Increased Glycosylation of Serum Albuminin Diabetes MellitusR. DOLHOFER AND 0. H. WIELAND

SUMMARYThe level of glycosylated albumin has been deter-mined in the serum of normal and diabetic subjectsafter purification of the albumin to apparent homoge-neity. The sugar was released from the albumin prep-arations as 5-hydroxymethylfurfural (HMF) after 24 hof hydrolysis in 2 N acetic acid at 92C, and it was as-sayed by the thiobarbituric acid reaction. The meanvalue for glycosylated albumin, expressed as pico-moles of HMF per nanomole of albumin, obtained from10 normal control and 65 diabetic subjects, was 64and 124, respectively. The level of glycosylated albu-min correlates with the mean blood glucose concen-tration (n = 55, r = 0.715), but not with the fasting bloodsugar concentrations. Moreover, a linear relationshipwas observed between the amounts of glycosylatedhemoglobin (HbA,a_c) and glycosyl-albumin (n = 74,r - 0.88). In an insulin-treated diabetic patient, therewas a different temporal relationship between bloodglucose concentrations and glycosylated hemoglobinand albumin levels. While HbA,a_c was lowered by only15% after 20 days, glycosylated albumin had droppedby more than 50% during the same time. Our resultsindicate that glycosylated albumin might provide avaluable tool to assess the average blood sugarlevels between shorter intervals, since the turnover ofserum albumin is considerably faster than that ofHbAla_c, DIABETES 29:417-422, June 1980.

The development of the late complications of diabe-tes is widely thought to depend on careful meta-bolic control. Therefore, methods that are not sub-ject to rapid fluctuations of the blood glucose level,

but rather reflect the mean blood glucose level prevailingover a longer period of time, are of clinical significance. The

From ihe Klinisch-chemisches Institut and Forschergruppe Diabetes, Stadt-isches Krankenhaus Munchen-Schwabing, Kolner Platz 1, D-8000 Munchen40. FRGAddress reprint requests to Prof. O. H. Wieland.Received for publication 15 October 1979.

glycosylated hemoglobins (HbA,a_c) have now become awidely used tool for assessing the long-term control of gly-cemia in diabetic patients.1 However, because of the longsurvival time of hemoglobin, corresponding to the life spanof the erythrocyte, HbAla_c levels may remain elevated formany weeks after successful therapeutic control of bloodsugar has been achieved. Searching for another more flexi-ble parameter which might permit an evaluation of the de-gree of blood sugar control within shorter periods of time,we have found, in agreement with another report,2 that, anal-ogous to hemoglobin, serum albumin is also glycosylatednonenzymatically.3

These and other studies from this laboratory4 also pro-vided first evidence that the amount of glycosyl-albumin isincreased in diabetic patients. It was the aim of the presentwork to substantiate our earlier finding and to establish ifglycosyl-albumin levels are a function of blood glucoseconcentration. Such a relationship would suggest the possi-bility of using glycosyl-albumin as a tool for the semi-long-term control of diabetes, since the half-life of human serumalbumin is much shorter (about 20 days5) than that ofHbAla_c.

MATERIALSHuman serum albumin, pblyspecific and monospecific anti-sera, and M-Partigen immunoplates were from Behring-werke, Marburg. Affi-Gel Blue was from Bio-Rad, Munich.Whatman DE 52 cellulose was obtained from Hormuth &Vetter, Wiesloch. 2-Thiobarbituric acid, 5-hydroxymethyl-2-furancarboxaldehyde, D(+) glucose, and all other chemi-cals were from Merck, Darmstadt.

METHODSSerum or plasma was obtained from 65 diabetic patients(ranging between 13 and 90 yr of age) and from 10 normalhealthy persons (21-48 yr old).

Albumin was purified by DEAE-cellulose and Affi-GelBlue chromatography as described,3 except that the DEAE-cellulose was washed with 0.045 M K+-phosphate buffer,pH 8.0, until A280 was near zero before the albumin was

DIABETES. VOL. 29, JUNE 1980 417

GLYCOSYLATED ALBUMIN IN DIABETES

0.2-1

FIGURE 1. Polyacrylamide gel electrophoresis of human albumin.Albumin from a diabetic's serum was purified as described in theMATERIALS AND METHODS section, before 30 M 9 o f protein in 0.01 ml0.9% NaCI was applied on the gel. Electrophoresis according to thesystem 6 was performed as described by Maurer.'8 The gel wasstained with amido black 10 B.

eluted with 0.1 M K+-phosphate buffer, pH 8.0. For the assayof glycosyl-albumin, 0.4 ml of albumin solution (correspond-ing to 4 mg protein) was mixed with 0.06 ml glacial aceticacid6 and was incubated for 24 h at a constant temperatureof 92C on an Eppendorf heating block (thermostat model3401). Earlier studies had shown that the recovery of HMF isgreatly increased at longer times of hydrolysis.4 After cool-ing for about 5 min in an icebath, 0.1 ml of 3 M trichloroace-tic acid (TCA) was added, and the sample was centrifugedfor 5 min (Eppendorf model 5412). To 0.4 ml of the superna-

LU

LL

X

o.i-

70 90 110Temperature (C)

FIGURE 2. Formation of 5-hydroxymethylfurfural (HMF) as a functionof temperature during hydrolysis. Commercial human albumin(10 mg/ml) was hydrolyzed in 2 N acetic acid for 24 h at the temper-atures indicated before the thiobarbituric acid assay for HMFwas performed as described in the MATERIALS AND METHODSsection.

tant, 0.2 ml of 0.05 M thiobarbituric acid (TBA) was added,the mixture was incubated for 30 min at 40C, and the ex-tinctionat 443 nm was measured. The concentration of 5-hy-droxymethylfurfural (HMF) was calculated using a molar ex-tinction coefficient of 4 x 104, as given by Keeney andBassette7 and confirmed in our laboratory. Immunoelectro-phoresis was carried out according to Grabar and Wil-

FIGURE 3. Formation of 5-hydroxymethylfurfural (HMF) as a functionof protein concentration. Purified albumin from diabetic's serum washydrolyzed at the protein concentrations indicated and analyzedfor HMF-formation as described in the MATERIALS AND METHODSsection.

CO

LU

E 0.35-

i

LL

5 15 25Protein (mg/ml)

418 DIABETES, VOL. 29, JUNE 1980

R. DOLHOFER AND 0. H. WIELAND

0.5n

400 420 440 460 480 500Wavelength (nm)

520 540

FIGURE 4. Absorption spectrum of (A)5-hydroxymethylfurfural(HMF)-thiobarbituricacid (TBA) adduct, prepared as described inthe MATERIALS AND METHODS section, afterhydrolysis of purified human albumin for 24 hin 2 N acetic acid; ( ) reaction product of anHMF standard with TBA; (O) TBA, 50 mM;and (A) HMF standard, 25 /*M.

liams.8 Quantification of albumin and c^-antichymotrypsin inpurified albumin preparations was performed with M-Parti-gen plates according to the supplier's instruction.

Blood sugar was determined by the glucose-dehydrogen-ase UV method9 (Merck, Darmstadt) and cholesterol and tri-glycerides enzymatically with test kits from Calbiochem(Lahn)10 and Boehringer (Mannheim),11 respectively. Hemo-globins A,a_c (HbA,a_c) were measured by the microcolumnmethod (Quick Sep, Panchem, Kleinwallstadt) according tothe supplier's instruction. Protein was determined by thebiuret method.12 Serum albumin concentration was calcu-lated from the total protein content of serum and the per-centage of albumin obtained after electrophoresis of theserum on cellulose acetate. Statistical evaluation was per-formed according to the instructions in reference 13.

RESULTSThe determination of glycosyl-albumin in normal and dia-betic sera depends on the fact that glycosyl-albumin doesnot behave differently from albumin during the purificationprocedures employed. Our studies with commercial humanalbumin showed that glycosylation in vitro with [14C]-glu-cose affected neither the precipitation by (NH4)2SO4 nor theelution patterns from DEAE-cellulose or Affi-Gel Blue, re-spectively (results not shown).

The purification procedure from both normal and diabeticsera yielded highly purified albumin preparations. Electro-phoresis on polyacrylamide gels showed a single proteinband (Figure 1). Analysis by immunoelectrophoresis dis-played only trace amounts of contaminating protein. Usingmonospecific antisera, this was identified and quantifiedto be arantichymotrypsin, amounting to 0.68 0.07%(x SEM, n = 73). Previous work has shown that glucose is

incorporated in vitro nonenzymatically into albumin in aform that yields HMF after acid hydrolysis.3 The conditionsof the formation of this TBA-reactive material have beenstudied now in more detail. The HMF-formation during 24 hproved to be extremely dependent on the incubation tem-perature, as shown on Figure 2. We have chosen a tempera-ture of 92C (precision < 0.5C, provided by our Eppendorfequipment), because, under these conditions, all albuminsamplesirrespective of the degree of glycosylationcould be assayed at the same protein concentrations, theabsorption of the TBA reaction product remaining within thelinear range of the standard curve. It should be noted in thiscontext that the formation of HMF depends also on the dura-tion of hydrolysis4 and on the nature of the acid used.6 Thusthe HMF determined by the TBA assay represents a relativerather than an absolute value of protein glycosylation,depending on the experimental conditions.

As illustrated in Figure 3 the amount of HMF was a func-tion of the albumin concentration employed, being linear upto 20 mg/ml. Thus, our routine assays at 10 mg/ml proteinwere well within the linear range. Figure 4 illustrates that theabsorption spectrum of the TBA-reactive material released

TABLE 1Levels of glycosyl-albumin (expressed as picomoles ot HMF pernanomole of albumin) in normal and diabetic subjects

Normal controlsDiabetic subjects

n

1065

pmol HMFnmol Alb(x S.D.)

64 4.6124 37

Range

56-7261-255

DIABETES, VOL. 29, JUNE 1980 419

GLYCOSYLATED ALBUMIN IN DIABETES

260-

c

c1LL

O

Q.

130-

r =0.715y=0.Ux*l9.3

MBG (mmol/l)22.2

FIGURE 5. Glycosyl-albumin levels in diabeticsubjects. MGB is the mean blood glucoselevel of the preceding 4 or 5 days. For furtherdetails, see text.

from albumin was nearly identical to that obtained from anHMF standard.

Our previous finding,3'4 that the level of glycosyl-albuminis increased in the serum from diabetics, is substantiated bythe data recorded in table 1. The amount of glycosyl-albu-min, expressed as picomoles of HMF per nanomole of albu-min, ranged from 56-71 and 61 -255 in normal and diabeticsera, respectively. The highest degree of albumin glycosyl-ation was displayed by the patients suffering from diabeticcoma (x SD = 187 8 pmol HMF per nanomole of albu-min, n = 5). The relationship between glycosyl-albuminand blood glucose within the diabetic group is illustrated inFigure 5. The mean blood glucose (MBG) level, shown onthe abscissa, was calculated from all blood sugar determi-nations (usually taken at 0530 h, 0900 h, 1300 h, 1600 h, and2300 h) within 4 or 5 days before blood was withdrawn for al-bumin analysis. Statistical analysis of the data from our dia-betic patients (excluding those suffering from diabeticcoma) revealed a clear correlation (r = 0.715, n = 55) be-tween the mean blood glucose level and albumin glycosyla-tion. This finding is comparable to a similar interrelationshipalready demonstrated between the degree of glycemia andthe level of HbA,a_c.

14'15 It was therefore of interest to investi-gate whether increased albumin glycosylation is associatedwith elevated levels of HbA,a_c. The analysis of blood from74 subjects, with respect to both parameters, establishedthe regression line (r = 0.88) in Figure 6. In contrast with ourresults on the basis of MBG, no such correlation was de-monstrable between glycosyl-albumin and the fasting bloodsugar (r = 0.3, n = 28). However, nearly the same result aswith MBG was obtained when the amount of glycosyl-albu-min was correlated (r = 0.733, n = 57) with the highest sin-gle value of blood glucose found in the preceding interval of4 or 5 days. The level of glycosyl-albumin was apparentlyindependent of the age and body weight of the patients, theduration and therapy of their diabetes, and the blood levelsof cholesterol and triglycerides. Moreover, glycosyl-albu-min and total serum albumin contents were not interrelated.

Because of the relatively short half-life of serum albumin,the determination of glycosyl-albumin could be expected toprovide a more flexible means of diabetes control than

could HbA,a_c. This view is supported by the data sum-marized in Figure 7, obtained from a patient whose HbA,a_cand glycosyl-albumin levels were followed in parallel dur-ing insulin therapy for the indicated period of time. WhileHbAla_c was decreased only slightly (from 20.7 to 17.5%)within 20 days of therapy, a marked lowering of glycosyl-al-bumin (from 208 to 91 pmol HMF per nanomole of albumin)occurred within the same time period.

DISCUSSIONThe measurement of glycosyl-albumin as a tool to assessthe semi-long-term control of diabetes appears reason-able, since albumin can be purified relatively easily from

FIGURE 6. Correlation between the levels of HbAla_c andglycosyl-albumin in 10 normal and 64 diabetic subjects. Thepurification of albumin and the measurement of HbA,,_c wereperformed as described in the MATERIALS AND METHODS section.

260

o 130-

LL

X

"oCL

Hb A10

Ia-c (%)20

420 DIABETES, VOL. 29, JUNE 1980

R. DOLHOFER AND 0. H. WIELAND

IS)oo

O

-210 r21

E

O

C

O

EQ.

0 ' 1 I 2 I 3 I A I 5 I 6 I 7 I 8 I 9 I 10 I 11 I 12 I 13 I 14 I 15 I 16 I 17 I 18 I 19 I 201 ""0 L Duration of therapy (days)

FIGURE 7. Temporal relationship between the levels of blood glucose ( ) , HbAla_c (O), and glycosyl-albumin (A) in a diabetic patient duringinsulin therapy. For experimental details, see the MATERIALS AND METHODS section.

serum, and the TBA-reaction provides a simple and rapidmeans for the determination of HMF derived from glycosyl-albumin. As we have reported elsewhere, HMF can also bedetermined by high performance liquid chromatographyusing electrophoretically purified albumin samples from100 fx\ serum or less.4 Perhaps most important in this con-nection, however, is the fact that the circulating half-life ofhuman serum albumin is about 20 days.5 Although it has notbeen specified so far whether this rate of turnover is charac-teristic for both the glycosylated and the nonglycosylatedforms of human albumin, it is worthwhile to note that, in anycase, it is much faster than that of HbA|a_c. (For rat serumalbumin, which has a half-life of 2 days, no difference inturnover was found between glycosylated and nonglycosyl-ated forms.16)

As to the sensitivity of albumin to monitor an increase inblood sugar, Figure 5 clearly shows that the level of glyco-syl-albumin is already significantly increased at bloodsugar concentrations of 8.3-11.1 mmol/l. Moreover, it isnoteworthy that the level of glycosyl-albumin is a linearfunction of the mean blood glucose, at least up to 16.6mmol/l. From this fact, and the data shown in Figure 7, it ap-pears that the measurement of glycosyl-albumin may be-come a useful diagnostic tool capable of monitoring thestate of glycemia during shorter yet still long enoughperiods of therapy. After submission of this paper, we be-came aware of the most recent results by McFarland et al.17

on glucosylation of total serum protein. Apart from the factthat, in agreement with our earlier findings,3 albumin was re-ported to represent the most abundant but not the only gly-cosylated serum constituent, these authors found that thedegree of serum protein glycosylation is significantly corre-lated to the level of HbA,a_c and also to the fasting bloodsugar concentration. While our results agree with respect to

the correlation with HbA,a_c, we could not demonstrate asignificant correlation between glycosyl-albumin and fast-ing blood glucose levels. The reason for this discrepancy isnot known.

ACKNOWLEDGMENTSThis work was supported by the Deutsche Forschungsge-meinschaft, Bad Godesberg. The skillful technical assist-ance of Miss M. Schotsch and A. Militz is gratefully acknowl-edged. We also thank Professor H. Mehnert and his staff ofthe Third Medical Department of the Schwabing City Hospi-tal for their cooperation in the studies of diabetic patients.

REFERENCES1 Bunn, H. F., Gabbay, K. H., and Gallop, P. M.: The glycosylation of

hemoglobin: relevance to diabetes mellitus. Science 200:21-27, 1978.2 Day, J. F., Thorpe, S. R., and Baynes, J. W.: Non-enzymatically gluco-

sylated albumin. J. Biol. Chem. 254:595-97, 1979.3 Dolhofer, R., and Wieland, O. H.: Glycosylation of serum albumin:

elevated glycosyl-albumin in diabetic patients. FEBS Lett. 703:282-86, 1979.4 Wieland, O. K, Dolhofer, R., and Schleicher, E.: Glycosylation of he-

moglobin and of serum albumin in diabetes. 10th Congr. Int. Diabetes Fed.Vienna 1979, in press.

5 Schultze, H. E., and Heremans, J. F.: Molecular Biology of HumanProteins, Vol. 1. Amsterdam, New York; Elsevier/North-Holland, 1966, p. 475.

8 Gottschalk, A.: Some biochemically relevant properties of N-substi-tuted fructosamines derived from amino-acids and N-arylglucosylamines.Biochem. Z. 42:455-60, 1952.

7 Keeney, M., and Bassette, R.: Detection of intermediate compoundsin the early stages of browning reaction in milk products. J. Dairy Sci. 42:945-60, 1959.

8 Grabar, P., and Williams, C. A.: Method permettant l'6tude conjuge"edes proprie'te's e"lectrophor6tiques et immunochimiques d'un melange deproteines Application du s6rum sanguin. Biochim. Biophys. Acta /0:193-941953.

9 Dolhofer, R., Weiss, L, and Wieland, O. H.: Erfahrungen mit der Glu-cose-Dehydrogenase-UV-Methode zur Bestimmung der Blutglucose. J. Clin.Chem. Clin. Biochem. 74:415-17, 1976.

DIABETES, VOL. 29, JUNE 1980 421

GLYCOSYLATED ALBUMIN IN DIABETES

10 Allain, C. C, Poon, L, Chan, S. G., Richmond, W., and Fu, P.: Enzy-matic determination of total cholesterol. Clin. Chem. 20:470-75, 1974.

" Bucolo, G.. and David, H: Quantitative determination of serum tri-glycerides by the use of enzymes. Clin. Chem. 79:476-82, 1973.

12 Beisenherz, G., Boltze, H. J., Bucher, T., Czok, R., Garbade, K. H.,Meyer-Arendt, E., and Pfleiderer, G.: Diphosphofructose-Aldolase, Phos-phoglyceraldehyd-Dehydrogenase, Milchsaure-Dehydrogenase und Pyru-vat-Kinase aus Kaninchenmuskulatur in einem Arbeitsgang. Z. Naturforsch.8b:555-77, 1953.

13 Sachs, L.: Angewandte Statistik. Berlin, Heidelberg, New York:Springer-Verlag, 1974, pp. 298-344.

14 Koenig, R. J., Peterson, C. M., Jones, R. L, Saudeck, C, Lehrman,

M., and Cerami, A.: Correlation of glucose regulation and hemoglobin A,c indiabetes mellitus. N. Engl. J. Med. 295:417-20, 1976.

15 Dolhofer, R., Stadele, A., and Wieland, 0. H.: Clinical and biochemi-cal studies on the significance and formation of hemoglobins A,c and A,a+b indiabetes mellitus. Klin. Wochenschr. 55:945-54, 1977.

16 Day, J. F., Thornburg, R. W., Thorpe, S. R., and Baynes, J. W.: Non-enzymatic glucosylation of rat albumin. J. Biol. Chem. 254:9394-9400, 1979.

17 McFarland, K. F., Catalano, E. W., Day, J. F., Thorpe, S. R., andBaynes, J. W.: Nonenzymatic glucosylation of-serum proteins in diabetesmellitus. Diabetes 28:1011-14, 1979.

18 Maurer, H. R.: Disk-Elektrophorese. Berlin, Walter de Gruyter & Co.,1968, pp. 42-43.

422 DIABETES, VOL. 29, JUNE 1980