Glycerated hemoglobin, alpha 2A beta 2(82) (EF6) N epsilon-glyceryllysine. A new post-translational modification occurring in erythrocyte bisphosphoglyceromutase deficiency.

A new minor Hb fraction initially designated Hbx, has been found in the hemolysate of an erythremic patient that we have previously described with a complete erythrocyte bisphosphoglycerate mutase (EC 5.4.2.4) deficiency. Hbx (3.5% of the total) was detected by isoelectric focusing and exhibited electrophoretic and chromatographic properties similar to those of several variants of the Hb central cavity. By density fractionation of red cells, it was demonstrated that Hbx was an aging hemoglobin as in the case of glycated Hb A1c. Functional studies revealed a low oxygen affinity and almost complete inhibition of the allosteric effect of the organic phosphate effectors. Structural studies demonstrated an absence of tryptic cleavage between the peptides beta T9 and beta T10 suggesting the presence of an adduct on Lys beta 82 or on a neighboring residue. Fast atom bombardment mass spectrometry and a specific enzymatic assay with glyoxylate reductase demonstrated that the beta 82 adduct was a glycerate moiety. It was concluded that Hbx was a glycerylated Hb, alpha 2A beta 2(82) (EF6) N epsilon-glyceryllysine, to our knowledge the first example of glycerylated protein. The mechanism of formation of glyceryl Hb, which was found in the four studied subjects with a bisphosphoglyceromutase deficiency, remains to be determined.

As opposed to most eukaryotic proteins the human hemoglobin chains remain unmodified after translation. Only a small proportion of human globin chains undergo a posttranslational modification: 10% of the N-terminal amino groups of fetal y chains are acetylated in cord blood (1) and 4% of the N-terminal valine residues of the /3 chain are glycated in adult Hb AI, (2). In addition some a-or eNH2 groups of Hb A are also glycated during red cell aging (3). A low percentage of hemoglobin is also pyridoxylated in normal red blood cells (4). In individuals with diabetes mellitus, Hb glycation increases markedly (5) and the level of Hb Ale, the main glycated Hb, is a good index of diabetes severity (6). A post-translationally modified hemoglobin has also been recently described in subjects with chronic alcoholic intoxica-* This work was supported by grants from the Institut National de la Santd et de la Recherche Mddicale, from the Centre National de la Recherche Scientifique, and from the MinistGre de la Recherche et de la Technologie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. tion (7), and aspirin binds to several Hb lysines (8). This ability of Hb to form different adducts is currently used in the aim of developing anti-sickling drugs, and several chemically modified Hbs have been obtained (9). In contrast, the only case of a genetically determined post-translational Hb modification has been described in Hb Providence where Lys p82 is replaced by asparagine of which 60% is deamidated after the chain synthesis (10). The present paper reports a novel form of genetically determined post-translational Hb A modification, involving the presence of an Hb fraction with a glycerate adduct on the t-NH2 of lysine 1682. This glycerylated Hb was discovered in a patient with a complete bisphosphoglyceromutase (EC 2.7.5.4) deficiency previously described in our laboratory (11). Bisphosphoglyceromutase catalyzes the formation of glycerate-2,3-P2. Its defect produces an absence of red blood cell glycerate-2,3-P2 and induces a high oxygen affinity of the red blood cells responsible for a secondary erythrocytosis. The modified hemoglobin is present in all the family members carrying the bisphosphoglyceromutase deficiency. The functional characteristics of this modified Hb and the elucidation of its structure are described below. polyacrylamide gel (pH 6-9) (14), globin chain analysis on acid-urea-Triton acrylamide gels ( X ) , chromatography on aminopfienyl boronated agarose, in order to retain glycated hemoglobins using Glyco-Gel B resin (16), thiobarbituric acid (17), and alkali denaturation (18) tests. Hb AI, and the novel abnormal form, designated Hb,, were separated from Hb & by Bio-Rex 70 chromatography according to McDonald (19). Approximately 8 g of HbCO were applied to a 9 X 35-cm column which was developed with a linear gradient of sodium produced with 5 liters of a 0.048 M starting sodium phosphate buffer, pH 6.75, and 5 liters of the same buffer containing 0.3 M NaC1. The flow rate was 1.6 liters/h. Fractions containing Hb AI, and H b , were pooled and concentrated in a Sartorius SM 16566 ultrafiltration apparatus (Sartorius, Gottingen). Hb A1, and H b , were separated by DEAE 52 chromatography according to Abraham (20). A solution containing 250 mg of HbCO was applied to a 2.5 X 40-cm column which was developed with a linear sodium gradient obtained with 250 ml of a 0.2 M starting glycine buffer, pH 7.85, and 250 ml of the same buffer containing 0.1 M NaCl. The flow rate was 100 ml/h. Purity of the hemoglobin fractions was assessed by IEF.

Functional Studies
Oxygen Equilibrium Studies-Oxygen-binding studies in red cell suspensions were performed at 37 "C in 0.14 M NaCl, 50 mM bis-Tris buffer with an automatic continuous method (Hemox Analyzer, TCS, Southampton, PA) as previously described (21,22). Oxygenation measurements for the purified H b , component were carried out at 25 "C. The buffer solutions were made of 50 mM bis-Tris or Tris buffer (for pH values higher than 7.5), 100 mM NaCl, 0.5 mM EDTA, and 20 pg/ml catalase to avoid oxidation of the hemes during the runs. The oxygen-binding curves were recorded in solutions containing 150-200 p~ Hb on a heme basis. A spacer was introduced into the optical cuvette to reduce the light path from 1.2 to 0.2 cm. The methemoglobin formed during the 45-min duration of each deoxygenation curve never exceeded 4%. Experiments were performed at different pH values or after addition of glycerate-2,3-P~ (Na salt), inositol hexaphosphate, or NaCl as indicated in Tables I and 11. For the oxygen-binding experiments, the recording system has been interfaced to an HP 85 microcomputer which was programmed to store on tape up to 500 values of absorbance and POP. The PW and nm values representing, respectively, the partial pressure of oxygen and the Hill coefficient at half-saturation were computed from the experimental points in the range of 40-60% saturation by linear regression analysis. The amount of protons released by H b , upon oxygenation in the range of pH 6.5-9 (Bohr effect) was calculated from the linkage equation relating log Ps0 to pH (23), using a nonlinear least squares fitting procedure (24,25).
Density Gradients for Fractionation of Red Cells-Fractionation of the red cells was performed according to the method described in Ref. 26 with a discontinuous "Percoll" gradient which allows the separation of the cells according to their density. The relationship between density and age was ascertained by measuring, in hemolysates of each fraction, the activity of the glutamate oxaloacetate transaminase (27).
The hemolysates corresponding to each fraction (F, to F6 in order of increasing density) were further analyzed for their Hb content by thin plate isoelectric focusing.

Structural Studies
After desheminization by acid-acetone precipitation, the abnormal p" chain was prepared by CM-cellulose urea 8 M chromatography (281, aminoethylated, stripped of urea on a Bio-Gel Pz column, and hydrolyzed with trypsin L-1-tosylamido-2-phenylethyl chloromethyl ketone, Worthington) according to Ref. 29. Tryptic peptides were characterized by HPLC using a Beckman system 343 with a Kratos (Kratos Analytical Instruments, Ramsey, NJ) wavelength detector set at 214 nm. The column system was a Waters pBondapak C18 (10 pm, inner diameter 3.9 mm, 30 cm) (Waters Associates, Milford, MA). The solvent system was made according to Ref. 29 with some modifications using a solvent A, 0.02 M ammonium acetate (pH 5.7) and a solvent B, 0.01 M ammonium acetate (pH 5.7) mixed 5050 with acetonitrile. The gradient was linear from 0 to 25% in 30 min, from 25 to 50% in 10 min, from 50 to 60% in 15 min, and from 60 to 100% in 5 min. Amino acid composition was determined on a Blotronik 6000 1E (Biotronik, Munich) after 20 h of hydrochloric acid hydrolysis of the separated peptides. Pepsin hydrolysis of fl9-TlO was performed according to the method described in Ref. 30 using Worthington pepsin in 0.5 ml of 0.01 N HCl and an enzyme/substrate ratio 1:25 for 2 h at room temperature. The characterization of pepsic peptides was performed as described for the tryptic peptides with the following gradient from 0 to 50% in 50 min and from 60 to 100% in 5 min.
NMR Analysis-NMR analysis of H b , was performed in order to detect a possible phosphorylated adduct. The NMR measurements have been performed with an XL-100 (Varian Associates, Inc., Pal0 Alto, CA) operating in the Fourier transform mode at 40 MHz. The spectra width was 5000 Hz and the flip angle 45". The number of transients that were accumulated for the hemoglobin samples ranged from 30,000 to 130,000. During these experiments the probe temperature was regulated to 4 "C. Glucose-6-P (Glc-6-P) was dissolved in D20 (200 nM). Glc-6-P Hb was obtained according to the method described in Ref. 31. Glc-6-P Hb and H b , were deionized by chromatograpy on a column of AG-501-X8 (Bio-Rad) and concentrated before use.
Mass Spectrometry-The mass spectrometer was a MM-ZAB 2F instrument (VG Instruments) equipped with a FAB source. A xenon atom beam of 8 keV energy was focused on the target loaded with 3-10 pg of the peptide dissolved into 2 pl of a 1:l mixture of glycerol and thioglycerol. The spectra were recorded on sensitive UV paper at a speed of 1 mass unit/s. For CID/MIKE measurements, helium was introduced into the gas cell in order to reduce the main beam to about 30% of its original value. The electric sector was scanned within 150 s in the energy range 8-1 keV. In order to identify a putative adduct and its binding to Hb. , high resolution FAB mass spectrometry associated with CID/MIKE measurements was performed on the pepsic peptide 082-85 according to the method described in Ref. 32. Precise mass measurements were made at V.G. Analytical Ltd. (Manchester) on a V.G. 7070E instrument fitted with an alternating FAB probe connected to a data system. A mixture of alkaline iodides was used as reference compounds. The accurate mass determination was obtained by interpolation between the cluster ions =Rb2 RbI: (m/z 510.5417) and cesium =Rbz 1; (m/z 556.5380) using the computer averaging system.

Assay of Glycerate
Glycerate was measured by the specific enzymatic method using the glycerate dehydrogenase activity displayed by the glyoxylate reductase from spinach leaves (33,34). The method used was an adaptation of the technique described by Sallach (35) for the determination of the glycerate dehydrogenase activity of the enzyme according to the following equation.
The oxidative reaction is shifted to the right by carrying out the reaction at pH 9.2 and in the presence of a carbonyl-trapping reagent, i.e. hydrazine sulfate. The assay medium contained 0.1 M Tris-glycine, 0.03 M hydrazine sulfate (pH 9.2) in an adequate sample volume. The reaction was started at 37 "C by adding 7.5 pg of glyoxylate reductase (Sigma) and was allowed to reach completion. The resulting values were compared to those of a reference curve of absorbances versus increasing glycerate concentrations. The reaction was sensitive enough to detect 15 nmol of D-glycerate in the cuvettes. This reaction is specific and does not occur with glyceraldehyde, dihydroxyacetone, glyceraldehyde 3-phosphate, 3-phosphoglycerate, or 2-phosphoglycerate.

RESULTS
A minor abnormal hemoglobin was detected during the reevaluation of a previously described subject presenting a complete red cell bisphosphoglyceromutase deficiency (11). In brief, the main features of the patient were: a well tolerated erythrocytosis nevertheless requiring episodic venipunctures, the absence of detectable red cell glycerate-2,3-P2, and of bisphosphoglyceromutase activity leading to large modifications of the glycolytic intermediate concentrations and particularly a large increase in fructose 1,6-Pz, dihydroxyacetone phosphate, and glycerate-3-P. The propositus hemoglobin pattern previously studied by cellulose acetate electrophoresis and by IEF did not exhibit a major abnormal component.
More recent analyses by Bio-Rex 70 chromatography revealed a high level of Hb AI, of 9.6% (control, 5.4 f 0.4). This result contrasted with the absence of an abnormal level of Hb F, as estimated by alkali denaturation, and of diabetes mellitus. This prompted us to study more carefully the minor Hbs of the propositus. The isofocusing pattern of the propositus hemolysate ( Fig. 1, samples 2 and 7) and of the Bio-Rex 70 fraction eluting with the Hb A,, retention time ( Fig. 1, sample 6) demonstrated the presence of a normal amount of Hb AI, and of an abnormal minor component (Hb,) representing 3.5% of the total Hb and 40% of the Bio-Rex fraction. The PI of Hb, was intermediate between those of Hbs A,, and Alb. At this PI a faint band is present in normal control (Fig. 1, sample I ) and is sometimes increased in stored red blood cell samples (not shown). The propositus Hb, level remained constant over a period of several months. Hb, was also found at the same level in the three propositus sisters who also displayed complete bisphosphoglyceromutase deficiency. A lower amount of Hb, was detected among the members of this family as well as of another unrelated family presenting a partial bisphosphoglyceromutase deficiency (36). Large quantities of Hb, mixed with Hb Al, were isolated from the propositus hemolysate by chromatography on Bio-Rex 70. Hb, was separated from Hb A,, by chromatography on a DEAE-52 column, and the properties of the purified abnormal hemoglobin were studied. On citrate agar electrophoresis Hb.
mobility was similar to that of Hb A, and Hb F, a property which together with the behavior of these Hbs on Bio-Rex is observed in several abnormal hemoglobins with a substitution of one of the residues lining the central cavity between the 2 p chains (37,38). In urea buffer, the migration of the , f 3 chains in cellulose acetate electrophoresis was more anodal than PA and chains. The / 3 chains were separated from a, / 3, ,f3A~a, -yA, and -yG chains by acrylamide electrophoresis in urea-Triton buffer. H b , was not retained on affinity chromatography column Glyco-Gel.B, indicating that in contrast to Hb A,, it did not contain a cis-diol adduct (19). No reaction was obtained with the thiobarbituric test (17) indicating that Hb, was not a glycated Hb. Since several minor Hb fractions have been suspected to contain phosphorylated adducts (19) we submitted Hb, to NMR analysis. Compared to control Glc-6-P Hb, the NMR spectrum for Hb, did not exhibit any peak corresponding to a phosphorylated derivative (Fig. 2).
Kinetics of Hb, Formation-To explain the presence of Hb, in the propositus family it could be postulated that the abnormal component was the product of a p" gene not yet detected or a post-translationally modified normal hemoglobin. To test this latter possibility, the kinetics of formation of Hb, were studied. A density gradient separated the propositus red blood cells into six different fractions (Fig. 3). The glutamo-oxalo-transaminase red blood cell activity decreased from the lightest fraction (Fl) to the densest (Fs), indicating that the cells have been separated according to age, the oldest being the densest as expected. The concentrations of Hb A,,    that line the entrance to the DID2 cleft in normal Hb A (40).
The effects of heterotropic cofactors for H b , and Hb A are summarized in Table 11. Whereas the effects of glycerate-2,3-Pz and inositol hexaphosphate on the functional properties of Hb, are nearly abolished, the chloride effect is only halved. The latter result is in agreement with the existence of two classes of binding sites for chloride in Hb A, one with a high affinity at the alaz N-C interface and the second, with low affinity, at the +amino group of Lys (EF6) p82 (41,42). This result suggested that a residue of the central cavity might be involved in the abnormal function of Hb, and focused attention on lysine (EF6) p82. The oxygen Bohr effect was measured in the range of pH values between 6.5 and 9. Compared to Hb A the maximum proton release upon oxygenation of Hb, was slightly diminished (Table I1 and Fig. 4).
Structural Studies-The structural study of the Hb, was performed on a sample isolated by the chromatographic steps already described. The globin chains were separated by CMcellulose chromatography in urea buffer. The elution profile was not strikingly different from that produced by globin chains from Hb A. The chains were aminoethylated, submitted to a tryptic hydrolysis, their peptides separated by HPLC on a pBondapak C18 column, and their amino acid composition determined. Fig. 5 presents the HPLC profile of the tryptic peptides of the aminoethylated DX chains. It is similar to that of PA aminoethylated chains (not shown) except the presence of a new peak eluted at the end of the chromatography and presented an amino acid composition identical to that of the sum of the amino acids present in T9 and T10 (Table 111). The small quantities of T9 and T10 present most probably arose from a slight contamination by pA1o peptides as indicated by the presence of TIC. The absence of tryptic hydrolysis in p" of the bound between Lys 82 and Gly 83 suggested that an adduct was bond to Lys 82 or to an amino acid in its immediate vicinity.
Mass Spectrometry Analysis-To assess the precise location of the putative adduct and identify its structure it was decided to use mass spectrum analysis. Since the size of the 29-residue PT9-10 was too large for suitable results with this method, 30 nmol of the p"T9-Tl0 were hydrolyzed by pepsin, an enzyme  which releases the tetrapeptide 82-85: Lys-Gly-Thr-Phe (30). Fig. 6 shows the pattern of the /3T"l?-T10 pepsic hydrolysate obtained after HPLC. Among the 12 peaks obtained, the amino acid composition of the peak 5 (Table IV) corresponded to the 82-85 sequence. This peak was three times lyophilized and subjected to mass spectrum analysis. The positive FAB mass spectrum, using thioglycerol as matrix, is presented in Since the nominal m/z value for MH+ was an even number, the molecule should contain an odd number of nitrogen atoms. Thus, a computational examination of the possible elemental compositions indicated that the best fitting was C24H3809N5 (theoretical 540.26695). The error was less than 2 ppm. It was already established that acidic hydrolysis of the modified peptide gave the same amino acid composition as the unmodified one (Table IV). Thus, the modified peptide 82-85 included, in addition to amino acids lysine, glycine, threonine, and phenylalanine, a new moiety corresponding to C3H403.
In order to detect the position of the modification on the sequence a CID/MIKE spectra was measured on m/z 540. It was shown that a partial N-terminal sequence could be obtained (Fig. 8). Two main ions were observed the former corresponded to the loss of a phenylalanine residue at m/z 375; the latter to the loss of both phenylalanine and threonine residues at m/z 274. Thus, the modification was associated with the remaining dipeptide, i.e. Lys-Gly, but no further characteristic fragmentation could be detected. The most likely hypothesis to explain these findings is a modification of the lysine residue on its €-amino group by amidation with a molecule of glyceric acid (Scheme 1). This proposal takes into account the recovery of unmodified lysine and glycine upon acid hydrolysis of the p82-85 pepsic peptide (Table IV). To confirm this "glycerate hypothesis" we developed a specific enzymatic assay described under "Methods" and looked at the presence of glycerate on the pT9-10 peptide.
Final Evidence for Glycerate as the Adduct in Hb,"Since 6 N HC1 hydrolysis of the modified peptide released stoichio-CHOH I CH, OH I GLYCERIC ACID SCHEME 1 metric amounts of nonsubstituted amino acids, the adduct, if not destroyed, should be found free in the hydrolysate. In a preliminary control we observed that glycerate added to a mixture of amino acids was quantitatively recovered after 22 h of hydrolysis. Subsequently, sufficient quantities of pT9-T10 were isolated by HPLC and subjected to 6 N HC1 hydrolysis for 22 h at 110 "C. The hydrolysate was lyophilized three times and divided into two equal aliquots for amino acid analysis and glycerate determination. In two separate experiments glycerate was present in stoichiometric quantities compared to the peptide, confirming the hypothesis formulated according to the mass spectrometry analysis. According to these results we concluded that the structure of Hb, was azAP282 (EF6) N-glyceryllysine, and we propose to name it glyceryl Hb.

DISCUSSION
This report describes a new post-translational modified hemoglobin: the glyceryl hemoglobin. This novel minor hemoglobin fraction is progressively formed in the red blood cells of patients with total bisphosphoglyceromutase deficiency. The formation kinetics of this hemoglobin are close to those of Hb AI, and as in the case for this later form, the level of glyceryl Hb is maximal in the oldest red cells. The arguments used to assess that Hb A, is an aging hemoglobin, nonenzymatically synthesized during the red cell life span (39), can be applied to the synthesis of glyceryl Hb. In contrast to Hb AI,, which is found in normal nondiabetic subjects, it is difficult to assess whether glyceryl Hb is formed in subjects with normal bisphosphoglyceromutase activity. A minor band at the position of Hb, is sometimes present in the IEF pattern from normal red cells, but it is very faint and scarcely more visible in stored red cells.
Glyceryl Hb formation is effectively under genetic control since it is present at similar levels in all the patients with complete bisphosphoglyceromutase deficiency. This Hb was also detected in the hemolysates from subjects with partial deficiency who originated from different families (36). In the eukaryotes, hemoglobin represents the most studied genetic system, and a large number of abnormal hemoglobins have been described that are produced by genetic events involving the globin gene domain. Glyceryl Hb is to our knowledge the first example of abnormal hemoglobin whose synthesis is linked to an abnormal nonglobin gene. It is interesting to note, that the bisphosphoglyceromutase gene involved is func-tionally closely related to Hb function since it synthesizes the enzyme producing and hydrolyzing the main allosteric effector of Hb oxygenation: the glycerate-2,3-P2.
The presumptive localization of the adduct on lysine ,882 was relatively straightforward because the peptidic bond between ,8T9 and pTl0 normally hydrolyzed by trypsin with a high yield in BA chain was not cleaved in case of glyceryl Hb.
Such an absence of action of trypsin has been previously described in another example of eNH, lysine ,882 in vitro modification involving adduction with glyceraldehyde (8). Before mass spectrometry analyses had been performed on the abnormal peptide it was not possible to identify the chemical structure of the adduct, and only partial information indicated that the adduct was not a compound bound to the Lys ,882 by a Schiff base, since it was not reduced by tritiated borohydride. By contrast, the high resolution FAB mass spectrometry associated with CID/MIKE measurements was extremely efficient in identifying the adduct and its binding to the lysine on small quantities of purified material. It is a new example of the efficiency of the FAB mass spectrometry which has been used already successfully for fast characterization of genetically determined abnormal hemoglobins on microquantities of material (43). Such a method would be very useful for future determinations of other adducts of hemoglobin, particularly if, as in the present case, a specific enzymatic determination can be used together with the mass spectrometry analysis for the final identification of the adduct.
Precise oxygen-binding measurements revealed that the glyceryl Hb has a low oxygen affinity, slightly reduced Hill coefficient, and an almost complete inhibition of the allosteric effect of the organic phosphate effectors; these results are expected from the disappearance of the positive E-NH, of lysine ,882, one of the positive groups involved in the binding of glycerate-2,3-P2 to deoxy Hb (40). Similar results including the observed halved effect of chloride on oxygen affinity were obtained with Hb Providence (44) in which Lys ,882 is absent. Moreover, the decreasing order of oxygen affinities for various hemoglobins allelic at the ,882 locus, Hb Rahere (Lys + Thr) (45) > Hb Helsinki (Lys + Met) (46) > Hb A (Lys) > Hb Providence I (Lys + Asn) > glyceryl Hb (Lys + N'-glyceryllysine) > Hb Providence I1 (Lys + Asp) demonstrated the marked effect on the oxygen affinity of the nature of the side chain of the residue ,882. The polarity of the lateral chain of glyceryl Hb is effectively intermediate between that of Hb Providence I which is neutral and of Providence I1 which is acidic. Indeed, precise information concerning the location of the adduct, the nature of its linkage, and even its structure could be deduced from careful analysis of some electrophoretic, chromatographic, and functional properties of glyceryl Hb.
The mechanism of formation of glyceryl Hb remains to be determined, and it would probably be difficult. The absence of bisphosphoglyceromutase activity in the propositus erythrocytes produces large modifications of the glycolytic intermediates involving increases of fructose-1,6-P2, of triose phosphates, and of glycerate-3-P. Preliminary results indicated that these perturbations did not produce an elevation of glycerate which was undetectable in the propositus erythrocytes as in the controls. A good candidate, as a primer, could be the glycerate-1,3-P2. The concentration of glycerate-1,3-Pz cannot usually be determined in the red cells because of its extreme instability, but it is reasonable to assume that its level may be raised in the absence of bisphosphoglyceromutase. This ester phosphate could bind to positive groups lining the central cavity between the two p chains. One may postulate that its CI carboxylic group will amidify eNH2 of lysine PS2 after hydrolysis of its highly unstable phosphate group. Since no phosphate was found in glyceryl Hb such a reaction would be followed by the hydrolysis of the remaining phosphate group which could occur under the action of an acidic phosphatase of the red cells or through the esterasic action of Hb itself recently described by Elbaum (47), an esterase activity which involves precisely the lysine @82.