Role of sugar chains in the expression of the biological activity of human erythropoietin.

Various deglycosylated derivatives of recombinant human erythropoietin (hEPO) were prepared and used to determine the role of the sugar chains in the expression of its biological activity in vivo and in vitro. Three N-linked oligosaccharides of hEPO have been partially or fully removed to obtain N-glycan (NG) (2)-, NG(1)-, and NG(0)-hEPO carrying two, one, and no N-linked sugar chains, respectively. The preparation lacking only O-linked sugar chain O O-glycan (OG) (0)-hEPO was also used. As de-N-glycosylation proceeded, the in vivo activity of the hormone decreased drastically, and the activity of these derivatives was correlated with the number of sialic acids bound to them. On the contrary, the in vitro activity was increased by the de-N-glycosylation; NG(0)-hEPO showed a 3-fold higher specific activity than the intact hormone. This was confirmed by binding experiments of the derivatives to target cells. The in vitro activity and the affinity also correlated with the number of sialic acids bound to the deglycosylated hEPO preparations. On the other hand, OG(0)-hEPO was as active as the intact hormone in vivo and in vitro. In conclusion, the N-linked sugar chains are not required for in vitro activity but required for in vivo activity, acting as anchors for the essential terminal sialic acids. The O-linked sugar chain has no essential role in the biological activity of the hormone in vivo or in vitro.

that the major carbohydrate chain of the hormone is sialylated tetraantennary oligosaccharides with or without N-acetyllactosaminyl repeats. The importance of the terminal sialic acid residues for the expression of the biological activity in vivo and i n vitro has been reported using crude ; the sialic acids are essential i n vivo but not required i n uitro. The complete loss of hEPO activity by desialylation i n vivo has been confirmed to be a result of hepatic removal of the asialo-hEPO from circulation (13,14).
We have demonstrated in a previous paper (15), using highly purified recombinant hEPO, that the asialo-hEPO shows a 3-6-fold higher specific activity than the intact hormone i n vitro and that the increased activity is caused by the increase of affinity between asialo-EPO and EPO receptors (15). We have also shown, by using various partially or fully desialylated hEPO, that the i n vivo activity of the hormone fully depends on the number of sialic acid residues.
Recently, Tsuda et al. (16) and Takeuchi et al. (17) have reported the biological activity of hEPO in which three Nlinked sugar chains had been fully removed. They have shown that the i n vivo activity of fully de-N-glycosylated hEPO is completely lost. However, the significance of the sugar chains in the expression of i n vitro activity remains confused Tsuda et al. showed that the i n vitro activity remained after removal of N-linked or total sugar chains, but Takeuchi et al. showed that the activity i n vitro is completely abolished. This contradiction may be derived from not having isolated the deglycosylated derivatives and not fully characterizing them physicochemically. Before evaluation of the biological activity, it is essential to confirm that the deglycosylated derivatives have not been damaged in the polypeptide moiety, as determined by physicochemical characterization. In this paper, we describe the contribution of N-linked and 0-linked oligosaccharide chains to the biological activity of hEPO i n vivo and i n vitro using the deglycosylated preparations fully characterized physicochemically. We also discuss the site-specific N-glycosylation of hEPO analyzed by sugar mapping.

DISCUSSION
We have successfully clarified the role of the N-linked and 0-linked sugar chains in the biological activity of hEPO using various deglycosylated forms of hEPO. The preparations used were NG(2)-, NG(1)-and NG(0)-hEPO carrying two, one, and no N-linked sugar chains, respectively (Fig. 1). We also used the intact hEPO as a control, OG(0)-hEPO (lacking only * Portions of this paper (including "Experimental Procedures," Figs. 1-8, and Tables 1-111) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. 0-linked sugar chain but containing three N-linked sugar chains), and the fully deglycosylated hEPO. Physicochemical analyses demonstrate that undesirable degradation did not occur during preparation of the sample and Tables   I and 11). In conclusion, N-linked sugar chains are essential for the expression of the biological activity of hEPO in vivo, but their role is restricted to a role as anchors for the terminal sialic acids. On the other hand, the N-linked sugar chains are not required for the in vitro activity, and the 0-linked sugar chain does not have affect any on either activity. The significance of sialic acids bound to the 0-linked sugar chain is thought to be different from that bound to the N-linked sugar chains.
In Vivo Activity-The importance of the N-linked sugar chains in the in vivo activity of hEPO has been confirmed; as de-N-glycosylation proceeded, the in vivo activity of the hormone decreased drastically (Table IV). Results obtained here indicate that all the N-linked sugar chains contribute almost equally to the in vivo activity, and that there are no specifically important N-linked sugar chain. A linear relationship between the logarithm of in uiuo activities of partially and fully de-N-glycosylated hEPO and the number of sialic acids bound to them was obtained. If a specific N-linked sugar chain contributed to the activity, either NG(2)or NG(l)-hEPO would not have been shown an in vivo activity correlated with the number of sialic acids. We have also clarified the role of the 0-linked sugar chain of the hormone in vivo using OG(0)-hEPO. This showed almost the same specific activity as the intact hEPO, indicating that the 0-linked sugar chain has no effect in the expression of the in vivo activity. However, this lack of contribution of the 0-linked sugar chain was assessed in the presence of the N-linked sugar chains. The presence of the N-linked sugar chains may mask the function of the 0-linked sugar chain.
Then, we compared the activities of NG(0)-hEPO, which carries one 0-linked sugar chain but no N-linked sugar chains, asialo-NG(0)-, and fully deglycosylated hEPO. NG(0)-, asialo-NG(0)-, and fully deglycosylated hEPO showed in vivo activities of 800, 600, and 300 units/mg, respectively, although these values were quite low (0.14-0.38%) when compared with that of the intact hEPO (2.13 X lo5 units/mg). From these results, in terms of the role of sugar chains in the in vivo activity, it is concluded that the three N-linked sugar chains of hEPO are important, being required as anchors for the essential terminal sialic acids, but the 0-linked sugar chain does not contribute. The 0-linked sugar chain carries about 1.5 mol of sialic acid. However, the biological significance of sialic acids bound to the 0-linked sugar chain is thought to be different from those bound to the N-linked sugar chains because NG(0)-and asialo-NG(0)-hEPO showed almost the same specific activity.
The loss of the in vivo activity of asialo-hEPO is explained by the hepatic removal from circulation; however, a portion of the derivative may escape from capture via the lectin in the liver to still reach bone marrow. It is widely believed that asialo-hEPO and de-N-glycosylated hEPO completely lose their in vivo activity. Even fully desialylated hEPO showed an in vivo activity, which, however, is only 0.1-0.2% that of the intact hEPO. The question is how partially or fully de-Nglycosylated hEPO are cleared from circulation. We can not explain this phenomena satisfactory at present, but we have some data that the disposition of de-N-glycosylated hEPO is different from asialo-hEPO. It has been found that the de-Nglycosylated hEPO are uptaken and degradable mainly in the kidney but not in the liver.3 The half-life of NG(2)-hEPO is almost the same as the intact hEPO, but further de-Nglycosylation caused shorter half-lives.
I n Vitro Activity-The role of the N-linked sugar chains of hEPO in vitro is quite different from that of in vivo. Results shown in Figs. 6 and 7 and Table IV demonstrate that the derivatives have a higher specific in vitro activity than the intact hormone. In disagreement with these results, however, Takeuchi et al. (17) have reported that N-glycanase digestion of the intact hEPO resulted in almost a complete loss of the activity in vitro. They have concluded that the core portion of the Asn-type sugar chain is necessary for erythropoietin to have its full biological activity in vitro and suggest that removal of the core portion of the sugar chains destroys the active conformation of erythropoietin. Their results are quite opposite to those that we have obtained. We used the hEPO derivatives after they had been fully characterized. Furthermore, de-N-glycosylated hEPO showed almost the same CD spectra, indicating that full de-N-glycosylation does not affect to the structure of hEPO even though up to about 40% total mass of the hormone had been removed (Fig. 4). The binding ability of the fully deglycosylated hEPO to human erythroleukemic cells increased. Therefore, it is reasonable to think that the N-linked sugar chains of hEPO are not required for the in vitro activity.
Binding activity of the deglycosylated hEPO to FMLC was evaluated by competition for the iodolabeled intact hEPO using the AccuFit Competition program. Affinity between the deglycosylated hEPO and the specific receptors for hEPO increased by de-N-glycosylation, but the number of binding sites was unchanged. The ratio of the Kd values and the halfmaximal doses for CFU-E colony-formation of de-N-glycosylated hEPO to the intact hEPO showed a good accordance (Table IV and Fig. 8). OG(0)-hEPO showed almost the same Kd value as the intact hormone. These results demonstrated that the higher specific activities of the de-N-glycosylated hEPO depend on the affinity between it and the receptors. Why then is the affinity between hEPO and its receptors increased by deglycosylation? One possible explanation is that the negative charges of the intact hormone carried by the H. Kinoshita, N. Ochi, N. Oh-ishi, M. Kato, S. Tokura, and A.
Okazaki, manuscript in preparation.

Role of Sugar
Chains in Human Erythropoietin 7705 sialic acids inhibit the interaction between the hormone and the receptors. When the affinity between the FMLC and lz5I-hEPO or '251-asialo-hEP0 was determined by Scatchard analysis, the affinity was increased but the binding sites per cell were not changed by desialylation. Furthermore, the intact hEPO showed almost the same affinity to the neuraminidase-treated FMLC as asialo-hEPO. The intact hEPO has a high negative charge (pZ = 3.0-4.2) but the deglycosylated hEPO does not. Therefore, the explanation that the negative charges of the sialic acids inhibit the interaction between the hormone and its receptors can be applicable to the deglycosylated hEPO. Site-specific N-Glycosylation-hEPO carries three N-linked sugar chains (N24, N38, and N83) attached at Asn-24, -38, and -83. The Asn-83 sugar chain had the highest susceptibility to glycopeptidase F out of the three sugar chains because NG(2)-hEPO lacks only the Asn-83 sugar chain but has the Asn-24 and -38 sugar chains intact. The Asn-24 and -38 sugar chains have a similar susceptibility to the enzyme because NG(1)-hEPO has been shown to consist of an equal amount of the derivatives that carry the Asn-24 or -38 sugar chains. This may be explained by the two glycosylation sites being closely located to one another.
Recombinant hEPO has been reported to carry over 10 different kinds of N-linked sugar chains as does human urinary EPO (6-8). We have analyzed site-specific N-glycosylation to determine the structure of certain sugar chains to see if these certain sugar chains are essential for the expression of the biological activity of hEPO in vivo. We have previously reported site-specific glycosylation of the hormone using fast atom bombardment mass spectrometry and reported that the sugar chains attached to Asn-83 are composed mainly of tetraantennary without N-acetyllactosaminyl repeats (32). However, we now demonstrate that the Asn-83 sugar chain is as heterogeneous as the Asn-38 sugar chain by the sugar mapping employed here (See Fig. 5 and Table 111). The difference is derived from the difference of the peptide samples analyzed. The glycopeptide K4, containing the Asn-83 sugar chain, shows multiple peaks close together on the peptide map (see Fig. 3a). This is caused by the heterogeneity of the sugar chains. In our previous study, we used the major peak of K4, but have used all peaks in the present study. The N-linked sugar chains were obtained by enzymatic digestion of the hormone using glycopeptidase F, not by chemical cleavage of hydrazinolysis. An advantage of the enzymatic method is that the degree of release of the sugar chain can be easily monitored by sodium dodecyl sulfate-polyacrylamide gel electrophoresis since the protein moiety remains intact. The other advantage of sugar mapping using reverse-phase HPLC and ion suppression amine adsorption-HPLC is that once the structure of each peak has been identified, we can estimate the structure of the sugar chains with a high probability by simple HPLC.

27.
28. The column was developed with the same buffer at a flow rate of hEPO wcrc separately poalcd, and were applied to a Vydac protein C4 reverse-phase 3.75 mllmin. The rcsulting three fractions corresponding to NG(2)-. NG(I)-. and NG(0)column (4.6 x 250 mm) equilibrated with 30% acetonitrile in 0.1% trifluoroacetic acid. The dcglycosylated derivatives were eluted with a linear gradient of acelonilrile from prepared from NG(0)-hEPO by the Successive digestion with neuraminidase (Seikagaku 30% to 80% over 20 min at a flow rate of 1.0 mllmin.

Role of Sugar C h i n s in Human Erythropoietin
Fully deglycosylalcd hEPO was kogyo Co. Tokyo) and endo-a-N-aectylgalactosaminidasc (Boehringer Mannhcim) and purified by reverse-phase HPLC as described above. hEPO lacking the 0-linked sugar chain but containing three N-linked sugar chains was purified accordmg to the method described previously from a medium conditioned by transfected CHO cells wilh a cDNA clone for hEPO(5). The inlact hEPO was also subjected to gel-permeation chromatography and reverse-phase HPLC under the same conditions as described above, to be used as a control. The derivatives. ercepl for the fully deglycosylatcd one, were lyophilized and dissolved in 10 mM sodium phosphate buffer, pH 7.4. containing 0.01% Tween 20, and stored at -80°C until USC.
lsololion of N-linked oligosacchoridrr-----Asialo hEPO (500 pg). prepared by neuraminidase digestion of the hormone and purified by reverse-phase HPLC as described previously (l5). was denatured in 480 pI of 0.5% SDS and 2% 2mercaptoethanol for 5 mi" in a boiling water bath. 500 pl of 0.5 M sodium phosphate buffer, pH 8.6. containing 10 mM EDTA and 250 pl of 7.5% NP-40 was then added to the digest was subjected to a Vydac Protein C4 column equilibrated with 15% acetonitrile in rcaction mixture, and digested with 10 mU of Glycopeptidase F at 3 7 T for 16 h. The passed-through fraction was pooled and lyophilized. Lyophilized aEialooligoraeeharlde.
0.1% trifluoroacetic acid to remow NP-40. SDS and the deglycosylated protein, and the P-4 column ( I x SO cm) equilibraled with 20 mM acctie acid to remove 2-dissolved in 500 p1 of distilled water. were further purified by gel filtration on a Bio-Gel mereaptoethanol and salts.
Lyophilized K2 (70 "mol) was dissolved in respeclivcly. were obtained by reverse-phase HPLC using a Vydac Peptide C18 column 500 pl of 0.1 M ammoniumbicarbonate. pH 8.3, and 10 pg of Endoproteinase GI"-C from Sraphylococcus aweus V8 (Boheringcr Mannheim) was added. After incubation at 3 7 T for 5 h. K2 was digested with additional 5 p g of the enzyme at 37°C for 19 h. Glycopeptides K2V1 (residues 21-31) and K2V2 (32.45). Le., those carrying the N24 and N38 sugar chains. respectively, were obtained by reverse-phase HPLC. KZVl and KZV2 (I5 nmols) were digesled with 10 mU of Glycapeptidase F for 24 h as described above to obtain the N24 and N38 sugar chains. The N83 sugar chain war also obtained from the K4 peptide in the same manner as described abovo. The column was equilibrated with 80% accmnitrile in 3% acctie acid-subjected to ISAA-HPLC, P size fractionation HPLC, using a YMC PA-03 column (4.6 x triethylamine. pH 7.5. at a flow rate of 1.0 mllmin and at 25°C. The saccharides were eluted with a decreasing linear gradient of acetonitrile from 8 0 to 56% in the same emission wavelength of 380 nm.

Pyridylominafion of N-linkcd oligosocehorides-----The
buffer over 60 min. and were detected at the excitation wavelength of 310 nm and thc Physicochemical onolysis-----Amino acid and amino sugar composition wers determined in a Pieo-Tag system(Watcrs Associatc)(21). Samples were hydrolyzed in evacuated tubes in 6 N HCI containing 1% phenol at 11O'C for 24 h for amino acid composition and in 4 N HCI at l 0 0 T for 6 h for amino sugar composition analysis. The phcnylthiaarbamoyl derivatives were quantified in a Pica- Tag  Biosystems 470A protein sequencer equipped with a model 120A on-line PTH-analyzcr.
sequence analysis was carried out by automated Edman degradation using an Applied SDS-PAGE was performed by the method of Laemmli(24) on 13.5% gclr, and isoelectric focusing was carried out on 4 % polyacrylamide gels containing 2% ampholite(25) by using an LKB multiphor electrophoresis unit. CD spectra were taken at 2 5 T in a Jarco I-5OOA recording spccrropolarimeter equipped with a DPJOON data proecssor, Ccllr obtained by cenlrifugation after washing were lyzcd in 1.0 ml of distilled water. added to 1.2 ml of Drahkin's solution and 2.5 ml of butanone. and were shaken for 10 min to extracl heme. One milliliter sampler were withdrawn from the bulanonc layer. and the radioactivity was measured in an Aloka autogamma ARC- bovine serum and were suspended in a binding medium (a-minimum essential medium IO6) were incubated with 0.5 nM I l~l l l h E P O and various dcglycorylated hEPO derivatives (0.024-6.25 nM) as compelitorr at 15°C for 3 h in 100 p1 of the binding medium.
The incubation mixtures were then tranrfered onto 200 pl of 10% S U E~O S C solution in polyethylene lubes and cenlrifuged for I mi" al 4 T and 5000 x g. The radioaetivily was counted in an Aloka autogamma ARC-300. Binding data were tubes were cut off just above Ihe cell layer after freezing at -80°C. and cell-associated analyzed with the computer program '"AccuFilTM" (Bcckman Instruments. h e . ) . Nonspecific binding Was mcasurcd in the presence of a 200-fold amount of unlabellcd hEPO. binding.
Specific binding of fully dcglycosylalcd hEPO to human erythroleukemic cells Specific binding was defined as the diffcrcncc between total binding and non-specific was determined in the same manner as described above except that 2.5 x 106 cells were used.
were prepared by a partial digestion of hEPO wilh Glycopeplidase F. Fig.  I shows the separation of these derivatives from the digest on a TSK G3000SW column.

The fractions indicated
with bars were separately pooled. and subjecled to reverse-phase HPLC to remove the rcagenlr. Fully dcglycorylated hEPO and OG(0)-hEPO were prepared as described in "EXPERIMENTAL PROCEDURES'. Homogcneity of these dcrivalivcs and the intact hEPO is shown in Fig. 2. and their molewlar weights were estimated to be 32, 28.
22. and 2lkDa for NG(2)-. NG(1)-, NG(0)-. and fully dcglycorylated hEPO. rcspectivcly. The intact hEPO and OG(0)-hEPO showcd the Same molecular weight of about 3SkDa. Thcse deglycosylatcd hEPO showed microhcterogcncity as seen in the intact hEPO by isoelectric focusing. as described previourly(l5). except for the fully deglycosylatcd hEPO, which showed a single hand a1 p l 6.5. Homogeneity of those preparations was further dcmonstraled by amino acid composition analysis as shown in Table I: the amino acid composition of the deglycosylated hEPO was in good accordance with theoretical V~U C E calculated from the amino acid sequence. Table  I also shows that no proteolysis in the polypeptide moiety of the hormone occurred during the preparation of these derivatives. Table  II  These derivativcs were subjected to peptide mapping. and the amino acid sequences of the glycosylated peptides wcre determined. to show which sugar chains were removed. Fig. 3 shows the peptide maps of the intact, NG(2)-. NG(I)-. NG(0)-. and OG(0)-hEPO. The glycosylated peptides, K2, K4 and K6, showed delayed retenlion times as de-N-glycosylation proceeded (Fig.3b-3d) or when de-0-glycorylation occurred (Fig.  3e). The amino acid sequence analysis of K2 obtained from NG(2)-hEPO demonstrated that it contains the N24 and N38 sugar chains. because no Asp was detected at the posilions corresponding to Asn24 and 38. On the contrary. Asp was dclcctcd quantitatively a1 the position corresponding lo Am83 in the K4 peptide. NG(1)-and NG(0)-hEPO were determined in the same way as described above that the farmer carrics N24 or N38 sugar chain and the latter carries no N-linked sugar chain. These results demonstrate that NGO)-hEPO lacks the N83 but carries the N24 and N38 sugar peptide maps showcd that undesirable degradation in the polypeptidc moiety did no1 chains. and NG(I)-hEP0 carries the N24 or N38 sugar chains. Furlhermore. these occur. Deglycosylated hEPO dcrivalivcs gave almost the same CD spectra as the intact hEPO at wavelengths from 200 Io 260 nm as shown in Fig. 4. These results demonstrated that no essential conformational changc occurred in the polypeptide backbone even though up to approximately 40% of thc total mass of the hormonc had been removed.

AnolySiS of Sire-Specific Clyeosylorion---Physieochemieal
analyses dcmonstratcd chat dcrivativcs which carry the N24 or N38 sugar chains. To know the structural NG(2)-hEPO lacks the N83 sugar chain and NG(1)-hEPO consists of an equal amount of diffcrences of the N-linked sugar chains at each attachmenl site. the N-linked sugar chains attached to Asn24. 38, and 83 were separately prepared and analyzed by sugar mapping. Fig.  5 shows the sugar maps of Ihe N24, N38. and N83 sugar chains on revcrsc phase HPLC in comparison with the map of thc N-linked sugar chains obtained from the intact hEPO. Eighteen peaks were observed in the sugar map of whole oligosaccharide chains (Fig. 4a). and the structures of the I I major peaks of 1. 2, 6. 7, 8. 9. 13. 14. IS. 17. and 18 were determined by IH-500 MHz NMR using the preparations obtained from 20 mg of hEPO as described in the legend of Fig.5. Fig. 5b-5d rhowr that the tctraantennary saccharndc without N-acctyllaetoraminyl repeats (peak 9) are dominant at each N-glycosylation site and that biantcnnary and triantcnnary saccharides are found IO be richer in the N24 than the N38 or N83 sugar chains. Distribution of the saccharides at each N-glycosylallon w e was further analyzed as to them sire by ISAA-HPLC. and the results are summarized sn Table 111.
In vivo ocriviry of dcglycosylared hEPO..-The biological activity of various deglycosylatcd hEPO in vivo have been compared with that of the intact hormone. As de-N-glycorylation proceeded. the activity lowered: NG(2)-. NG(I)-and NC (0) On the other hand.
as de-N-glycosylation proceeds. the activity increased: the 5.09 x I@ unitslmg. respectively.
These results. ,.e.. that de-N-glycorylatton causes an increase of the rn virro activity. were confirmcd by CFU-E colony-formation assay as shown in Fig. 6.
Partially or fully de-N-glycosylated hEPO showed an enhanced activity. In particular. NG(0)-hEPO showed a three-times-higher spcciftc activity than the intact form.
were pooled. The saccharides were eluted with a series of the gradient of n-butanol from 0.06 to 0.2751 in the same buffer over 170 rnin and detected by their fluorescence at the excitation wavelength of 320 nm' and the emission wavelength of 400 nm. The structure of the N-linked sugar chain corresponding to elcven major peaks was determined by 500 MHz 1H-NMR using preparations obtained from 20 mg of hEPO. 1. an epimer of peak 9 formed during the pyridylarninatian: 2, triantennary saccharide: 6 and 7. triantennary saccharides with one N-acetyllactosaminyl repeat: 8, biantcnnary saccharide: 9. tetraantennary saccharide; 13 and 14. tetraantennary saccharides with one N-acttyllactosarninyl repeat; 15. triantennary saccharide differcnl from peak 2 in branching: 17, tetraantennary saccharide with two N-acelyllactorarninyl repeats;

18.
tetraantennary saccharide with three N-acetyllactosaminyl repeats. were incubated for two days at 37'C in a humidificd 5% C02 atmorphcrc. Colonies convatnine more than emht cells were scored usim an inverted microscooc. The assav was enmiid out in dupllcaye for each point.  Table IV were plotted against the number of sialic acids bound to the hEPO derivatives.