were separated by reverse phase HPLC. It was noted that, in the course of denaturation and hydrolysis of the hemoglobin, more than half of the bound acetaldehyde was recovered in the distillate collected during lyophilization of the digest. The results of peptide analysis of the lyophilized digest are shown in Fig. 6A. Contrary to expectation, radioactive peaks were scattered throughout the chromatogram. Having observed earlier that addition of the aldehyde to the p chains of hemoglobin & appeared to be limited to the free amino group of the P-1 valine residue and that this was likely to be true of the addition of the aldehyde to a chains as well, the appearance of a multiplicity of radioactive peaks suggested that, during the incubation with trypsin, acetaldehyde was transferred from the a-amino group of @T-1 peptide, the N-terminal octapeptide of @ chains, to the a-amino group of N-terminal amino acid residues of the other peptides released by trypsin. On this assumption the time of tryptic digestion was shortened in the expectation that less exchange would occur. The results of a 30-min digestion of the same sample are shown in Fig. 6B. Except for labile radioactivity at the start of the chromatogram and radioactivity associated with undigested polypeptides at the end, only 2 prominent peaks of radioactivity are seen. The radioactive peaks 1 and 2 from several runs were pooled and purified by rechromatography using ammonium acetate-acetonitrile elution. The results of amino acid analyses of the 2 peptide fractions are shown in Table 111. Peak 1 was observed to have the same composition as PT-1, confirming that the site of stable addition of acetaldehyde to / 3 chains is the a-amino group of the valine 1 residue. Peak 2 had the composition of aT-1 + 2 indicating that the free amino group of the Nterminal valine residue of a chains is also the site of stable acetaldehyde modification. Transfer of Acetaldehyde between Peptides-The transfer of acetaldehyde from one peptide to another was demonstrated as follows. The acetaldehyde derivative of PT-1 octapeptide was prepared by reacting a 12 mM solution of the peptide with 45 mM [1,2-3H]acetaldehyde for 72 h at 37 "C. The reaction mixture was Iyophilized and then analyzed by reverse phase HPLC, giving the results shown in Fig. 7A. The acetaldehyde derivative of the peptide gave the radioactive peak and double peak of UV absorbance shown. The specific activity of the adduct, calculated from the total radioactivity of the collected fractions and the UV absorbance of the doublet showed that it contained 1 molar eq of acetaldehyde/ molar eq of peptide. The product yield was 60% based on the starting amount of peptide. The adduct was stable to repeated lyophilization but unstable to incubation for 24 h at 37 "C at pH 3 or 8, decomposing to the extent of 90 and 40%, respectively. The transfer of acetaldehyde from the PT-1 adduct was demonstrated by incubating a 1.2 mM solution containing 60% adduct and 40% underivatized octapeptide with 12 mM Val-Gly-Gly at pH 8 and 37 "C for 24 h, i.e. under the conditions of time, temperature, and pH employed for tryptic hydrolysis. Peptide analysis of the reaction mixture at the end of the incubation gave the chromatogram shown in Fig.  7B. From the amount of radioactivity in the labeled fractions it was estimated that 90% of the radioactive aldehyde had been transferred from P' T-1 peptide to the tripeptide. Corresponding change in the peak areas of PST-l and Val-Gly-Gly remaining at the end of the reaction can be seen to have taken place.
The acetaldehyde derivative of psT-1 peptide, like the adducts of hemogIobin Ao, was not susceptible to reduction by sodium borohydride, indicating that the acetaldehyde derivative of @T-1 was a suitable model for a study of the mode of attachment of the aldehyde to the N-terminal valine residues of hemoglobin.
13C NMR Studies-The [1,2-13C]acetaldehyde derivative of @T-1 was prepared by incubation of 100 mM isotopic aldehyde and 6 mM PST-1 peptide at 37 "C for 72 h. Analysis of the reaction mixture by HPLC showed that 80% of the octapeptide was derivatized. The proton decoupled 13C NMR spectrum of the sample containing 2.4 mM adducts is shown in Fig. 8A. The spectrum consists of resonances from free acetaldehyde and from the peptide adducts. Thus, the doublets at 6 = 88.3 and 23.2 ppm arise from the C-1 and C-2 carbons, respectively, of the free acetaldehyde in the hydrated form. Only about 15% of the free acetaldehyde exists in the carbonyl form with chemical shifts of 207 and 30.2 ppm for C-1 and C-2, respectively (Fig. 8D). The resonances from the C-1 carbons of the adducts appeared as a pair of doublets at 6 = 70.3 and 69.5 ppm indicating the presence of two adducts. These probably correspond to the two closely spaced HPLC peaks observed for the adducts. No resonances were observed in the region of 140-170 ppm where the resonances for the imine carbons of Schiff bases are known to occur, and with chemical shifts of 70 ppm the C-1 carbons of the adducts were clearly not sp2 hybridized, confirming the results of sodium borohydride reductions. In the proton-coupled spectrum of the PST-1 adducts, the resonances from the C-1 carbons appeared as a pair of quadruplets, indicating a singly directly bonded hydrogen (Fig. 8B).
When Val-Gly-Gly was reacted with [1,2-13C]acetaldehyde at pH 7.2 under the same conditions, 100% of the peptide was converted to the acetaldehyde derivative. The proton decoupled spectrum of this sample containing 2.7 mM adduct is shown in Fig. 8C Table 111.
results imply that a peptide with a free a-amino group is the structural requirement for stable adduct formation. Because of their similarity in intensity and chemical shift, the pair of resonances were postulated to be due to a pair of diastereomeric adducts. To test this, the [1,2-13C]acetaldehyde adduct of tetraglycine was prepared and examined. A doublet from a single adduct was observed supporting this contention. The 13C NMR assignments are summarized in Table IV.
To test whether the adducts formed between acetaldehyde and hemoglobin were the same as those formed with peptides, a solution of 1 mM hemoglobin Go and 5 mM [1,2-13C]acetaldehyde was incubated under standard conditions, dialyzed, and analyzed by CM 300 HPLC. Hemoglobin-acetaldehyde adducts I-VI1 were formed in the same proportions as before and as a whole represented about 70% of the total hemoglobin. The unfractionated reaction mixture was made 0.01 M in sodium phosphate, pH 7.4, containing 20% DzO and 0.01 M KCN for a final concentration of 0.6 mM hernoglobin for 13C NMR analysis. The KCN was added to eliminate the potential line broadening and chemical shift effects on the carbon resonances from the adducts due to the presence of paramagnetic methemoglobin (32). A sample of 0.6 mM hemoglobin A,, was prepared in the same way to provide a measure of the background of the 13C NMR resonances due to the naturally occurring 13C nuclei of the protein. The proton decoupled 13C NMR spectrum of the hemog1obin-[l3C]acetaldehyde reaction mixture is shown in Fig. 9A. The notable observation is the strong resonance in the region of 70 ppm indicating the presence of acetaldehyde adducts similar to those made with peptides. The naturally abundant spectrum of unreacted hemoglobin & at the same concentration is shown in Fig. 9B where the resonance at 70 ppm is notably absent. To confirm the observation, the difference spectrum (spectrum A minus spectrum B ) was computed and is shown in Fig. 9C. Two broad resonances are clearly observed in the regions of 70 and 20 ppm which coincide exactly with the chemical shifts of the resonances from the C-1 and C-2 carbons, respectively, of the peptide adducts. With two sites of acetaldehyde addition to hemoglobin and two diastereomeric adducts forming at each site, the intrinsically broad line widths of the resonances from the different protein bound adducts could not be resolved from one another even in the spectrum obtained after 40,000 accumulations.

DISCUSSION
In reacting with free amino groups of hemoglobin Ao, acetaldehyde was observed to form tetrahedral addition products that were stable to prolonged dialysis in the cold and to lengthy preparative and analytical procedures. Schiff bases, which other observers have detected by reductive alkylation of hemoglobin in the presence of an excess of acetaldehyde (I), may also have been produced but presumably were dissociated by dialysis.
In concentrations slightly in excess of that of globin chains, acetaldehyde became stably bound to the free amino group of the N-terminal valine residues of hemoglobin a and @ chains. Hemoglobin tetramers were identified having the compositions afPz, CY&, and CYCY'& where x represents acetaldehyde addition to the chain. Chain analysis revealed 2 modified forms of a and @ chains, an observation consistent with the presence of pairs of hemoglobin tetramers having the same empirical structure, ie. afpz (Fractions VIIa and VIIb) and a& (Fractions V and VI). A possible explanation of this finding is deferred for later discussion.
In accord with the observations of others (17), addition of acetaldehyde to @ chains appeared to be favored over addition to a chains.
The observations reported here should not be interpreted to mean that acetaldehyde derivatives of other reactive groups than the free amino groups of N-terminal amino acid residues are not formed. Unstable modifications of -SH, a-amino, guanido-, and imidazolo-groups have been demonstrated by titration and in the case of t-amino groups by reductive alkylation. Labeled adducts of lysine, tyrosine, glucosylvaline, and glucosyllysine have been identified among the products of acid hydrolysis of borohydride-reduced acetaldehyde-modified hemoglobin (1). The absence of stable unreduced acetaldehyde adducts of the €-amino group of lysine residues indicates that the only acetaldehyde-hemoglobin linkages capable of undergoing spontaneous stabilization are those that are formed between the aldehyde and free a-amino groups. Attempted reduction of stable hemoglobin adducts to which 2 and 3 M eq of acetaldehyde/tetramer remained bound failed to reveal any correspondence between the number of bound aldehyde residues and reduction sites, perhaps an unnecessary verification of the fact that Schiff bases are indeed dissociated by dialysis. It is noteworthy, however, that, after dialyzing acetaldehyde-hemoglobin reaction mixtures to remove any free acetaldehyde, Stevens et al. found that the yield of reductively alkylated products of acid hydrolysis of the reduced modified hemoglobin increased with the length of time of reduction. Thus, during the reduction, slow transformation of stable adducts to Schiff bases must have occurred.
The relative basicity of acetaldehyde-hemoglobin adducts, as judged by the order of their elution from cationic exchangers or by isoelectric focusing, appears to be determined by the number of chains and the type of chain to which the aldehyde becomes bound. The most basic adduct, which eluted immediately adjacent to or with the leading edge of the hemoglobin & peak, was found to have 2 derivatized a chains and normal ( 3 chains. The same chromatographic behavior is exhibited by hemoglobin A having glycosylated a chains and normal ,6 chains (31). The most acidic adduct of known composition had 3 chains to which acetaldehyde was bound (da@;). Having properties intermediate between these extremes was a hemoglobin adduct having acetaldehyde bound to the N-terminal valine residues of both p chains, an adduct closely resembling in its chromatographic behavior hemoglobin A1,, whose / 3 termini are also the sole sites of chain modification. Although no hemoglobin adduct was identified having acetaldehyde bound to all 4 chains, it is likely that Fraction I (Fig. 1) contained an adduct of this composition.
With increasing acetaldehyde concentration not only does the amount of modified hemoglobin increase but so does the relative acidity of the product. Thus, virtually 100% of hemoglobin in erythrocytes exposed to 15 mM acetaldehyde was transformed to products which, according to the order of their elution from Bio-Rex 70, had lost much of their native basicity (3). Similar results have been reported by Stevens et al. (1) and by Abraham et al. (10). The present results indicate that the most acidic of these products are adducts having 3 and probably 4 derivatized chains.
The tendency of @ chains to form acetaldehyde adducts more readily than a chains has been observed in this laboratory and elsewhere. Since values of the pK of the a-amino group of the N-terminal valine residue of both a and @ chains are closely similar (33), it cannot be argued that the formation of chain adducts is favored by a higher concentration of free base at the attachment site. Furthermore, since the present experiments were carried out with oxyhemoglobin, salt bridges and hydrogen bonds that are present in deoxyhemoglobin are broken, so that differences in reactivity at the two sites cannot be said to be due to differences in their accessibility to the aldehyde. It is noteworthy that while the amino terminal groups of both chains are equally reactive in forming Schiff bases with glucose or glyceraldehyde, rearrangement of the Schiff base to the stable ketoamine is far more favored in the case of p than a chains (14). This has been attributed to the influence of the different local environments of the p-1 and a-1 valine residues on the rate-limiting rearrangement of the aldimine to the ketoamine (14). Similarly the preferential reactivity of @-I valine towards acetaldehyde may be due to environmental or conformational factors affecting the ratelimiting ring closure that is the final step in the synthesis of imidazolidinones (see below).
Because the stable adducts formed from acetaldehyde and hemoglobin were not susceptible to reduction, localization of the sites of attachment of the aldehyde could not be accomplished by the usual degradative procedures. On reacting acetaldehyde with hemoglobin A1,, however, it became evident that the a-amino group of the N-terminal valine residues was the probable site of addition of the aldehyde to a and @ chains. Tryptic hydrolysis, so long as it was relatively brief, provided information from which it was possible to deduce that acetaldehyde became bound only to the a-amino group of the N termini of both a and @ chains. Our results indicate that, at most, 4 M eq of acetaldehyde can form stable bonds with hemoglobin A.
The intermolecular transfer of acetaldehyde residues from one peptide to another is reason for caution in interpreting the results of experiments using the labeled aldehyde to pinpoint sites of attachment. Whether or not such transfers take place between the N termini of globin chains remains to be determined.
The stability to dialysis and the nonreducible nature of the bond between acetaldehyde and the N-terminal valine residues of globin chains rule out in these adducts Schiff base linkages between the aldehyde and amino groups. The results of 13C NMR spectrometric analysis of peptide and hemoglobin adducts of [ 1,2-13C ]acetaldehyde support that conclusion. Schiff base resonances were absent while resonance peaks attributable to a tetrahedral structure were clearly seen. Pairs of resonance peaks showed that addition of the aldehyde to the free a-amino group of the N-terminal valine residue of PST-1 peptide and to Val-Gly-Gly had created a new chiral center at the a-carbon of the attached aldehyde. Accordingly, diastereoisomers of the peptide adducts were produced. The same must also be true in the case of hemoglobin adducts. While formation of a carbinolamine, as in the reaction of formaldehyde with free amino groups of proteins, would account for these observations, other experimental evidence more strongly suggests that acetaldehyde-peptide and acetaldehyde-hemoglobin adducts are imidazolidinones. In this regard recent studies of the reaction of acetaldehyde with pentapeptide enkephalins are most pertinent (34).
A hypothetical reaction scheme for the synthesis of a 2methylimidazolidin-4-one from acetaldehyde and an N-terminal valine peptide is shown in Fig. 10. The reaction involves the initial formation of a Schiff base between the aldehyde and the a-amino group of the peptide followed by nucleophilic attack of the amide nitrogen of the peptide bond between the first and second residues of the peptide on the electrophilic imine carbon to form the cyclic imidazolidinone. In the cases of the enkephalins and the related Tyr-Gly-Gly, the structure of the adduct was proposed based on the loss of the proton NMR resonance of the amide hydrogen of the second residue and the appearance of resonances from the added CH&H bridging group (35). We have confirmed this observation by proton NMR of the acetaldehyde derivative of Val-Gly-Gly. 3 The structure was confirmed by analysis of the two-dimensional proton NMR spectrum (36). The adducts also resembled in stability the imidazolidinone derivative formed between oxytocin and acetone, the structure of which was characterized by infrared spectrometry and confirmed by synthesis (37, 38).
The observation by 13C NMR spectrometric analysis of a pair of adducts formed by reaction of acetaldehyde with BsT-1 and Val-Gly-Gly and only a single adduct of tetraglycine is consistent with the proposed formation of diastereoisomeric imidazolidinones from the aldehyde and the N-terminal valine peptides and an imidazolidinone enantiomer from the aldehyde and tetraglycine. The 13C NMR chemical shifts for the C-1 carbon of the aldehyde residue (C-2 of the imidazolidinone) are also consistent with a carbon bonded to amino and amido functions. However, we are not aware of 13C NMR studies of compounds proved by other methods of analysis to possess the imidazolidinone ring.
We found that Ala-Pro-Gly, which is unable to react with acetaldehyde to form an imidazolidinone, did in fact fail to yield a stable addition product with acetaldehyde whereas no difficulty was encountered in synthesizing stable acetaldehyde adducts of @T-1, Val-Gly-Gly, and tetraglycine.
The €-amino group of lysine residues of hemoglobin were found to be unreactive in forming stable adducts with acetaldehyde although Schiff base adducts of the same group have been demonstrated by reductive alkylation. This implies that the microenvironment of the reacting amino group is a factor affecting the production of stable acetaldehyde adducts. Whereas the distance from an €-amino group of a lysine residue to the nearest amide group is out of range for stable interaction, the nitrogen of the peptide bond between the Nterminal and adjoining amino acid residue is precisely situated to form a stable 5-membered ring with the postulated Schiff base or carbinolamine intermediate.
We have observed that the acetaldehyde adduct of B'T-1 octapeptide is not split by thermolysin, an endopeptidase that readily cleaved the underivatized octapeptide between histidine 2 and leucine 3 and which hydrolyzes such N-terminal blocked substrates as Cbz-Gly-Phe-amide and Cbz-Thr-Leuamide between the first and second amino acid residues (39). This suggests that the acetaldehyde adducts of BST-l are not simple N-alkylated derivatives. On the other hand, the conformational restraints imposed upon the peptide by the imidazolidinone ring might well prevent proper binding of the peptide to the enzyme and thus block hydrolysis. The foregoing evidence strongly supports the view that the structure of the attachment site of stable acetaldehyde-hemoglobin adducts is that of a 2-methylimidazolidin-4-one.
It has been known for some time that hemoglobins differing by chemical modification will form all possible hybrids of the type ( a B ) j ( a @ ) k when mixed in solution and that the plane of cleavage between the dimers is always the same (40). In the course of preparative or analytical separations of a mixture of modified hemoglobins, dissociation into dimers takes place with like dimers seeking the same position along the axis of R. C. San George and H. D. Hoberman, unpublished observations. separation. Reassociation of like dimers in these zones then occurs with the formation of symmetrical tetramers. The separation process itself has become the means by which like dimers, formed from unlike tetramers, segregate and reassociate.
In none of the experiments reported here was a hemoglobin adduct having a single modified chain detected among the products of the reaction between acetaldehyde and hemoglobin. Since 2 diastereoisomeric adducts are formed at each reaction site, four types of hemoglobin tetramers having a single chain modification are possible: d a h , or'&, a~@p, and a@'@, where x and y are diastereoisomers. The absence of hemoglobin tetramers of this form is regarded as evidence of dissociation into and reassociation of hemoglobin dimers. Subunit exchange occurring during the isolation process would yield only symmetrically modified tetramers, as illustrated 2axc& + 2 0 3 + Pa@ + a$& + a&. Accordingly, we assign the compositions azfi and a2& to Fractions V and VI and afPz and &Pz to Fractions VIIa and VIIb.
Doubly modified dimers, such as ax$, would be expected to form a hemoglobin tetramer having 4 modified chains. Although an adduct having this composition was not isolated in any of our experiments, it is more than likely that tetramers having this composition were formed but were eluted in Fraction I which, because of its high content of labile radioactivity and what was believed to be degraded hemoglobin, was not thoroughly analyzed. However, tetramers having 3 modified chains were identified. Although structures of this type would appear to contradict the principle of association of like dimers, the existence of these structures would seem to be evidence of incomplete resolution of dimers having similar but not identical chromatographic properties. Thus, the dimer ax$ might very well reassociate with the dimer ap" since, as indicated earlier on, modification of @ chains by acetaldehyde increases the acidic character of the chains to a far greater degree than does the same modification of a chains.
Accordingly, the structures daB and $a@ are assigned to Fractions I1 and I11 and aXaB and $aW are assigned to Fractions IVa and IVb. The formation of asymmetrically modified tetramers such as da$p cannot be excluded from consideration. It is evident that to an indeterminate extent the method of separation is itself an important factor in directing the reassociation of hemoglobin dimers and thus in the final composition of modified hemoglobins.
Examination of the 13C NMR spectra of separated fractions of ['3C]acetaldehyde-hemoglobin adducts is expected to shed further light on this interesting problem. 14.

18.
19. 20. Attention is directed to the modification l a t e r elution times than unmodifled chains. It will a150 be noted that 2 different 3 chains modifications were observed. Solid lines repcesent absorbance.
Cationic exchange HPLC of modified hemoglobins formed by reaction of ImM hemqlobin A0 and 4.2 mM 13H] acetaldehyde. Finer resolutmn of peaks I V and VI1 is seen. Solid line represents absorbance.   W absorbance due to the R ' T -1 adduct and the appearance of a new radioactive peak and a small W absorbance peak due to Eormtion of the acetaldehyde derivative of val-gly-gly. Solid lines represent absorbance.