THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 254, No. 10, Issue of May 25, pp. 3892~3898,1979 Printed in U.S. A.

Structural Heterogeneity of Human Hemoglobin A due to Nonenzymatic Glycosylation*

(Received for publication, November 20, 1978)

H. Franklin Bunn, Robert Shapiro, Michael McManus, Laura Garrick,$ Melisenda J. McDonald, Paul M. Gallop, and Kenneth H. Gabbay@ From the Howard Hughes Medical Institute Laboratory at Harvard Medical School, the Department of Medicine, Peter Bent Brigham Hospital, the Children’s Hospital Medical Center, and the Departments of Medicine, Biochemistry, and Pediatrics, Harvard Medical School, Boston, Massachusetts 02115

Human hemolysate contains several glycosylated mi- nor hemoglobms (Hbs AI,~, A Ia2, Am, and AI,) which can be chromatographically separated from the major com- ponent, Hb Ao. The glucosyl-ketoamine linkage in Hb AI, can be detected calorimetrically by the thiobarbit- uric acid test. When chromatographed on Bio-Rex 70 resin, Hb A0 is eluted as a single peak following Hb AI,. We have found that the leading edge of Hb Ao, as well as Hb AI,, contains carbohydrate as detected by the thiobarbituric acid test. Both glycosylated components were comparably increased in the diabetic. There was a corresponding increase in the incorporation of trit- ium from [3H]borohydride into diabetic Hb A+ After Hb A0 was incubated with [‘4C]glucose it was chromato- graphed on Bio-Rex 70 resin. The specific activity pro- file corresponded closely to the thiobarbituric acid test profile. Parallel incubation with glucose having 3H bound to the second carbon atom confirmed that both the synthetic Hb AI, and the glycosylated Hb Ao have undergone the Amadori rearrangement to the more stable ketoamine linkage. We estimate that 8 to 10% of Hb A0 in normal red cells is glycosylated. Autoradi- ograms of tryptic peptide maps indicate that several sites on both the a and /3 chains are modified, including the NH2 terminus of the (Y chain. Comparison of ion exchange liquid chromatograms of synthetic lysino-l- deoxysorbitol with those of acid hydrolysates of [3H]- borohydride-reduced native Hb A0 and [‘4C]glucosyl Hb A0 shows that the glucose is bound to lysines. This nonspecific reaction of glucose with lysine probably occurs in other proteins and may contribute to some of the long term complications of diabetes.

Human hemoglobin has probably been studied more exten- sively than any other macromolecule. Its structural and func- tional characterization has been simplified by the fact that human hemolysate contains one major hemoglobin compo- nent, Hb A0 (cy&), which comprises over 90% of the total

* This work was supported by the Howard Hughes Medical Insti- tute, Grants AM 18233, AM 15019, and HL 20539 from the National Institutes of Health, and the Nehemias Gorin Foundation. This work was presented in part at the annual meeting of the American Society of Biological Chemists, June, 1978 and was published in abstract form (1978) Fed. Proc. 37, 1390A and (1978) Blood 52(Suppl.), 119). 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 USC. Section 1734 solely to indicate this fact.

$ Present address, Department of Biochemistry, State University of New York, Buffalo, N. Y. 14214.

1 Established Investigator of the American Heart Association.

protein. It is widely assumed that our knowledge of the primary structure of Hb A0 is complete.

Besides Hb A”, two types of minor hemoglobin components have been identified in human red cells. Hbs A% (a&) and F (a2yn) differ from Hb Ao in that their non-a subunits are products of different globin chain genes. Other minor hemo- globin components in human hemolysate are post-transla- tional modifications of Hb A0 (1). Hemoglobin AI, is the most abundant minor component in human red cells, comprising about 5% of the total. Hb AI, has attracted considerable interest because it is elevated 2- to 3-fold in patients with diabetes mellitus (2-4). This hemoglobin is formed by slow nonenzymatic condensation of glucose with the NHz-terminal amino group of the p chain (5-8) (see Scheme 1). The en- hanced protonation of the ketoamine group at physiologic pH pulls the equilibrium to the right.

The apparent specificity of glucose for the NH2 terminus of the /3 chain is not readily explained. The formation of a Schiff base (aldimine) linkage is favored by the relatively low pK, of the NHn-terminal amino groups which allows them to be effective nucleophiles at physiological pH. Accordingly, one might predict that this site oxi the (Y chain would also be glycosylated. Furthermore, the e-amino group of lysine resi- dues, despite their relatively high pK, values, are also capable of forming aldimine and ketoamine adducts with glucose (9- 11).

In this report, we demonstrate unexpected structural het- erogeneity of the major human hemoglobin component (Hb Ao) which is due to the same kind of nonenzymatic glycosyl- ation that characterizes Hb AI,.

MATERIALS AND METHODS

Blood specimens were obtained from normal volunteers and pa- tients with diabetes mellitus. Red cell hemolysates, prepared by the method of Drabkin (12), were gassed with carbon monoxide and chromatographed on Bio-Rex 70 cation exchange resin (Bio-Rad, Inc., Calif.). Usually, 2 g of hemoglobin protein was applied to a 5 X 38 cm column. The non-hemoglobin protein and negatively charged minor hemoglobin components (Hbs AI,, , AI.,, AIM, and AI,) were eluted by the recently described procedure of McDonald et al. (13). The major hemoglobin component (Hb &) was then eluted by a linear NaCl gradient going from 0.1 M to 0.5 M. Column fractions containing isolated hemoglobin components were pooled and, when necessary, concentrated by pressure filtration (Amicon PM 10 membrane).

Poly(L-lysine) hydrobromide was purchased from Sigma Chemical Co. It had a molecular weight range of 30,000 to 70,000 and consisted of linear polymers of L-lysine linked by peptide bonds at a-NH2 and carboxyl groups.

Incubations-Purified Hb & (or in some cases polylysine) was incubated with uniformly labeled n-[‘4C]glucose, n-[2-3H]glucose, or with unlabeled D-ghCOSe for up to 26 days in a sterile solution of Krebs-Ringer phosphate, pH 7.4, at 37°C. In one experiment, intact

3892

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from

Structural Heterogeneity of Human Hemoglobin 3893

“7 =O "7 = N-%A CH -NH-BA HYOH

I 2 yH2-N+H2-%A HCOH Amadori c =o Hf

BA-NH~ + HOyH Z=f HO&I \ 5: =o

HFOH HdOH - HO& - HOCH

nyon HdOH HdOH H'OH

dH20H &OH

F

CH20H dHZOH HYOH

CH20H

Schiff Base Ketoamine Aldimine

Jt

H2C-NH-BA

"d---J Hd d ;H20H dH20n dH20H

GlUCOSS 5-N-valyl- 1-deoxyglucose

SCHEME 1

B-N-valyl- l-deoxyfructose

red cells were incubated with 12 mM L-[‘%]glucose in Krebs-Ringer phosphate’ and plasma (1:l) at 37°C for 7 days. Unbound glucose was separated from protein by passage through Sephadex G-25. The hemoglobin solutions were then analyzed by the chromatographic procedure described above and in Ref. 13.

Glucosyl valine (valyl-1-deoxyglucose) was prepared as described by Dixon (14) from [‘%]valine and unlabeled glucose. The adduct was then reduced with unlabeled borohydride.

Structural Analyses-Glucose covalently bound to either hemo- globin or polylysine by ketoamine linkage was measured by treatment with oxalic acid to generate 5-hydroxymethylfurfural which was then condensed with thiobarbituric acid (TBA) to form a colored adduct, having an absorbance maximum at 443 nm (7). These results can be expressed as specific color activity (SCA3 or A443 ,/mg of protein) (see TBA test, Table I).

Hemoglobin and glycosylated polylysine were treated with either unlabeled or 3H-labeled’ sodium borohydride. The enhanced stability of the amino-1-deoxysorbitol linkages enabled the carbohydrate to remain attached to the protein throughout the analytic procedures. In addition, the reaction with labeled borohydride was useful in the detection of aldimine and ketoamine linkages. The reductions were carried out as described by Bookchin and Gallop (5). Samples (0.3 to 1.0 mM (o/3 dimer) in 0.1 M potassium phosphate, pH 7.0) were reacted

with a 200-fold excess of NaBH4 for 10 min at room temperature followed by 50 min at IO’C. Excess borohydride was removed by acidification to pH 5, followed by gel filtration on Sephadex G-25 at 4°C in either 0.01 M NHdHCOJ pH 7.8 (polylysine) or 0.05 M Tris- HCl, pH 8.0 (hemoglobin). The polylysine samples were lyophilized.

Hemoglobin samples were converted to globin by acid/acetone precipitation, and separated into OL and /3 subunits by carboxymethyl- cellulose chromatography in 8 M urea (15). Purified globin subunits were aminoethylated and digested with trypsin as described previ- ously (16). The tryptic peptides were analyzed by two-dimensional maps on cellulose thin layer plates (16). Autoradiograms were pre- pared on X-omat film (Eastman Kodak Co.).

Globin subunits, glycosylated polylysine, and peptides were hydro- lyzed in constant boiling HCl, 108°C for 24 h. Some specimens were hydrolyzed in 2 N KOH and desalted by the method of Hauschka (17). Preparative separations of labeled amino acids were carried out by ion exchange chromatography as modified from Jones (18). Hy- drolysates were applied to a column (0.9 x 16 cm) of DC-GA resin (Durrum Chemical Co., Sunnyvale, CA) and eluted with a 200~ml linear gradient of 0.2 M pyridine acetate, pH 3.1, to 1.5 M pyridine acetate, pH 5.0. Column pressure averaged 300 psi.

Peaks from this column were pooled and dried down at 60°C under nitrogen. They were then redissolved in 1 ml of 0.4 M periodic acid in water to remove sugars so that the amino acid involved could be

~~~~ _________ ’ New England Nuclear Co., Boston, MA. ’ Krebs-Ringer phosphate contains 0.13 M NaCl, 0.013 M MgC12,

0.008 M CaCl2, 0.05 M KCl, and 0.01 M sodium phosphate, pH 7.4. ‘The abbreviations used are: SCA, specific color activity; TBA,

thiobarbituric acid.

(-) 4

0

- Normal Hemolysate

Diabetic Hemolysate

(---I 08

06 k

0.4 s Q

L 08 Q

k 06

04 8 k

0.2

3

FRAC JION NUMBER FIG. 1. Elution pattern of normal and diabetic hemolysates, chro-

matographed on Bio-Rex 70 (13). The concentration of hemoglobin is shown by p. - - - shows the presence of ketoamine-linked sugar detected by the TBA test (Ref. 7 and Table I). Hemoglobin-bound sugar was detected in the leading edge of the Hb Ao peak as well as in Hb Arc.

identified. After reacting for 2 h in the dark at room temperature, samples were lyophilized and subjected to amino acid analysis. Amino acid analyses were done on a Durrum D-500 analyzer by AAA Laboratories, Seattle, WA.

Aliquots of chromatographic fractions of hemoglobin, globin sub- units, peptides, and amino acids were counted in a liquid scintillation counter (Isocap 300, Searle Analytical) using Beckman Ready Solv liquid scintillation fluid. Counts were corrected using a standard quench curve.

RESULTS

Glycosylation of Intact Hemoglobin--The chromatograms from a normal and a diabetic hemolysate are shown in Fig. 1. As expected, the diabetic had about a 2-fold increase in Hb Arc. This figure also shows the ketoamine-linked carbohydrate of the hemoglobin components, determined by the TBA test of Fliickiger and Winterhalter (7). The Hb Arc peaks of the normal and diabetic are both strongly positive with specific color activities @CA-A/mg of Hb) of 0.13 and 0.17, respec-

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from

3894 Structural Heterogeneity of Human Hemoglobin

tively (Table I). In addition, a lesser amount of color was developed in the leading edge of the major hemoglobin com- ponent. Hb Ao of the diabetic contained about twice as much total TBA color as the normal. These results suggest that the major hemoglobin peak contains ketoamine-linked carbohy- drate in proportion to the concentration of Hb AI,. A similar pattern was observed when the trailing edge of purified Hb A0 was incubated for 21 days with 15 mM [‘4C]glucose (Fig. 2). In agreement with earlier studies (7, 13), Hb Ar, was formed. This synthetic component contained 0.9 mol of glucose/P chain. However, about 4 times as much [‘4C]glucose was incorporated into the leading edge of Hb A0 as into Hb AI,. The TBA test revealed considerably less specific color activity in the [‘4C]glucose-A,, than in the [‘4C]glucose-Ar, (See Fig. 2 and Table I). Thus, the TBA test was severalfold less sensitive in detecting the glucosyl linkages in Hb A0 than in Hb AI,. This difference could not be explained by decreased hydrolysis

TABLE I

Estimation of carbohydrate content of hemoglobins by the TBA test

The TBA test for amino-1-deoxyfructose was performed using a modification of the method of Fluckiger and Winterhalter (7). A mixture of 1.0 ml of sample (0.5 mM c$ dimer for Hb &, 0.1 mM for Hb Ai., 0.16 mM in glucose for polylysine incubated with [%]glucose) and 0.5 ml of 1.0 N oxalic acid was heated in a boiling water bath for 4 h in order to convert the bound sugar to free 5-hydroxymethylfur- fural. After cooling, the protein was precipitated by addition of 0.5 ml of 40% trichloroacetic acid and the mixture was passed through a 0.45-p Millipore filter. One milliliter of the fdtrate was added to 0.5 ml of 0.05 M thiobarbituric acid and incubated at 40°C for 30 min. The complex between thiobarbituric acid and 5-hydroxymethylfur- fural was detected by recording the spectrum from 400 to 500 nm on a Cary 118C recording spectrophotometer. SCA was calculated by dividing the absorbance at the peak (443 nm) by the amount of protein or bound glucose in the specimen. The SCA values for Hb A, incubated with [‘%]glucose were corrected by subtraction of the SCA value for the unincubated Hb &. In these cases, the Hb & was purified on Bio-Rex 70 and the fractions used for incubation were taken from a narrow region of the trailing edge where SCA values varv little with elution time.

Hemoglobin SCA

Arc (normal) Arc (diabetic) Ao (normal) Ao (diabetic)

A/mg Hb”

0.130 0.169 0.0019 0.0044

A/pm01 $3” A/p7d glucoseh

4.18 4.42 5.40 0.061 0.73 0.141

” Determined from unincubated (native) hemoglobins. * Determined from hemoglobins following incubation of [‘%]glu-

case with purified Hb Ao.

FRACTKJN NlMEER

FIG. 2. Synthesis of Hb Ai. and glycosylated Hb Ao. Hb Ao was isolated from normal hemolysate, incubated with 15 mM D-[‘%]gIu- case for 24 days under physiologic conditions (Krebs-Ringer phos- phate buffer, 37°C) and then chromatographed on Bio-Rex 70 (13). Hemoglobin concentration, M, radioactivity, o--O, TBA color, A-A. Glucose was incorporated not only into Hb AI,, but also into the leading edge of Hb Ao. Note that the efficiency of TBA color development was far greater for Hb AI, than for glycosylated Hb A,,.

-40 -I

fT?Acrw lw+fBEt?

FIG. 3. Incubation of Hb & with D-ghCOSe doubly labeled with 14C and 3H bound to the second carbon atom. The relative amounts of 3H and 14C in the labeled glucose solution are shown in the bar graph inset. After 22 days of incubation, ‘%-labeled Hb Ai, was formed as in Fig. 2. In addition, glucose was incorporated into the leading edge of Hb Ao (shown on a different scale). However, only a trivial amount of 3H radioactivity was detected in either hemoglobin. The hydrogen atom bound to the second carbon atom was expelled when the glucose-hemoglobin adduct underwent the Amadori rearrangement from the aldimine to ketoamine form.

of the sugar from the hemoglobin during oxalic acid treatment prior to reaction with TBA. Eighty-one per cent of the radio- activity from the & was recovered after hydrolysis, compared with 97% for AI,. As the following experiment shows, the difference was also not due to failure of the glucose linkages in A0 to undergo the Amadori rearrangement.

In order to elucidate the nature of this linkage, we incubated purified Hb & with a mixture of [2-“HIglucose and [‘“Cl- glucose under the same conditions described above (see legend of Fig. 3 for details). The ratio of 3H to 14C radioactivity incorporated into hemoglobin provides an accurate measure of the Amadori rearrangement of the aldimine to the more stable ketoamine linkage (7) as shown by the reaction scheme in the introduction. After 6 days of incubation, about 60% of the newly synthesized Hb AI, and Hb & had undergone the Amadori rearrangement. As shown in Fig. 3, after 22 days, very little 3H radioactivity was detected in either Hb AI, or Hb Ao. We estimate that 91% of the Hb Ai, and 93% of the glycosylated Hb A0 had undergone the Amadori rearrange- ment.

In order to determine whether glycosylation of Hb AI, and Hb A0 takes place in the intact red cell under controlled physiologic conditions, we incubated a sterile suspension of normal fresh human red cells with 12 mru L-[14C]glucose for 2 days at 37OC.4 Considering that hemoglobin has a 2-fold axis of symmetry, the L isomer should interact with hemoglobin in the same manner as the D isomer and yet would not participate in any enzymatic reactions such as glycolysis and the hexose monophosphate shunt. The radioactivity profile, shown in Fig. 4, indicates direct condensation of glucose with hemoglo- bin inside the intact red cell, resulting in the synthesis of Hb Arc. Incorporation of radioactivity into Hb & was also ob- served.

Structural Analysis of Glycosylated Hb Ao-Hemoglobin Ao, incubated with [‘4C]glucose under physiologic conditions, incorporated radioactivity into Hb AI, and into the leading edge of Hb A0 (See Fig. 2). Following reduction with unlabeled borohydride, the labeled Hb & was separated into (Y- and /3- globin subunits. As the elution profile in Fig. 5 (top) shows, the radioactivity was evenly distributed between LY and p chains. The glycosylated species were eluted slightly ahead of the unreacted globin, indicating that they were slightly more negatively charged. When Hb Ao was reduced with [3H]boro- hydride (without prior incubation with glucose), the elution

4 We thank Dr. George Cahill for suggesting this experiment.

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from

Structural Heterogeneity of Human Hemoglobin 3895

FRACTION NU!BEE

FIG. 4. Incubation of intact red cells with rJ4C]glucose. Incuba- tion conditions are described under “Materials and Methods.” After 48 h, the red cells were lysed and the hemoglobin was chromato- graphed on Bio-Rex 70. The metabolically inert sugar was incorpo- rated not only into Hb AIM, but also into Hb Ao. Hemoglobin concen- tration, M, radioactivity, 0- - -0.

Hb A, INCUEATED W/TH~4~-GLUCOSE y

0.8-

I i 5 04-

0.4 400

0 0 40 60 0 80

FRACTION NUMBER FIG. 5. Separation of globin a and /? subunits on CM-cellulose in

8 M urea. Top, glycosylated Hb &, following incubation with D- [‘%]glucose and reduction with unlabeled borohydride. Bottom, Hb A0 following reduction with r3H]NaB&. Protein concentration is measured by Azwn,,, (M), while radioactivity is shown by 0- - -0. In each case, radioactivity was located in the leading edge of (Y and B chains.

profile of the globin subunits showed a very similar pattern (Fig. 5, bottom) with maximal specific radioactivity at the leading edges of the (Y and p subunits. These results suggest that synthetic glycosylated Hb Ao has a distribution of ke- toamine-linked glucose similar to that of authentic Hb &. Hb Ao from a diabetic individual had 2.6-fold more incorporation

of tritium from [3H]NaBH4, compared to normal Hb A0 (data not shown).

In order to determine which residues are glycosylated dur- ing incubation of Hb AO with [“‘Clglucose, we treated the leading edge of labeled Hb A0 with unlabeled borohydride. The (Y and /3 subunits were isolated, acid-hydrolyzed, and subjected to amino acid analysis. As shown in Fig. 6, middle and bottom, a predominant radioactive peak was observed in a region following the elution of phenylalanine and preceding the elution of histidine, lysine, and arginine. The identification of this peak as a glucose-lysine adduct was established in two ways. First, synthetic lysino-1-deoxysorbitol was prepared by incubating polylysine with [%]glucose under conditions sim- ilar to those employed in the hemoglobin incubations; after separation of the protein from free [%]glucose by gel fiitra- tion, it was reduced with unlabeled borohydride and subjected to acid hydrolysis. The amino acid analysis in Fig. 6, top, shows that the synthetic lysino-1-deoxysorbitol eluted at pre- cisely the same place as the ‘%-labeled peaks obtained from the subunits of Hb Ao. Secondly, periodate treatment of the radioactive peaks obtained from hydrolysates of polylysine and a-globin yielded unlabeled lysine. In contrast, the radio- active amino acid obtained following acid hydrolysis of syn- thetic Hb AI, eluted with the solvent front, as did synthetic valyl-1-deoxysorbitol. Following treatment of the glucosyl va-

I 600 I

I PL

Phe

400- i “i’r

I

zoo-

a CHAINS

?

4 .8 CHAfNS 6a-

PI%

4co- + “i” r

zoo-

0 10 20 30 40 50 60 70

FRACTDAf NUMBER

FIG. 6. Separation of labeled amino acids by ion exchange liquid chromatography on cation exchange resin. Top, polylysine (PL) in- cubated with D-[‘4C]ghCOSe, reduced with unlabeled borohydride, and subjected to acid hydrolysis. Glycosylated Hb A,,, following incubation with D-[‘4C]ghCOSe, was reduced with unlabeled borohy- dride and separated into (Y- and P-globin subunits, each of which was acid-hydrolyzed. Middle, a chains; bottom, p chains.

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from

3896 Structural Heterogeneity of Human Hemoglobin

line with periodate, 82% of the radioactivity was recovered as valine.

Hemoglobin A0 was isolated from normal (unincubated) hemolysate on Bio-Rex 70 (13) and then treatedwithr3H]- borohydride without any prior incubation. Following acid hydrolysis of the separated (Y and p chains, the amino acids were separated as above. As shown in Fig. 7, a prominent “H- labeled peak corresponding to lysino-1-deoxysorbitol was found following acid hydrolysis of the isolated (Y chain. In contrast, the elution pattern of the /3 chain consistently re- vealed considerable heterogeneity with several radioactive peaks besides lysine-l-deoxysorbitol (Fig. 7). Similar hetero- geneity was seen following alkaline hydrolysis of the p chain. Apparently, the j3 chain takes up tritium at sites other than glucose-lysine adducts. This nonspecific uptake of tritium was circumvented when Hb A0 from a diabetic was reduced with [3H]borohydride. As Fig. 8 shows, the acid hydrolysate from normal A0 globin contained several labeled peaks besides that corresponding to lysino-1-deoxysorbitol while a comparable preparation of A0 globin from a diabetic showed a predomi- nant lysino-1-deoxysorbitol peak. Diabetic Hb A0 contains relatively more covalently bound glucose compared to normal Hb A,,. These results demonstrate the presence of the glucose- lysine adduct not only in Hb & incubated with glucose but also in naturally occurring Hb Ao.

Following incubation of Hb A0 with [‘4C]glucose and reduc- tion with unlabeled borohydride, the (Y- and /?-globin subunits were isolated, digested with trypsin, and then analyzed by two-dimensional peptide mapping on cellulose thin layers. Fig. 9 shows the ninhydrin-stained fingerprints of the (Y and ,L? chains and the corresponding autoradiograms. The p chain contained several prominent radioactive peptides, none of which correspond precisely with ninhydrin-positive peptides. Since only a small proportion of the p chains was glycosylated at each of these sites, it is not surprising that they were not detected by ninhydrin. The (Y chain, shown in Fig. 9A, con- tained one prominently labeled peptide. Analysis of this pep- tide following purification indicated that it was a-Tp-1 with a

~A o CHAINS

Phe

j.

,B CHAINS

0 10 20 30 40 50 60 70

FRACZ’OA’ NUMBER

FIG. 7. Separation of labeled amino acids as in Fig. 6. Normal Hb Ao was reduced with [“Hlborohydride, separated into (Y- and P-globin subunits, and then hydrolyzed. Top, a chains; bottom, ,6 chains.

2oco-

Phe

+

Phe

GLOB/N FROM D/ABET/C

FRACTION NUMBER

FIG. 8. Separation of labeled amino acids as in Fig. 6. Top, poly- lysine (PL) after incubation with unlabeled glucose followed by reduction with [“Hlborohydride and acid hydrolysis. Middle, normal globin reduced with [3H]borohydride and acid hydrolysis. Bottom, diabetic globin treated in the same manner.

blocking group at the NHP-terminal valine.5 In addition, sev- eral other radioactive (Y chain peptides were also noted. The relative intensity of 14C-labeled a-Tp-1 varied considerably with the sampling of hemoglobin from the leading edge of the Hb A0 peak. The hemoglobin that was eluted initially (and comprised the shoulder of TBA-positive material in Fig. 2) was relatively rich in glycosylated cy-Tp-1 (Fig. 74 while the relative amount of this labeled peptide decreased considerably when a larger portion of the labeled Hb A0 was analyzed. In view of the amino acid analyses shown in Figs. 6 and 7, the remaining labeled peptides of the LY chain and all those of the p chain are likely to be glycosylated at the e-amino group of lysine residues. It is unlikely that trypsin cleaves at these lysino-1-deoxysorbitol residues (19). We are currently at- tempting to prepare sufficient amounts of these purified la- beled peptides for structural analysis.

Relative Stabilities of Hb AI, and Glycosylated Hb Ao-Hb A0 was incubated with 15 IIIM [‘4C]glucose for 15 days as described under “Materials and Methods.” Following dialysis, the radioactive Hb Ai, and Hb A0 peaks were isolated and reincubated under identical conditions in the presence of 15 mM unlabeled glucose for 5 additional days. No radioactivity was lost from Hb Ai, during this period. The content of

’ FL Shapiro, M. McManus, and H. F. Bunn, unpublished obser- vation.

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from

Structural Heterogeneity of Human Hemoglobin

FIG. 9. Peptide maps and autoradiograms following tryptic digestion of (Y and p chains of ?Xabeled glycosylated Hb A”.

covalently bound glucose in Hb & decreased from 0.14 mol/ a/l dimer to 0.12.

Estimation of the Extent of Glycosylation of Hb Ao-The moles of glucose bound per L$ dimer of normal unincubated Hb A0 were estimated by two independent approaches: in vitro biosynthesis and the TBA test. 1) When Hb Ao was incubated with [‘*C]glucose sufficient to generate the per- centage of Hb Ai, found in normal hemolysate, 12 to 14% of Hb A0 became glycosylated. 2) When analyzed by the TBA test, glucose covalently bound to Hb & generates much less color compared to Hb AI, (Fig. 1 and Table I). The TBA SCA (A/mg of protein) of Hb A0 is about 1.5% that of Hb AI,. By performing the TBA test on Hb &, which has been incubated with [i4C]glucose, we found an average SCA of 0.727 (A/pm01 of glucose) compared with 4.42 (A/pm01 of glucose) in the AI, formed during the same incubation. Thus, the efficiency of color development for the glucosyl-lysine linkages in Hb Ao appears to be 6.1-fold lower than for the glucosyl-valine link- age in Hb Ai,. The SCA for Hb AI, formed during this incubation was very close to that of native Hb AI,. Thus, the TBA profile in Fig. 1 gives a falsely low impression of the amount of glucose covalently bound to Hb Ao. Correcting the SCA of & in normal hemolysate by the same factor, we arrive at an estimation of 9% for the amount of glycosylation in Ao. In light of our evidence that the glucose linkages in AO involve lysines, the differential SCA (A/pm01 of glucose) between AO and Ai, can be further examined by conducting the TBA test on polylysine incubated with [14C]glucose. In this case, we find a SCA of 0.603 (A/pm01 of glucose) on polylysine, about the same as that observed for the incubated &. We conclude that normal Hb & has an average of 0.09 to 0.13 mol of glucose/mol of ofl dimer. Most of the sugar is bound to the e-amino groups of several lysine residues on both (Y and p

chains. The fact that lysino-1-deoxyfructose linkages in Hb & appear to be slightly less stable than the valyl-1-deoxyfiuctose linkage in Hb Ai= tends to favor the lower estimate.

In their structural analysis of Hb Aio Bookchin and Gallop (5) found that following reduction of the purified minor com- ponent with [3H]NaBH4, the a! chain contained about 10% of the radioactivity of the p chain. We have obtained similar results which can be explained by the presence of these additional glucose-lysine adducts.

DISCUSSION

It has generally been assumed that chromatographically pure human hemoglobin A (Hb Ao) is homogeneous. Our studies reveal unexpected structural heterogeneity in Hb AO due to the linkage of glucose to the e-amino group of lysine residues on both (Y and /l subunits as well as to the NHY terminal valine of the (Y chain. The lysino-1-deoxyglucose linkages undergo the Amadori rearrangement to the ketoam- ine, lysino-1-deoxyfructose, identical to the structural transi- tion at the P-NH* terminus in Hb AI, (Fig. 3 and Ref. 6). Furthermore, the adducts formed at these lysine residues are nearly as stable as that at the P-NH2 terminus. The compar- ison of normal versus diabetic red cells (Fig. l), as well as the in uitro incubations with [“Clglucose (Fig. 2), show that the glycosylation of Hb A0 increases in parallel with the formation of Hb Ai,. Gabbay et al. (20) have compared a large number of normal and diabetic individuals and have found that the amount of sugar attached to Hb &, as detected by the TBA test, correlates with the level of Hb AI,. These results indicate that glycosylation of Hb A0 is also a slow, nonenzymatic process reflecting the average intracellular blood glucose level.

Normal human hemolysate contains about 4% Hb AI,. In addition, we estimate that Hb & contains about 0.1 glucose

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from

3898 Structural Heterogeneity of Human Hemoglobin

molecules per a/I dimer. In determining this relationship, two independent approaches were used, both of which are indirect and subject to considerable analytical error. From these esti- mates, we conclude that in normal red cells, the ,&NH2 ter- minus is the site most frequently glycosylated (0.04 mol/cup). The a-NH2 terminus is a somewhat less common site of attack. The relatively low pK, values of these two amino groups make them effective nucleophiles, condensing with the alde- hyde function of glucose to form the initial Schiff base linkage. In contrast, the much lower probability of glycosylation of individual lysine residues is probably due to the relatively high pK, of the e-amino group.

Even though glucose binds covalently to a number of dif- ferent amino groups on the (Y and /3 chains, only one modifi- cation, at the ,8 chain NH2 terminus, results in the formation of a chromatographically distinct minor hemoglobin compo- nent (Hb Ai,). Glycosylation of amino groups, when followed by rearrangement to the ketoamine form, does not remove the charge from these groups, but does lower their pK, values. Chromatography at neutral pH, near the pK, values of the a-amino groups, optimizes the chance of resolving hemoglo- bins glycosylated at the NH2 termini and probably explains the separation of Hb AI, as a discrete peak. By use of a shallow salt gradient, we have been able to enrich hemoglobin glyco- sylated at the (Y chain NH2 terminus to the leading edge of the Hb A0 peak, but have been unable to isolate it as a chroma- tographically distinct minor component. Thus, the change in pK, at this site due to glycosylation is slight. In the case of hemoglobin glycosylated at lysines, even a moderate change in pK, would not result in a significant separation of this component from the main Hb & peak since the chromatog- raphy is performed 4 pH units below the normal pK, of the e-amino groups. Thus, Hb Ai, differs from the other glycosyl- ated components in two respects: it is the most abundant and the glycosylation at this particular site happens to confer a decrease in isoelectric point in a pH range which enables it to be separated from the major component by chromatographic and electrophoretic methods.

These results argue strongly against specificity in the inter- action of glucose with hemoglobin. In contrast, glucose 6- phosphate attaches specifically to the NH2 terminus of the /I chain because its phosphate group serves as an affinity label, leading the molecule to the diphosphoglycerate binding site (21). The relative ubiquity of the glucose-hemoglobin inter- action suggests that other proteins are modified in a similar fashion. It has long been recognized that (y- and e-amino groups on protein can form adducts with glucose during in vitro incubations (22-24). Good candidates for this type of nonenzymatic post-translational modification in uiuo include proteins which have relatively long turnover times and are exposed to the high glucose concentrations found in plasma. Tamer et al. (9) have demonstrated hexose attached to the e-amino of hydroxylysine in bovine collagen while Bailey and his colleagues have found hexose-•-aminolysine both in col-

lagen (25) and in proteins of the normal erythrocyte mem- brane (10). Recently, preliminary results have suggested the presence of glucosyllysine in basic myelin protein of nerve (26) and also in crystallin following the incubation of lens with glucose (11). There is considerable interest in the possibility that this post-translational modification may contribute to the pathogenesis of the long term complications of diabetes.

Acknowledgments-We thank Mr. Michael Austin for expert sec- retarial assistance and Mr. Shesh Gerstein for donating blood samples.

5.

6.

7.

8.

9.

10.

11.

12. 13.

14. 15.

16.

17. 18.

19. 20.

21.

22.

23. 24.

25.

26.

REFERENCES

Bunn, H. F., Gabbay, K. H., and Gallop, P. M. (1978) Science 200,21-27

Rahbar, S. (1968) Clin. Chim. Acta 22,296-298 TriveUi, L. A., Ranney, H. M., and Lai, H. T. (1971) N. Engl. J.

Med. 284,353-357 Gabbay, K. H., Hasty, K., Breslow, J. L., EIIison, R. C., Bunn, H.

F.. and Gallon P. M. (1977) J. Clin. Endocrinol. Metab. 44. 859-864 -

Bookchin, R. M., and Gallop, P. M. (1968) Biochem. Biophys. Res. Commun. 32,86-93

Bunn, H. F., Haney, D. N., Gabbay, K. H., and Gallop, P. M. (1975) Biochem. Biophys. Res. Commun. 67, 103-109

Fluckiger, R., and Winterhalter, K. H. (1976) FEBS Lett. 71,356- 360

Koenig, R. J., Blobstein, S. H., and Cerami, A. (1977) J. Biol. Chem. 252,2992-2997

Tamer, M. L.. Fairweather, R.. and Gallop, P. M. (1972) Arch. Biochem. Biophys. 151, 137-141

Bailey, A. J., Robins, S. P., and Tanner, M. J. A. (1976) Biochim. Biophys. Acta 434,51-57

Stevens, V. J., Rouzer, C. A., Monnier, V. M., and Cerami, A. (1978) Proc. Natl. Acad. Sci. U. S. A. 75. 2918-2922

Drabkin, D. L. (1946) J. Biol. Chem. 164, 703-723 McDonald, M. J., Shapiro, R., Bleichman, M., Solway, J., and

Bunn, H. F. (1978) J. Biol. Chem. 253,2327-2332 Dixon, H. B. F. (1972) Biochem J. 129.203-208 Clegg, J. B., Naughton, M. A., and Weatherall, D. J. (1966) J.

Mol. Biol. 19,91-108 Jensen, M., Oski, F. A., Nathan, D. G., and Bunn, H. F. (1975) J.

Clin. Invest. 55.469-477 Hauschka, P. V. (1977) Anal. Biochem. 80, 212-223 Jones, R. T. (1964) Cold Spring Harbor Symp. @ant. Biol. 29,

297-308 Folk, J. E. (1956) Arch. Biochem. Biophys. 64, 6-18 Gabbay, K. H., Sosenko, J. M., Bunn, H. F., Shapiro, R., and

Gallop, P. M. (1978) Clin. Res. 26,528A Haney, D. N., and Bunn, H. F. (1976) Proc. N&Z. Acud. Sci. U.

S. A. 73,3534-3538 Lea, C. H., and Hannan, R. S. (1950) Biochim. Biophys. Acta 5,

433-454 Schwartz, H. M., and Lea, C. H. (1951) Biochem. J. 50,713-716 AbdeIIa, P. M., Ritchey, J. M., Tam, J. W. O., and KIotz, I. M.

(1977) Biochim. Biophys. Acta 490,462-470 Robins, S. P., and Bailey, A. J. (1974) Biochem. Biophys. Res.

Commun. 48.76-84 Fluckiger, R., and Winterhalter, K. H. (1978) In Biochemical and

Clinical Aspects of Hemoglobin Abnormalities (Caughey, W. S., ed) p. 205, Academic Press, New York

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from

GabbayH F Bunn, R Shapiro, M McManus, L Garrick, M J McDonald, P M Gallop and K H

glycosylation.Structural heterogeneity of human hemoglobin A due to nonenzymatic

1979, 254:3892-3898.J. Biol. Chem. 

  http://www.jbc.org/content/254/10/3892.citation

Access the most updated version of this article at

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/254/10/3892.citation.full.html#ref-list-1

This article cites 0 references, 0 of which can be accessed free at

by guest on April 11, 2018

http://ww

w.jbc.org/

Dow

nloaded from