Acetylation of Human Hemoglobin by Methyl Acetylphosphate

The development of chemical modification agents that reduce the tendency of sickle hemoglobin (HbS) to aggregate represents an important chemotherapeutic goal. Methyl acetylphosphate (MAP) has been reported to bind to the 2,3-diphosphoglycerate (2,3-DPG) binding site of hemoglobin, where it selectively acetylates residues, resulting in increased solubility of HbS. We have prepared [1-13C]MAP and evaluated the adduct formation with hemoglobin using 1H-13C HMQC and HSQC NMR studies. These spectra of the acetylated hemoglobin adducts showed 10–11 well resolved adduct peaks, indicating that the acetylation was not highly residue specific. The chemical shift pattern observed is in general similar to that obtained recently using [1′-13C]aspirin as the acetylating agent (Xu, A. S. L., Macdonald, J. M., Labotka, R. J., and London, R. E. (1999) Biochim. Biophys. Acta 1432, 333–349). Blocking the 2,3-DPG binding site with inositol hexaphosphate (IHP) resulted in a selective reduction in intensity of adduct resonances, presumably corresponding to residues located in the 2,3-DPG binding cleft. The pattern of residue protection appeared to be identical to that observed in our previous study using IHP and labeled aspirin. Pre-acetylation of hemoglobin using unlabeled MAP, followed by acetylation with [1′-13C]aspirin indicated a general protective effect, with the greatest reduction of intensity for resonances corresponding to acetylated residues in the 2,3-DPG binding site. These studies indicated that both MAP and aspirin exhibit similar, although not identical, acetylation profiles and target primarily the βLys-82 residue in the 2,3-DPG binding site, as well as sites such as βLys-59 and αLys-90, which are not located in the β-cleft of hemoglobin.

Therapeutic approaches for the treatment of sickle cell disease include induction of fetal hemoglobin expression, bone marrow transplantation, and the use of pharmacological agents that interact either non-covalently or covalently with HbS 1 to inhibit aggregation (1). The latter approach remains a potentially important therapeutic strategy, which has been under exploration by a number of groups (2)(3)(4)(5)(6), although to date no agents of this class have been successfully developed for clinical use. Since the first report of hemoglobin carbamylation by cyanate as a potentially effective anti-sickling agent (7), a number of other agents have been designed, and their suitability as anti-sickling drugs has been investigated (e.g. Refs. 5 and 8). Among these reagents is methyl acetylphosphate (MAP), an acetylating agent that was first designed as an inhibitor of D-3-hydroxybutyrate dehydrogenase (9). Based on the investigation of the modification reaction using conventional protein chemistry techniques, MAP was reported to target residues in the 2,3-DPG binding site: ␤Lys-144, ␤Lys-82, and ␤Val-1 (Refs. 10 and 11 and references therein). No residues of the ␣-chain were reported modified in the same study (10). The modification of HbS by MAP was reported to yield adduct hemoglobin with oxygen affinity similar to that of hemoglobin A but with a reduced gelation tendency. The decrease in the affinity of 2,3-DPG and in gelation tendency is thought to result from the MAP selective acetylation of amino groups located in the ␤-cleft to which 2,3-DPG binds. Among the set of criteria first proposed by Walder et al. (3) for a suitable antisickling agent, specificity for selected target residues of hemoglobin is of critical importance in order to limit the introduction of undesired structural and functional changes of the hemoglobin as well as the toxicity of the agent resulting from modifications of other proteins. Multidimensional NMR in conjunction with isotopically labeled agents has been shown to be an effective approach for studying chemical modification of hemoglobin (12) and ubiquitin (13). We report here our NMR study of the sites/residues of human hemoglobin acetylated by MAP. Results from this study are important for evaluating the effects of MAP acetylation on hemoglobin and hence are useful for the evaluation of this compound as a possible anti-sickling agent.

EXPERIMENTAL PROCEDURES
Inositol hexaphosphate sodium salt (IHP) was from Calbiochem-Novabiochem; salicylic acid was from Fluka (Ronkonkoma, NY); and Slide-A-Lyzer dialysis membrane cassettes (molecular mass cut-off, 10 kDa) were obtained from Pierce. Centriprep and Centricon membrane concentrators were from Amicon (Beverly, MA). All other reagents were of analytical grade. Human hemoglobin A was prepared using the protocol reported in our previous study (12). [1Ј-13 C]Aspirin was synthesized using the procedures reported elsewhere (12,13). [1-13 C]Methyl acetylphosphate was prepared by a reported synthesis procedure (14) using [1-13 C]acetyl bromide as the precursor for isotopic labeling.
(COHbA) or oxy form (OxyHbA) where indicated was allowed to react with a 10-fold molar excess of MAP in phosphate-buffered saline (PBK; ϳ35 mM phosphate, 140 mM KCl, pH 7) at 37°C for 1 h. At the completion of the incubation, the hemoglobin sample was dialyzed immediately against a 1000-fold excess volume of cold PBK (20 mM phosphate) at 4°C overnight, with two to three changes of buffer to quench the reaction and to remove the unreacted MAP and acetate. The dialyzed hemoglobin adduct solutions were then reconcentrated for the NMR experiments using a Slide-A-Lyzer membrane microconcentrator. Total hemoglobin concentration was determined by measuring the concentration of cyanomethemoglobin (CNHbA) using the Total Hemoglobin kit from Sigma. Acetylation of hemoglobin by aspirin was achieved under similar conditions except that the reaction was allowed to proceed for 12 h prior to dialysis. The HMQC (15) and HSQC (16) experiments were adapted to acquire the spectra of the chemically modified hemoglobin adducts. Details of spectral acquisition, processing, and analysis are reported elsewhere (12).

RESULTS AND DISCUSSION
In this study, 1-13 C-labeled methyl acetylphosphate was synthesized and used to acetylate human hemoglobin. The presence of the isotopic label allowed detection of the adduct resonances of large proteins such as hemoglobin using 1 H-13 C HMQC or HSQC spectroscopy. Detection utilizing the relatively weak polarization transfer between the carbonyl 13 C and methyl protons of the acetyl group was facilitated by the high mobility of the acetyllysine side chains, and the carbonyl resonances were found to yield greater shift dispersion than the methyl resonances observed using [2-13 C]acetyl-labeled precursors. The 1 H-13 C HMQC spectrum of COHbA acetylated using a 10-fold molar excess of [1-13 C]MAP is shown in Fig. 1A. Approximately 10 -11 well resolved peaks were apparent in the spectrum, indicating only a moderate residue specificity of acetylation. A small and relatively sharp peak labeled "Ace" arises from free [1-13 C]acetate in the sample, possibly present as a result of incomplete removal by dialysis. A similar adduct spectrum was acquired for OxyHbA modified in HEPES buffer under the conditions used by Manning and co-workers (10) (Fig. 1B). In general, we have avoided the use of HEPES and other amino group-containing buffers, which are readily acety-lated, leading to a loss of buffering capacity and to the release of amino protons upon formation of amidated buffer.
To investigate the relative level of COHbA modification as a function of MAP concentration, a series of HSQC spectra was acquired on the COHbA modified with increasing molar ratios of [1-13 C]MAP:COHbA. The spectral intensities are shown as a function of the ratio of [1-13 C]MAP:COHbA in Fig. 2. In this figure, we have plotted the sum of the intensities of resonances 5, 6, and 9 for the reasons discussed below. Subject to some scatter, the intensities of the adduct peaks are roughly proportional to the MAP:COHbA ratio, as expected. In particular, it does not appear that the most highly reactive sites become fully . The spectra were acquired at 37°C with 2048 t 2 data points and 128 complex t 1 data points, with 16 scans per t 1 increment. A 90°shifted sine-squared function was applied in both dimensions, and the data were zero-filled to 4096 ϫ 2048 points prior to Fourier transformation. The 1 H and 13 C chemical shifts were set to 0.0 and Ϫ2.0 ppm for 3-(trimethylsilyl)-1-propanesulfonic acid. Use of a small spectral window to improve the 13 C spectral resolution resulted in folding of the [1-13 C]acetate resonance at ϳ182 ppm into the region of interest. The peaks are labeled from 1 to 11 arbitrarily for identification purposes. B, 1 H-13 C HMQC spectra of 0.8 mM OxyHbA acetylated using a 10-fold excess of [1-13 C]MAP in 20 mM HEPES, pH 7.5, at 25°C for 1 h. The adduct sample was dialyzed in 20 mM potassium phosphate-buffered KCl (140 mM), pH 7, prior to the NMR experiment. The spectrum was acquired as described in A but with 64 scans/t 1 increment. Ace, denotes the resonances of [1-13 C]acetate. acetylated before significant modification of less reactive sites. Hence, these data also indicate that although there is significant variation in the rates of acetylation of the various hemoglobin residues, these differences are not sufficient to achieve a high degree of residue selectivity.
Assignment of the adduct resonances in acetylated hemoglobin represents a difficult problem. We have recently performed extensive studies on hemoglobin acetylated with [1Ј-13 C]aspirin (12) and assigned the resonances using a broad range of approaches. These included protection of the 2,3-DPG binding site with both covalent and non-covalent agents, modification of ␤Cys-93 with a spin label, the use of site-directed hemoglobin mutants, and prediction of dipolar shifts in paramagnetic CNHbA. These studies have allowed assignments of a number of the adduct resonances and also demonstrate that the most highly modified residue is ␤Lys-82 rather than ␤Lys-144 as previously reported (17). The acetyl ␤Lys-82 adduct has been shown to give rise to several resonances (12). The latter phenomenon occurs in more highly modified hemoglobin, apparently as a result of the additional modification of nearby residues leading to a chemical shift inequivalence (12). In general, the HMQC spectra of hemoglobin acetylated using [1Ј-13 C]aspirin or [1-13 C]MAP were fairly similar, indicating that [1Ј-13 C]aspirin and [1-13 C]MAP share major acetylation sites. Indeed, our recent studies (12) and investigations by Manning and co-workers (Ref. 10 and reference therein) show that both agents acetylate ␤Lys-82 in the 2,3-DPG binding pocket (10,12). Based on this spectral comparison and in light of our recent study of asprin acetylation (12), we assign resonances 1, 3, and 6 to acetyl ␤Lys-59, acetyl ␣Lys-90, and acetyl ␤Lys-82, respectively. Resonance 8 is tentatively assigned to acetyl ␤Lys-144. Further, as in the aspirin study, it is likely that resonance 9 arises from hemoglobin molecules acetylated at both the ␤Lys-82 and ␤Lys-82Ј on the second ␤-chain, because the latter represents the potential acetylation site nearest to ␤Lys-82. Resonance 5 may also arise from acetyl ␤Lys-82 in hemoglobin, which is additionally acetylated at other nearby or interacting sites. This conclusion is supported by several lines of evidence, particularly the identical shifts observed for resonances 6 and 9 in acetylated paramagnetic cyanomethemoglobin (12).
To confirm the assignment of acetyl ␤Lys-82 and other adduct residues in the 2,3-DPG pocket, acetylation of hemoglobin with [1-13 C]MAP was carried out in the presence of a 1-4-fold molar excess of IHP. IHP was utilized to block access to the 2,3-DPG binding site of COHbA because of its greater affinity for this site (K D ϭ 10 Ϫ4 M) (18), compared with 2,3-DPG (K D ϭ 2.4 ϫ 10 Ϫ3 M) (19). The binding of IHP resulted in a substantial decrease in intensity for peaks 6, 8, and 9 (Fig. 3), consistent with the assignment of these resonances to acetyl ␤Lys-82 and acetyl ␤Lys-144 in the 2,3-DPG binding pocket. However, no significant decreases of the intensities of peaks 1, 2, 3, 4, 7, 10,

FIG. 4. 1 H-13 C HMQC spectra of 2 mM COHbA acetylated with a 10-fold molar excess of [1-13 C]aspirin at 37°C for 12 h without (A) and with (B)
prior reaction with a 10-fold molar excess of unlabeled MAP at 37°C for 1 h. After treatment of the COHbA with MAP, the unreacted MAP was extensively dialyzed away prior to further acetylation by [1Ј-13 C]aspirin. The spectra were acquired in a manner similar to that described for Fig. 1. The more significant decreases of spectral intensity resulting from MAP pre-treatment are shown by the boxes.

FIG. 3. 1 H-13 C HSQC spectra of 1.5 mM COHbA acetylated by a 10-fold molar excess of [1-13 C]MAP at 37°C for 1 h with (left) or without (right)
prior incubation with a 4-fold molar excess of IHP at 37°C for 30 min. The spectra were acquired at 25°C in a similar manner to those of Fig. 1 but with 64 scans/t 1 increment. The resonances showing the most significant reduction of intensity as a result of pre-incubation with IHP are boxed. and 11 were observed, indicating the high selectivity of the protective effect of IHP against subsequent acetylation by [1-13 C] MAP. Thus, the lack of a significant reduction in the intensities of peaks 1, 2, 3, 4, 7, 10, and 11 ( Fig. 3) indicates that the corresponding residues are located outside of the 2,3-DPG binding pocket. This result appears to be in direct contrast to the early report by Manning and co-workers (Ref. 10 and references therein), according to which no acetylation of residues outside the 2,3-DPG binding site was noted.
To further compare the specificities of aspirin and MAP as acetylating agents, COHbA was exposed initially to a 10-fold molar excess of unlabeled MAP, which was followed by exposure to [1Ј-13 C]aspirin. The HMQC spectra of COHbA acetylated with and without prior exposure to unlabeled MAP are shown in Fig. 4. The adduct spectrum corresponding to [1Ј-13 C]aspirin-acetylated COHbA (Fig. 4A) has been labeled with primes to distinguish the resonances from the adduct spectra derived from [1-13 C]MAP treatment of the hemoglobin. A comparison of Figs. 1A and 4A indicates that very similar adduct species were generated, as discussed above. A comparison of Fig. 4, panels A and B, reveals a substantial decrease in intensity of the adduct resonances 5Ј, 6Ј, 9Ј, and 8Ј as a result of the pre-exposure to MAP, in addition to smaller decreases in intensity for the other adduct resonances, which is consistent with the greater intensities for [1-13 C]MAP adduct peaks 5, 6, 9, and 8 shown in Fig. 2. These results indicate that both acetylating agents appear to be capable of targeting a similar group of residues, and both show a similar preference for acetylation of ␤Lys-82, located in the ␤-cleft involved in the binding of 2,3-DPG.
In summary, acetylation of oxyHbA or COHbA with methyl acetylphosphate under the conditions used in previously reported studies (4 -6, 10, 20, 21), as well as in a phosphate buffer, yields HMQC spectra with 10 -11 adduct resonances. In general, the pattern of acetylation appears to be similar to that obtained using [1Ј-13 C]aspirin (12). Observation of the resonance intensities at various MAP:COHbA ratios indicates a preference for acetylation of ␤Lys-82, as well as a significant degree of acetylation of the ␤Lys-59 and ␣Lys-90 residues, which are also acetylated by aspirin. Hence the acetylation profile obtained using MAP is qualitatively similar to that obtained using aspirin, with MAP acetylating a number of residues located outside of the 2,3-DPG binding pocket. Further, we have recently found that exposure of human serum albumin to MAP results in the formation of multiple adduct resonances (results not shown here). The broad reactivity of methyl acetylphosphate with various residues of hemoglobin outside of the 2,3-DPG binding site, as well as with human serum albumin, indicates that this compound would not be a suitable therapeutic agent for treatment of sickle cell disease as has been previously suggested (Ref. 4 and references therein).
Based on the reported selectivity of methyl acetylphosphate for residues in the 2,3-DPG binding pocket, Jones et al. (8) have evaluated a series of bis(methyl phosphate) cross-linking agents that stabilize the hemoglobin tetramer. These agents were reported to be highly selective in cross-linking the aminocontaining residues of the 2,3-DPG binding cleft (␤Val-1 and ␤Lys-82) as well as acylating ␣Val-1 and ␣Lys-99. Some of the same cross-linked adducts have previously been reported for bis(salicylate) derivatives (22,23). The present studies call into question how much more selective the bis(methyl phosphate) cross-linking agents actually are. A comparative study, utilizing the same conditions, for the analogous bis(methyl phosphate) and bis(salicylate) agents would appear to be desirable to address this question.