Int. J. Peptide Protein Res. 9, 1917,241-248 Published by Munksgaard, Copenhagen, Denmark. No part may be reproduced by any process without written permission from the author(s)

SOME OBSERVATIONS ON THE CARBOHYDRATE COMPOSITION O F PURIFIED T R A N S F E R R I N K.-L. WONG & E. REGOECZI

Plasma Protein Research Laboratory, McMaster University, Health Sciences Centre, Hamilton, Ontario, Canada

Received 2 December 1975 Five subfractions were isolated on DEAE-cellulose from samples o f a commercially available human transferrin preparation and their carbohydrate composition was analysed. Hexosamine, galactose and total hexose were determined in four subfractions and sialic acid in all five. The data obtained indicate that the excess number of electrophoretic bands observed in transferrin from this source is due to the loss o f carbohydrates which only affects sialic acid and none of the other sugar types. The availability o f the penultimate galactose residues as the terminal residues in the subfractions deficient in sialic acid was also confirmed by a biological test utilizing the rat liver. The reason for the gradual loss o f sialic acid from transferrin is unknown. Freezingthawing and lyophilization did not detectably affect the sialic acid content of purified transferrin. However, free sialic acid did appear in some preparations on storage. It is concluded that similar changes in the carbohydrate composition o f other plasma glycoproteins before, during, or after purification can be expected to exert an adverse effect on their usefulness as metabolic tracers.

The human transferrin preparation produced by Ekhringwerke (Marburg, Germany) is widely renowned for its purity (Morgan et al., 1967) +nd high metabolic quality (Cromwell, 1964). Transferrin from this source has been used in h i s laboratory for some time as a starting material to obtain asialotransferrin (Regoeczi e l al., 1974; Regoeczi & Hatton, 1974). In the , course of our studies it was noticed that when electrophoresed in alkaline polyacrylamide gels, Qehringwerke transferrin consistently resolved p t o at least four bands which were spaced not unlike incompletely desialylated transferrin 4Parker & Bearn, 1962; Schermann et al., 1971). Accounts by previous workers (Jamieson

1965; Graham & Williams, 1975; Heide & Haupt, 1964) conflict on the number of electrophoretically distinguishable components that can normally be expected in purified human transferrin type CC. Nonetheless, since the largest number reported did not exceed three (one major and two minor components: [Heide & Haupt, 19641 ), the electrophoretic heterogeneity of Behringwerke transferrin appears excessive and warrants investigation. We have therefore devised a chromatographic technique for the isolation of the individual transferrin components, and comparatively determined their carbohydrate composition. The findings from these studies are reported below.

24 1

K.-L. WONG and E. REGOECZI

MATERIAL AND METHODS Transferrin

Protein samples from four shipments of Behringwerke transferrin (series 2469B and 2469E) were studied. To obviate the effect of any unequal distribution of bound iron on the chromatographic behaviour (Lane, 1971) and electrophoretic mobility (Aisen er al., 1966) of transferrin, samples to be analysed were either saturated with iron or, less frequently, fully converted into apotransferrin. Saturation with iron was carried out in a borate buffer system using ferriccitrate as the metal donor under conditions described before (Regoeczi er al., 1 9 7 5 ~ ) .To obtain apotransferrin, transferrin was dissolved in 0.1 M sodium citrate containing 0.1 M EDTA-Na2 and exhaustively dialysed against a 50-fold excess of the same solvent. Chroma tography on D EA E-cellulose

After adjusting their iron content as outlined above, samples (approx. 100 mg) of transferrin were exhaustively dialysed against 0.01 M tris-HCl pH8.0 at 4°C and then absorbed at room temperature on to a column (1.4 cm x 45 cm) of DEAE-cellulose (Whatman DE 52) equilibrated with the same buffer and containing 5% (w/w) Hyflo Supercel. The column was developed with an asymptotic gradient which was established by running the limit solvent (0.2 MTris-HCl pH8.0) into a mixing chamber containing 500 ml of the equilibrating buffer. The flow rate was approx. 20ml/h. The absorbance of the effluent was monitored at 280 nm and the fractions (6 ml) containing the individual protein peaks were pooled, concentrated by ultrafiltration, exhaustively dialysed against 0.02 M ammonium acetate pH 7.0, and freezedried. The degree of separation achieved was monitored by electrophoresis (cf. below) and, if necessary, contaminations with protein from adjacent peaks were eliminated by rechromatography. Carbohydrate analyses

Sialic acid content was determined by the method of Aminoff (1961) after hydrolysing the protein samples in 0.05 M H 2 S 0 4 at 85OC for 1 h. The standard, N-acetyl-neuraminic 242

acid (NANA), was subjected to hydrolysis in the same way as the protein. Total hexose content was determined by the H2SO4 -phenol reaction (Dubois er al., 1956) using D-galactose and D-mannose as a standard at a molar ratio of 1 : 1. Hexosamines were measured by the method of Rondle & Morgan (1955). The protein was hydrolysed at a concentration of 0.2-0.4% in 3.5 M HCl for 3.5h at 105OC. Excess HCI from the hydrolysis was neutralized by 4 M NaOH. The standard was N-acetyl-glucosamine and blanks were hydrolysed protein samples treated in the same way except for boiling after the addition of acetylacetone was omitted. To liberate galactose, transferrin was desialylated and freeze-dried as described pre-. viously (Regoeczi er al., 1974). Approx. 2mg of the asialoprotein were incubated with 4 mg emulsin 0-glucosidase (Sigma; 0-galactosidase specify activity 1520 units/mg/h) in 0.5 ml' of 0.1 M sodium citrate pH 6.0 under an atmosphere of toluene. After 4 8 h at 37OC, 0.5 ml of 0.2 M Tris-HC1 pH 8.0 were added and the precipitate formed was removed by centrifugation. Aliquots of the supernatand were used t o determine the degree of desialyation and the amount of galactose released. To estimate galactose, 0.5ml of the supernatant were incubated for 1 min at 37°C with 0.1 ml of a 0.1 M solution of nicotinamide adenine dinucleotide (Sigma) and 0.3 ml, of 0.2 M Tris-HC1 pH 8.0. Then 50 pg galactose dehydrogenase from Pseudomonas fluorescenl (Sigma; 6.3 unitslmg) were added in 0.1 ml o the above buffer and the absorbance at 34 nm was followed at 15-sec intervals for 2 min. The blank was incubated enzyme alone and standards consisted of incubated mixtures & enzyme and galactose. A calibration curve wah obtained by plotting the initial rate of change in absorbance of standards of known strengths# Transferrin concentrations were determined spectrophotometrically using an A;$ value of 13.0 (Lane, 1971). To relate carbohydrate tq protein, a molecular weight of 76,000 (Palmour & Sutton, 1971) was assumed for transferrin.

6

Measuring the hepatic clearance o f transferrin fractions prepared on DEAE-cellulose

This was done in adult Sprague-Dawley a@

CARBOHYDRATE COMPOSITION OF PURIFIED TRANSFERRIN

Wistar rats using a technique described elsewhere (Regoeczi, 1975). In brief, groups of rats were given a mixed intravenous injection of a transferrin fraction (30-40 pg/lOO g body weight) labelled with 1311 and human serum albumin [Behringwerke; 10-2Opg/lOO g body weight), labelled with "'I, as a trapped plasma marker. Five minutes later the liver was removed and homogenized. Aliquots of the homogenate as well as of the plasma from a terminal blood sample were assayed in duplicate in a Packard model 5986 multichannel analyser. From the results of these measurements the net hepatic uptake of transferrin was calculated using a formula given elsewhere (Regoeczi, 1975). Similar studies were also performed in groups ,of rats in which the hepatic asialoglycoprotein receptor (Hudgin et al., 1974) was blocked by ,injecting chicken al acid glycoprotein (2.9 k 0.4 mg/100g body weight) 1 min prior to the transferrin (Regoeczi et al., 19756).

(Regoeczi et al., 1974) human transferrin (2mg/ml of distilled H 2 0 ) was allowed to freeze at -14OC and then thaw at 37°C five times; freshly prepared human transferrin was freeze-dried from 0.2 M ammonium acetate pH 7.0 in a Virtis (Gardiner, N.Y.) lyophilizer; seven fractions prepared from Behringwerke transferrin by DEAEchromatography and freeze-dried as already outlined were stored at -14'C for a period of 6 months. In addition Behringwerke transferrin as obtained from the manufacturer was also analysed for free sialic acid. Other techniques

Analytical electrophoresis was performed in 7.5% polyacrylamide gels in 0.005M Trisglycine pH 8.3 at 4 m A per gel for 1 h. After staining with amido black and destaining in 7% acetic acid, gels were scanned in a Clifford model 345 recording densitometer. Protein iodinations were performed using The effect of various treatments of transferrin iodine monochloride (McFarlane, 1958) and up e n its sialic acid content to 1.5 atoms of iodine were incorporated per The presence of free sialic acid in transferrin molecule of protein. breparations was tested by applying the technique of Aminoff (1961) either to nonhydroRESULTS lysed or hydrolysed samples before and after dialysis against distilled H20at 4OC for 16h. Separation of the components of'Behringwerke The effects of the following experimental transferrin on DEA E-cellulose conditions were investigated: Freshly prepared The extent of resolution obtained with the

O.*

I t

\

z FIGURE 1 Chromatography of a sample of Behringwerke transferrin, hturatcd with iron, on DEAEcellulose. Fractions (6 ml b c h ) are numbered from the start of the gradient. Molarity of the gradient is indicated /@ the broken line.

2a

0

g Q

0.2;)

I

I

I-

I

En

- 0.05

I

FRACTION NUMBER

243

K.-L.WONG and E. REGOECZI

present technique is illustrated by the chromatogram of 2Fe-transferrin in Fig. 1. Five peaks are apparent and are numbered 0-IV in accordance with the nearest integers calculated for their NANA contents as residues per 76,000 daltons (cf. Discussion). The same number of peaks was observed also when apotransferrin was used as the starting material. The unfractionated protein yielded four bands in polyacrylamide gel electrophoresis and a comparison of the individual chromatographic peaks with this gel pattern showed that each peak eluted from the DEAEcolumn corresponded to one of the electrophoretic components of unfractionated transferrin (Fig. 2). Thus, the peak IV chromatographic material corresponded t o the transferrin band with the highest anodic mobility, peak 111 to the band with the second-highest mobility, and so on. No band could be seen in the electropherogram of whole transferrin corresponding to the peak 0 protein, probably because of the low proportion of this material in unfractionated transferrin (cf. Table 1). Peak 0 transferrin,

pooled from several fractionations, migrated more slowly than the slowest visible band in whole transferrin, but faster than asialotransferrin prepared from whole transferrin. The relative amounts of the individual transferrin components were calculated from the absorbance at 280nm in four chromatograms, and the values obtained are compared with the densitometric evaluation of the starting material (electrophoresed in triplicate) in Table 1. Carbohydrate composition o f the transferrin subfractions

The data are summarized in Table 2. It is seen that the hexose, hexosamine and galactose contents of subfractions I-IV varied withid a relatively narrow range. The significance of, the differences was tested by analysis of the variances, followed by Duncan’s multiple range’ test as modified by Kramer (1956) for groups of unequal size. These calculations indicated that none of the subfractions differed signifa

FIGURE 2 Electropherograms of the transferrin and its chromatographic subfractions from the experiment shown in Fig. 1. From right to left: starting material, peak IV, peak 111, peak 11, peak I and peak 0 transferrins. Running conditions are given in the Methods section.

244

CARBOHYDRATE COMPOSITION OF PURIFIED TRANSFERRIN TABLE 1 Relative amounts o f the components in BehrinKwerke transferrin as derived from chromatograms and electrophoretic scans Technique

Chromatography

peak 0

peak I

1.5 0.2)

3.9 1.8) 9.7 (i 0.2)

(t

17.1 2.0) 25.5 (f 0.7)

48.1 4.7) 40.0 (+ 1.5)

(i

(f

invisi blc

Gel scan

% Transferrin corresponding to peak 11 peak 111 (2

peak IV 29.4 5.4) 24.7 (f 1.1) (f

Values are means with standard deviations in parentheses.

cantly (P> 0.01) from another as far as these three sugar types were concerned. In sharp contrast, the sialic acid contents of the subfractions differed significantly (P< 0.01) and they clearly paralleled the anion binding ,capacity of the chromatographic peaks. Only small quantities of the subfraction 'designated as peak 0 could be collected, thus making a complete carbohydrate analysis of this material impossible. However, the identity uf this protein peak as transferrin was readily verified immunologically and it contained b.39% (w/w) sialic acid. Further properties of peak 0 transferrin could be inferred from its behaviour in the rat liver test.

Hepatic u p t a k e of the human transferrin subfractions in rats

When labelled samples of the different transferrin subfractions were injected in trace quantities into rats, the fraction of the dose recovered with the liver after 5min related inversely to the sialic acid content of the protein. The values, corrected for radioactivity due to trapped plasma in the liver, are listed in Table 3. It is seen that peak 0 transferrin was taken up most rapidly, the portion of this material recovered with the liver being only slightly smaller than the 35.9% calculated from earlier studies (Regoeczi er al., 197%) for desialylated whole transferrin. In contrast, the relative amounts of peak 111 and peak IV proteins taken

TABLE 2 Carbohydrate composition o f the transferrin subfmctions obtained b y DEAE-cellulose chromatography

*Carbohydrate 4YQe &alic acid

I

x S.D. n

Amount (k w/w)in transferrin peaks I1 111

i

0.99 0.13 8

f

2.63 0.29 4

0.58

* 0.16 7

i

IV

1.33 0.18 8

f

1.63 0.08 7

2.31 0.13 5

2

2.41 0.17 5

ifexose

2 S.D. n

2.60

* 0.38 4

+_

Hexosamine

x S.D.

f

1.88 0.27 4

f

2.08 0.26 4

f

2.13 0.22 4

i

2.09 0.23 4

t

0.95 0.03 4

f

0.87 0.08 4

f

0.85 0.05 4

i

0.87 0.10 4

n

Gdactose

x S.D.

n

+ h e s listed for each sugar are the mean (X) and its standard deviation (S.D.), and the number of estimations ( n ) .

245

K.-L. WONG and E. REGOECZI

up by the liver during the same interval were tents of the chromatographic subfractions were 0.46 (peak 0), 1.4 (peak I), 2.4 (peak 11) negligible. To see if the hepatic uptake was mediated 3.2 (peak IIJ) and 3.9 (peak IV). The molecular by exposed galactosyl groups on transferrin, forms of transferrin arising froni the gradual the behaviour of peak 0, I and I1 type trans- loss of sialic acid carry sufficiently different ferrins was studied in rats in which the hepatic charges to be separated chromatographically' asialoglycoprotein receptor had been blocked on a preparative scale. Values obtained for the by the injection of chicken al-acid glyco- relative proportion of the individual compoprotein. As apparent from the data in Table 3, nents are about the same, regardless of whether this measure invariably resulted in substantial they are calculated from polyacrylamide gel reductions in the uptake of transferrin by the scans or from chromatograms, suggesting that' the chromatographic procedure neither initiated, liver. nor augmented the degradation of the oligoFree sialic acid saccharide chains of transferrin. (The slight, No measurable loss of bound sialic acid occursystematic divergence between the two sets red in transferrin samples which were subjected of values in Table 1 probably reflects the to either lyophilization or repeated freezing- negative influence of sialic acid on protein, thawing, and no free sialic acid could be detec- staining with amido black.) Needless to say, ted in the transferrin obtained from the manufrom a generalised point of view, the contents facturer. However, small quantities of free sialic of Table 1 are just an example, since different acid, corresponding to 1-6% of the total sialic degrees of deterioration in the carbohydrate? acid content, became detectable in the chromacomposition of transferrin would give rise to tographic fractions prepared from whole different proportions for the subfractions. transferrin which have been stored for 6 Results of both the carbohydrate a n a l y s a months in the lyophilized state. and the biological assays for exposed galactose TABLE 3 residues support the conclusion that the ob? served change in the carbohydrate composition Uptake in vivo by the rat liver of the transfernin subfractions prepared on DEAE-cellulose of the commercial transferrin preparation is limited to the sialic acids. This is particularly Transferrin No. of Net uptake Treatment evident from the behaviour in vivo of the peak no. animals (% dose) two subfractions having the greatest deficiency ' 0 4 32.6 (* 2.4) in sialic acid (peak 0 and peak I transferrinsh 0 1 4.6 + their uptake by the rat liver linearly approaches I 5 24.8 (t 2.6) that of freshly prepared asialotransferrifi I 5 1.3 (* 0.2) + (Regoeczi, 1975; Regoeczi et al., 19791). & I1 5 8.6 (i 2.0) the uptake of asialoproteins by the hepatocyte I1 5 0.5 (* 0.1) + has been shown to be mediated by exposed. I11 5 3.8 (t 2.6) galactose residues (Ashwell & Morell, 1971 IV 5 1.5 (i 0.5) the enhanced clearance of these transferr 1 Values are means with standard deviations in parenth- subfractions from the circulation by the rat eses. Treatment consisted of an intravenous injection clearly demonstrates the presence of galactose. of chicken a,acid glycoprotein (2.9 i 0.4 mg/lOOg Furthermore, since galactose occupies a penult# body wt.) 1 min prior to the labelled dose. mate position at the non-reducing end of the DISCUSSION oligosaccharide chains of transferrin (Jamieson' From the presented data partial loss of sialic et al., 1971), from the results of the rat livq acid emerges as the most likely explanation tests, the lack of any further alteration in the for the excess number of electrophoretic carbohydrate attachments of transferrin c& components in the commercial transferrin safely be inferred. preparation studied, and also for its chromaThe precise reason for Behringwerke transtographic heterogeneity. Expressed as residues ferrin containing partially desialylated sub-, per 76,000 dalton mol.wt., the NANA con- fractions is unknown. No free sialic acid could ~

~

~

~~~~

P

246

CARBOHYDRATE COMPOSITIOIN OF PURIFIED TRANSFERRIN

be detected in the lyophilized protein as et al., 1974), experimental evidence indicates received from the manufacturer, indicating (Morell et al., 1971) that under similar circumthat the change had taken place before the stances the metabolic performance of many purified protein reached the last stage of pro- other glycoproteins would severely be impaired. duction. The fractionation scheme for this preparation of transferrin involves exposure ACKNOWLEDGEMENTS t o rivanol, ammonium sulphate and Al(OH)3 gel at pH values between 5 and 8 (Heide & These studies were supported by the Medical Research Hauptmann, 1964). In unpublished experi- Council of Canada by grant No. MT-4074. We wish ments we found no detectable effect by to thank Profesor H.G. Schwick and Dr. N. Heimburger rivanol or ammonium sulphate on the sialic of Bchringwerkc AG for providing much helpful acid content of human transferrin but we information relevant to the subject of the present lack experience with the use of AI(OH)3 investigations. at pH 5. As an alternative to partial desialylation REFERENCES of transferrin during purification, it is conAisen, P., Leibman, A. & Reich, H.A. (1966) J. Biol. ceivable that the loss occurs during the interChem. 241,1666-1671 val between blood donation and plasma frac- Aminoff, D. (1961) Biochem. J. 81,384-392 tionation. Thus on one occasion we isolated Ashwell, G. & Morell, A.G. (1971) in Glycoproteins of from aged starting material by another fracBlood Cells and Plasma (Jamieson, G.A. & Greentionation procedure (Regoeczi et af., 1974) Walt, T.J., eds.) pp. 173-189, Lippincott, Philadelphia transferrin which exhibited electrophoretic heterogeneity similar to the one observed Cromwell, S. (1964) Prorides Biol. Fluids 11, 484486 with the Behringwerke preparation. Leukocytes arc known to contain neuraminidase (Gielen Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A. & Smith, F.A. (1956) Anal. Chem. 28, 350et al., 1970; Yehetaf., 1971 ;Gielen & Schaper, 356 1973), fractions of which may leak under ,Gelen, W., Schaper, R. & Pink, H. (1970) Hoppethe metabolic conditions prevailing in stored Seylers %. Physiol. Chem. 351,768-770 blood. Gielen, W. & Schaper, R. (1973) Blur 26, 54-60 The appearance of small quantities of free Graham, I. & Williams, J. (1975) Biochem. J. 145, sialic acid in lyophilized transferrin fractions 26 3 -279 during prolonged storage at -14°C is probably Heide, K. & Haupt, H. (1964) Behringwerk-Mitt. 43,161-194 a parallel, yet unrelated phenomenon, to the observations with the commercial transferrin Hudgin, R.L., Priccr, W.E., Ashwell, G., Stockert, R.J. & Morell, A.G. (1974) J. Biol. Chem. 249, preparation. Whether it represents a low-grade, nonenzymic dissociation of the NANA 6 's6 5536-5543 GAL linkage remains to be shown in the future. Jamieson, G.A. (1965) J. Biol. Chem. 250, 29142920 It is noteworthy in this connection that, as Jamieson, G.A., Jett, M. & Debernardo, S.L. (1971) inferable from studies with acid hydrolysis J. Biol. Chem. 246,3686-3693 of glycoproteins, glycosidic bonds involving Kramer, C.Y. (1956) Biometrics 12,307-310 sialic acid distinguish themselves as being the Lane, R.S. (1971) Biochim. Biophys. Acta 243, intrinsically weakest ones (Marshall & Neu193-202 Marshall, R.D. & Neuberger, A. (1972) in Glycoberger, 1972). proteins (Gottschalk, A. cd.), part A, pp. 351Assuming that the present observations with 359, Elsevier, Amsterdam transferrin are not unique to this protein, the McFarlane, A S . (1958) Nature (Lond.) 182,53 tentative conclusion from our studies is that the carbohydrate structure of glycoproteins, Morell, A.G., Gregoriadis, C., Scheinberg, H., Hickman J. & Ashwell, G. (1971) J . Biol. Chem. 246, prepared from aged starting material or allowed 1461-1467 t o age after purification, should not be taken Morgan, E.H., Masaglia, G. Giblett, E.R. & Finch, for granted. Although loss of sialic acid has C.A. (1967)J. Lab. Clin. Med. 69,370-381 been shown to affect the catabolic rate of trans- Palmour, R.M. & Sutton, H.E. (1971) Biochemistry iferrin in man to a slight degree only (Regoeczi 10,4026--4032 247

K.-L.WONG and E. REGOECZl Parker, W.C. & Bearn, A.G. (1962) J. Exp. Med. 115, 83-105 Regoeczi, E. (1975)J. Nucl. Biol. Med. 19,149-154 Regoeczi, E. & Hatton, M.W.C. (1974) Can. J. Biochem. 52,645-651 Regoeczi, E., Hatton, M.W.C. & Wong, K.-L. (1974) Can. J. Biochem. 52,155-161 Regoeczi, E., Wong, K.-L. & Hatton, M.W.C. (197%) Can. J. Biochem. 53,1070-1077 Regoeczi, E., Hatton, M.W.C. & Charlwood, P.A. ( 1 9 7 9 1 )Nature fL.nnd.) 254,699-701

248

Rondle, C.M.J., & Morgan, W.T.J. (1955) Biochem. J. 61,586-589 Scharmann, W., Bruckler, J. & Blobel, H. (1971) Biochim. Biophys. Acta 229,136-142 Yeh, A.K., Tulsiani, D.R.P. & Carubelli, R. (1971) J. Lab. Clin. Med. 78,771 -778 Dr. E. Regoeczi Department of Pathology McMaster University Hamilton, Ontario L8S 4J9 Canada