Eur. J. Biochcm. 188,405-411 (1990) 0FEBS 1990

The role of carbohydrate in recombinant human erythropoietin Eisuke TSUDA', Gosei KAWANISHI', Masatsugu UEDA', Seiji MASUDA' and Ryuzo SASAKI'

' Research Institute of Life Science, Snow Brand Milk Products Co., Tochigi, Japan Department of Food Science and Technology, Faculty of Agriculture, Kyoto University, Japan (Received July 2/0ctober 6, 1989) - EJB 89 0916

1. Recombinant human erythropoietin has N-linked sugar [Tsuda et al., (1988) Biochemistry 27, 5646 - 56541. Here we have demonstrated the presence of 0-linked sugar (0.85 mol/mol erythropoietin) composed of sialic acid and GalP(1- 3)GalNAc. 2. To investigate the role of these sugars, erythropoietins deglycosylated to different extents were prepared using specific glycosidases. Sugars are not essential for in vitro biological activitiy of erythropoietin, because the fully deglycosylated erythropoietin had the full activity when assayed with in vitro bioassay methods. Asialylation yielded erythropoietin with higher affinity to the receptor than the undigested hormone and therefore an increased in vitvo activity. Although erythropoiein from which N-linked or total sugars were removed also had higher affinity for the receptor, their in vitro activity remained unchanged compared with that of the undigested erythropoietin for unknown reasons. On the other hand, removal of sialic acids or N-linked sugar abolished the in vivo biological activity completely, indicating that the presence of N-linked sugar with terminal sialic acids is required for the hormone to reach target sites; full deglycosylation resulted in total loss of the in vivo biological activity of erythropoietin. 3. Incubation of asialo-erythropoietin and fully deglycosylated recombinant human erythropoietin at 70 "C for 15 min decreased the biological activity to 35% and 11YOof the initial activity, respectively, while the undigested erythropoietin lost no activity. Thus resistance of erythropoietin to thermal inactivation is largely due to the presence of sugars, and terminal sialic acids greatly contribute to the stability.

clearance by binding of asialylated EPO to hepatic receptors recognizing exposed galactose residues. Digestion of human urinary EPO by endoglycosidase F, by which most N-linked oligosaccharides were removed, yielded EPO with no in vivo activity, while it retained the in vitro activity [5].It has been reported, using EPO partially purified from serum of anemic sheep, that the in vivo activity was completely lost by asialylation but the in vitro activity was increased when asialo-EPO was assayed at low levels [8]. Thus, it appears from these results that terminal sialic acids and N-linked carbohydrates of EPO are not essential for manifestation of the in vitro activity, although the mechanism by which asialo-EPO gains an increased in vitvo activity is poorly understood. No information on a role for 0-linked sugar in EPO is available. Production of rHuEPO in large quantities, and the recent isolation of a highly active endo-a-N-acetylgalactosaminidase that can remove 0-linked carbohydrates attached to the asialylated protein [9, 101, have allowed us to examine the Correspondence to M. Ueda, Research Institute of Life Science, function of 0-linked sugar in EPO action. Snow Brand Milk Products Co. Ltd, 519, Shimo-ishibashi, IshibashiIn this paper, we describe the preparation of EPOs that machi, Shimotsuga-gun, Tochigi, 329-05 Japan have lost carbohydrates to different extents by enzymatic Abbreviations. EPO, erythropoietin; rHuEPO, recombinant hu- deglycosylation of rHuEPO using sialidase, 8-galactosidase, man erythropoietin; CFU-E, colony-forming units of erythroid. B-N-acetylhexosaminidase, endo-a-N-acetylgalactosaminiEnzymes. Endoglycosidase F (EC 3.2.1.96); endo-cc-N-acetylgalactosaminidase (EC 3.2.1.97); sialidase (EC 3.2.1.18); P-galac- dase and N-glycanase, and measured their biological activities. tosidase (EC 3.2.1.23); P-N-acetylhexosaminidase (EC 3.2.1.52); N- Binding of these preparations to EPO receptors on target cells glycanase (EC 3.5.1.52); trypsin (EC 3.4.21.4); V8 protease (EC was investigated to find how the activity was altered through 3.4.21.29) enzymatic deglycosylation. The contribution of carbohydrates Erythropoietin (EPO), a highly glycosylated 34 - 38-kDa protein, regulates red blood cell levels by promoting maturation and growth of erythroid precursor cells. The molecular mass of the peptide chain deduced from its cDNA is about 18 kDa [l] and therefore about 50% of the mass of the fully glycosylated EPO is made up of Carbohydrates. The presence of three N-linked carbohydrate chains and their structures have been reported for EPO from human urine [2-41, and recombinant human erythropoietin (rHuEPO) from baby hamster kidney cells [3] and Chinese hamster ovary cells [2,4] Initial analyses of the carbohydrate composition of human urinary EPO found no 0-linked sugar [5] but this type of sugar was found later in both urinary EPO and rHuEPO [2, 61, (and this paper). Removal of terminal sialic acids from carbohydrate chains of urinary EPO [7] and rHuEPO [6] increased their in vitro activity but abolished the in vivo activity completely. The latter is thought to be due to rapid metabolic

406 to resistance against heat-inactivation of EPO was also investigated.

MATERIALS AND METHODS Recombinant human erythropoietin

rHuEPO from the culture supernatant of baby hamster kidney cells [6] which had been engineered to produce it was purified to homogeneity as described previously [ 111. Glycosidase digestion of recombinant human erythropoietin

rHuEPO from which carbohydrates were removed enzymatically to various extents was prepared as described below (see also Fig. 1). Asialo-EPO was prepared by digesting EPO (300 pg) with 30 mU sialidase from Streptococcus sp. (Seikagaku Kogyo Co. Ltd, Tokyo, Japan) for 3 h at 37°C in 200 pl 100 mM sodium acetate pH 6.5, containing 10 mM CaC12. After incubation, the reaction mixture was dialysed against distilled water at 4°C and then lyophilized. Exposed P-Gal residues in asialo-EPO were removed by the action of P-galactosidase. Asialo-EPO (100 pg) was incubated with I mU P-galactosidase from Streptococcus sp. for 3 h at 37°C in 200 pl 50 mM sodium phosphate pH 5.5. P-Gal and P-GlcNAc residues were removed by double digestion of asialo-EPO (100 pg) with 1 mU P-galactosidase and 100 mU j-N-acetylhexosaminidase (from jack bean; Seikagaku Kogyo Co. Ltd) by incubating for 24 h at 37°C in 200 pl 50 mM sodium phosphate pH 5.5. The N-linked carbohydrates were removed by incubating EPO (100 pg) with 10 U N-glycanase from Flavobacterium menigosepticum (Genzyme Corporation, Boston, MA) for 24 h at 37°C in 200 pl 100 mM sodium phosphate pH 8.6, containing 20 mM EDTA. Fully deglycosylated EPO was prepared by successive digestion with sialidase, endo-a-N-acetylgalactosaminidase, and N-glycanase. EPO (100 pg) was incubated with 5 mU sialidase for 3 h at 37°C in 250 p1 100 mM sodium acetate pH 6.5, containing 10 mM CaC12. After incubation, the buffer was changed to 25 mM sodium acetate pH 4.5 by ultrafiltration dialysis using Centri Cut Mini V-10 (Kurashiki Boseki, Osaka, Japan) and the volume adjusted to 300 pl by the addition of the above buffer. Then 15 mU endo-a-N-acetylgalactosaminidase (the kind gift of Dr K. Yamamoto, Kyoto University [9, 101) in 10 pl 600 mM potassium phosphate pH 6.0 was added to the mixture which was kept at 37°C for 3 h. After digestion, the buffer was changed to 100 mM sodium phosphate pH 8.6 containing 20 mM EDTA by ultrafiltration dialysis and the volume adjusted to 80 pl. N-Glycanase ( 5 U) in 20 pl 50% glycerol containing 2.5 mM EDTA was added and the mixture kept at 37°C for 24 h. A control experiment for each digestion was done under the same conditions as described above except that glycosidases were omitted. All preparations were stored at -20°C until used for measurement of EPO concentrations and activity. SDSlpolyacrylamide gel electrophoresis

To confirm the completion of each glycosidase reaction, the preparations containing digested EPO were analysed by SDS/polyacrylamide gel electrophoresis [12] with a separating gel containing 13% acrylamide. After electrophoresis, protein bands were detected by staining with Coomassie brilliant blue.

Erythropoietin assay

Concentrations of EPO were measured with a sandwichtype enzyme immunoassay in which two monoclonal antibodies that recognized different epitopes on the peptide chain [ l l ] were used. The in vitro biological activity was assayed with two methods: the stimulatory effect on incorporation of [,H]thyrnidine into DNA in cultured fetal liver cells of BALB/ c mouse [13] and on formation of CFU-E colonies in semisolid culture of mouse bone marrow cells [14]. The in vivo activity was measured using starved rats [15]. Briefly, fiveweek-old female Wistar rats, weighing about 120 g, were starved and EPO samples, dissolved in NaCl/Pi (10 mM sodium phosphate pH 7.4 containing 145 mM NaC1) containing 0.1 % rat serum albumin, were injected subcutaneously on the 2nd and 3rd days after the start of starvation. On the 4th day, 1 pCi 59Fe was injected subcutaneously. After 18 h, blood samples were taken from the hearts and radioactivity of red blood cells was measured. Four rats per assay were used, and eight rats injected with 0.1% rat serum albumin in NaCl/P, were used as a control group. Binding of 1251-labeledrecombinant human erythropoietin to its receptor

Binding of '251-rHuEP0 to EPO receptors on mouse erythroleukemia cells was assayed as described previously [16] with a minor modification. A cell suspension (100 pl) containing 5 x lo6 TSA8 cells was mixed with 50 pl NaC1/Pi containing 60 mM Hepes pH 7.2, 0.3% bovine serum albumin, 0.6% NaN,, and 1251-EP0(3.6 nM). After incubation for 3 h at 15 "C, the cells were pelleted, washed once with NaC1/Pi, and suspended in 200 p1 NaC1/Pi. The suspension was layered on 800 p1 NaC1/Pi containing 10% bovine serum albumin and the cells were separated from unbound ligand by centrifugation. The tube contents were frozen in solid CO,/cthanol, and the tips were cut off just above the cell pellet to count the radioactivity. Analyses of 0-linked oligosaccharides in recombinant human erythropoietin

rHuEPO (5.5 mg) was desialylated by incubating at 90'C for 60 min in 0.01 M HCl. 0-linked oligosaccharides were removed using endo-a-N-acetylgalactosaminidasewhich released the disaccharide, GalP( 1 - 3)GalNAc, from glycoproteins possessing serine or threonine 0-glycoside linkages [9, 101: the asialo-EPO was incubated with 100 mU enzyme at 37°C for 24 h in a total of 1 ml 25 mM citrate pH 4.5. The amount of carbohydrates released was measured using the Morgan-Elson reaction [17] that detected carbohydrates containing an N-acetylhexsosamine in which C-I and C-4 are not substituted; 10 vol. ethanol chilled at -20°C was added to precipitate proteins in the reaction mixture. After centrifugation, the supernatant containing carbohydrates was dried in a rotary evaporator. The dried material was dissolved in 1 ml water and purified on a Bio-Gel P-4 column (1.6 x 90 cm). The presence of carbohydrates in the fractions was detected by TLC of each fraction, visualizing with orcinol/H2S04. The purified oligosaccharides were hydrolysed to monosaccharides by incubation in 2.5 M trifluoroacetic acid at 100°C for 6 h in an evacuated sealed tube. The monosaccharides were separated by HPLC on a Shimadzu model LC-6A apparatus with a column of ISA-O7/S2504(4 x 250 mm) eluted with 0.25 M potassium borate pH 8.5 at a flow rate of 0.5 ml/

407

i

( II

2

3

1 Fuc(a1-6)

(Sialic Acid)

1

( Ill

\

Man(pl-4)GlcNAc(pl-4)Gl~NAc

2

3

- Asn

t

4 3

1 (Sialic Acid) 1

5

Fig. 1 . Action of glycosidases on oligosaccharides of recombinant human erythropoietin. Cleavage sites of oligosaccharides in rHuEPO by glycosidases are shown. (I, 11) Structures of representative complex-type N-linked oligosaccharides of rHuEPO ; (111) 0-linked sugar. The lines with arrowheads numbered 1 - 5 indicate the cleavage sites by glycosidases: 1, sialidase; 2, B-galactosidase after sialidase digestion; 3, /I-galactosidaselp-N-acetylhexosaminidase after sialidase-digestion; 4, N-glycanase; 5, sialidase/endo-a-N-acetylgalactosaminidaselNglycanase

min at 60 "C. The separated monosaccharides were detected by a post-column fluorometric method [18]. The monosaccharides were also analyzed with an amino acid analyzer, Hitachi L-8500. The purified oligosaccharides from EPO were reductively aminated with 2-aminopyridine using sodium cyanoborohydride [19]. The pyridylamino derivatives of oligosaccharides thus prepared were gel-filtered on a Sephadex G-10 column to remove reagents used for amination. The derivatives were fractionated and identified by HPLC on a reverse-phase Shimpack CLC-ODS column (6 x 150 mm, Shimadzu) developed with 10 mM sodium phosphate pH 5.0 at a flow rate of 1.0 ml/min at 55°C [20] and on an amide-adsorption TSK gel Amide-80 column (4.6 x 250 mm, Tosoh) developed with a solvent composed of 3 % acetic acid in water with triethylamine pH 7.3/acetonitrile (20: 80, by vol.) at a flow rate of 1.0 ml/min at 45°C [21]. GalB (1 -3)GalNAc was prepared from fetuin by the same procedures as the oligosaccharides were prepared from EPO. The pyridylamino derivatives of maltose, maltotriose, maltotetraose, melibiose, and GalP(1- 3)GalNAc were used as standards in HPLC to identify the compound of interest. RESULTS The presence of 0-linked oligosaccharides in recombinant human erythropoietin

We have suggested [6] that human urinary EPO and rHuEPO produced by two types of hamster kidney cells contained an 0-linked oligosaccharide, because most of the N-acetylhexsosamine from these EPOs was GlcNAc but about 10% of the total was always GalNAc. With rHuEPO produced by baby hamster kidney cells, here we have demonstrated the presence of an 0-linked oligosaccharide and identified its structure. Asialo-EPO was digested with endo-a-Nacetylgalactosaminidase to detach 0-linked oligosaccharides,

if they are present, from the peptide chain. Because of the substrate specificities of the glycosidase, some of the 0-linked oligosaccharides might not be cleaved and would escape analysis. The reaction was completed with the concomitant release of about 0.85 mol/mol of an oligosaccharide that contained an N-acetylhexosamine residue at the reducing end. No carbohydrates were released from EPO which had not been desialylated. It was found from size estimation with TLC that the carbohydrate consisted of two or three monosaccharide residues. When this oligosaccharide was hydrolyzed to the constitutive monosaccharides and then analyzed by HPLC, hexosamine and galactose were found in a molar ratio of 3 : 1. Further analysis of the monosaccharides with an amino acid analyzer indicated that the hexosamine was galactosamine. Little glucosamine was found. The oligosaccharide prepared from the asialo-EPO using endo-a-N-acetylgalactosaminidase was labeled with a 2aminopyridine and the resulting pyridylamino derivative of the fluorescent compound was analyzed by HPLC on two columns, ODS silica and Amide-80. The sample behaved identically on both columns to GalP(1- 3)GalNAc obtained from desialylated fetuin. From these results, we concluded that EPO contained one 0-linked oligosaccharide residue composed of sialic acid and GalP(1- 3)GalNAc (Fig. 1). It is not known whether the sialic acids are linked to Gal or GalNAc or both; the structure shown in Fig. 1 is putative in terms of the residues to which sialic acids attach. Peptide fragments were prepared from EPO by digestion with proteases (trypsin and V8 protease) and purified. When all of the fragments that were confirmed to contain serine and threonine residues by amino acid analyses were analyzed with a protein sequencer, we found both amino acids in all the expected positions (see [l])of the fragments except those containing Ser126. Serine could not be sequenced in most of the fragments thought to contain Ser126 but in some fragments it could be. These results indicate that 0-linked glycosylation of EPO occurs at Ser126 but it is incomplete.

408 Table 1. Biological activities of recombinant human erythropoietin deglycosylated enzymatically rHuEPO was treated with or without glycosidases as described in Materials and Methods. After treatment, molar concentrations of 49.6 recovered EPO were measured with a sandwich-type enzyme immunoassay [I I]. The concentration of EPO treated in the absence of 3 7 . 2 --+ glycosidases was defined as 100%. Biological activity in vitro was assayed using two methods: [3H]thymidine incorporation into cellular 2 4 . 8 -b DNA [Ill and CFU-E colony formation [14]. Dose/response curves of the two assay methods were drawn for the untreated EPO and each deglycosylated EPO. The curves were also drawn for the control EPOs which were treated in the absence of glycosidases; corrections, when necessary, for nonenzymatic alteration of the activity that occurred 12.4 during treatment for deglycosylation could be made. The molar concentration of the untreated EPO giving 50% of the maximum activity Fig. 2. Analysis of glycosidase-digested recombinant human erythro- was calculated from the curve and defined as 100%. Biological activity poietins by SDS/polyacrylamide gel electrophoresis. rHuEPO protein of a deglycosylated EPO at this concentration was calculated and ( 5 pg) was used. Lane 1, standard proteins; lane 2, untreated rHuEPO; the concentration of EPO required to express the same activity was lane 3, sialidase-digested; lane 4, sialidase/P-galactosidase- digested; calculated from the dose/response curve drawn for the untreated EPO. lane 5 , sialidase/B-galactosidase/B-N-acetylhexosaminidase-digested;The concentration thus obtained was compared with that of the lane 6, N-glycanase-digested; lane 7, fully deglycosylated (sialidase/ untreated EPO described above. In vivo activity of untreated and endo-a-N-acetylgalactosaminidase/N-glycanase-digested) deglycosylated EPOs was assayed using 0.54, 1.08, and 2.16 pmol EPO. The activity of the untreated EPO was defined as 100%. rHuEPOs are as follows: EPO-0, untreated rHuEPO; EPO-1, sialidase-digested (asialo-rHuEPO); EPO-2, sialidase/P-galactoDeglycosylation of recombinant human erythropoietin sidase-digested; EPO-3, sialidase//$galactosidase/P-N-acetylhexosrHuEPO produced by baby hamster kidney cells has N- aminidase-digested; EPO-4, N-glycanase-digested; EPO-5, fully delinked and O-linked oligosaccharides [6] and their structure glycosylayted (sialidase/endo-a-Nacetylgalactosaminidase/N-glycahas been analyzed [3] (and this paper). Oligosaccharide chains nase-digested rHuEPO) (see also Fig. 1 )

--+

have sialic acid residues in their nonreducing terminals. Approximately 80% of the N-linked oligosaccharides are of the tetra-antennary complex type with or without N-acetyllactosamine repeating units. Others are tri- or biantennary complex type oligosaccharides. The structure of the 0-linked oligosaccharide was identified as GalP(1- 3)GalNAc with sialic acids. Fig. I shows representative structures of N-linked and 0-linked oligosaccharides of rHuEPO and cleavage sites of the enzymes used here. Sialidase removes terminal sialic acids, producing asialo-EPO. P-Galactosidase cleaves off the exposed Gal residues of N-linked oligosaccharides in asialoEPO but does not split the /31-3 linkage between Gal and GalNAc in the 0-linked oligosaccharide. Double digestion of asialo-EPO with P-galactosidase and /3-N-acetylhexosaminidase removes all of the exposed Gal and GlcNAc residues from N-linked oligosaccharides and the trimannosyl core structure is exposed. N-Glycanase completely eliminates Nlinked oligosaccharides from the untreated EPO, and endo-ccN-acetylgalactosaminidase removes 0-linked oligosaccharides from asialo-EPO; therefore successive digestion of EPO with sialidase, endo-a-N-acetylgalactosaminidase,and Nglycanase yields the fully deglycosylated EPO. Completion of the enzyme reactions were confirmed with SDS/polyacrylamide gel electrophoresis. The untreated rHuEPO migrated with a molecular mass of 37 kDa. EPO digested with sialidase, sialidaselj-galactosidase, sialidaselj-galactosidase/ /3-N-acetylhexosaminidase, N-glycanase and fully deglycosylated EPO had the apparent molecular masses of 31,28,24, 19, and 18 kDa, respectively (Fig. 2). The minor band of 18 kDa in the N-glycanase-digested preparation seems to correspond to fully deglycosylated EPO, agreeing with the finding described in the previous section that some molecules lack the 0-linked oligosaccharide. Erythropoietin activity of glycosiduse-digested recombinant human erythropoietin

Digestion of EPO by glycosidases included complicated procedures, as described in Materials and Methods, and there-

rHuEPO

Recovery of CPO after glycosidase digestion

In vitro activity [3H]thyrnidine CFU-E incorporation colony formation

In vivo activity

YO EPO-0 EPO 1 EPO - 2 EPO - 3 EPO - 4 EPO - 5 ~

-

118 112 98 93 88

100 252 252 266 81 106

100 180 225 225 90 67

100 0 0 0 0 0

fore preparations containing high concentrations (0.3 - 1.5 mg/ml) were used for digestion to avoid loss of protein during processing. To investigate alteration of the biological activity by glycosidase digestion, it is necessary to know EPO concentrations accurately in digested samples. We have developed a sandwich-type enzyme immunoassay using two monoclonal antibodies that recognized different peptide epitopes on EPO [ll]. With this method we can measure concentrations of the peptide chain in the samples after digestion. EPO was digested with glycosidases and the same amount treated identically except that glycosidases were omitted. The EPO peptide chain in both samples was measured by enzyme immunoassay (Table 1). A high recovery of deglycosylated EPO indicates that deglycosylation does not alter the immunological properties with respect to interaction with the monoclonal antibodies used here. Biological activity in vitro of the deglycosylated EPOs was assayed using their stimulatory effects both on ['Hlthymidine incorporation into DNA of cultured mouse fetal liver cells and on CFU-E colony formation in semi-solid culture of mouse bone marrow cells. Asialo-EPO and EPO digested with sialidaselb-galactosidease and sialidase/fi-galactosidase/fi-N-

409 I

1

w

50 0

?!m

0 Unlabelled rHuEPO

( nM )

I n c u b a t i o n T i m e ( min

Fig. 3. Effects of deglycosylation on the affinity of recombinant human Fig. 4.Effects of deglycosylation on thermal inactivatianof recombinant erythropoietin for its receptor. Binding of 1251-labelled untreated human erythropoioetin. rHuEPO (50 ng/ml) in sodium phosphate conrHuEPO to the receptor was assayed using mouse erythroleukemia taining 0.1% bovine serum albumin was incubated at 70°C. At intercells as described in Materials and Methods. Inhibition of the binding vals, samples were measured for biological activity in vitro using by the untreated and deglycosylated EPOs was examined to estimate [3H]thymidine incorporation into DNA of fetal mouse liver cells [13]. alteration of the affinity by deglycosylation. Binding mixtures The initial activities of rHuEPOs were defined as 100%. The logarithm contained 1.2 nM '251-labelled untreated EPO and varied amounts of activities was plotted against incubation time. (0) Untreated (0-12 nM) of the unlabelled EPO. B, represents specific binding rHuEPO; ( 0 ) sialidase-digested; ( A ) sialidaselp-galactosidase-diwhich was calculated by subtraction of non-specific binding (in the gested ; (A) sialidase//3-galactosidase/fl-N-acetylhexosaminidase-diN-glycanase-digested; (I) fully deglycosylated (sialidase/ presence of 200-fold unlabelled EPO) from total binding (in the ab- gested; (0) sence of unlabelled EPO); here total binding and non-specific binding endo-a-N-acetylgalactosaminidase/N-glycanase-digested) were 16 500 cpm and 2 500 cpm, respectively, and therefore specific binding was 14000 cpm, defined as 100%. B shows specific binding in the presence of the unlabelled EPO at each concentration as indi- deglycosylated EPOs (asialo-EPO, sialidaselp-galactosidasecated in the abscissa. (0)Untreated rHuEPO; ( 0 )sialidase-digested; digested EPO, and fully deglycosylated EPO) showed even ( A ) sialidaselp-galactosidase-digested ; (A)sialidase/fl-galactosidase/ higher affinity, fourfold higher than the undigested protein. /3-N-acetylhexosaminidase-digested; (0)N-glycanase-digested; (I) fully deglycosylated (sialidaseleudo-a-N-acetylgalactosaminidase/N-Heat stability of deglycosylated recombinant human erythropoietin glycanase-digested)

acetylhexosaminidase showed 2.5-fold more activity than the undigested EPO (Table 1). N-Glycanase-digested and fully deglycosylated EPOs had similar in vitro activity to the untreated EPO. In vivo biological activity of all preparations digested with glycosidase disappeared. Addition of glycosidases to the assay mixtures of EPO at the concentrations that were expected to be derived from the EPO samples after glycosidase digestion had no effect on the activity of the undigested EPO; the possibility that these alterations in biological activity of EPO by deglycosylation were caused by the action of glycosidases on target cells, which might remain active in EPO samples after glycosidase digestion, was excluded. Binding affinity of deglycosylated erythropoietin to its receptor

To discover which mechanism caused an increase in the in vitro activity of these partially deglycosylated EPOs, their

binding to EPO receptor was studied. Changes in the affinity of deglycosylated EPOs to the receptor were surveyed by their inhibitory effect on binding of the native '''I-rHuEPO to the receptor using a mouse erythroleukemia cell line, as shown in Fig. 3. All of the deglycosylated EPOs had higher affinities than the undigested EPO. EPO from which N-linked sugar had been completely removed by digestion with N-glycanase showed a twofold increase in binding affinity and other

To investigate the effects of glycosylation on EPO structure, the resistance of glycosidase-digested EPOs to heat inactivation was examined. Untreated and glycosidase-treated EPOs were incubated at 70°C and at intervals EPO activity of samples was measured using the stimulatory effect of EPO on [3H]thymidine incorporation into DNA of fetal mouse liver cells. As shown in Fig. 4, the untreated EPO was the most stable. Digestion of EPO by sialidase, N-glycanase, sialidaselp-galactosidase, and sialidase/B-galactosidase/fkNacetylhexosaminidase and full deglycosylation decreased heat stability in that order. When we compared the activity of EPOs incubated at 70°C for 15 min, the undigested EPO had not lost any activity but the activities of fully deglycosylated EPO decreased to 11YOof the undigested EPO; thus, carbohydrates in EPO contribute largely to the resistance to heat inactivation. Asialylation yielded EPO with an activity of 35% of the undigested EPO, indicating that sialic acids are important for stability of EPO in heating. N-Glycanase-digested EPO is more resistant than EPOs deglycosylated by combination of the enzymes. This higher resistance is probably due to the presence of sialic acids in 0-linked oligosaccharides but a contribution of the core structure, Galp(1- 3)GalNAc, is not excluded because we have not tested EPO which had only the core structure as carbohydrate. DISCUSSION In previous papers, we reported the presence of N-linked oligosaccharides in rHuEPO produced by baby hamster kid-

410 ney cells [6] and their complete structures [3]. In this paper, we have shown first, the presence of an 0-linked sugar composed of sialic acid and Galp(1- 3)GalNAc, and, second, examined the role of sugars attached to EPO towards biological activity, interaction with EPO receptor, and resistance against heat inactivation. For this, we prepared EPOs deglycosylated to different extents by specific glycosidases. Analyses of products of deglycosylation reactions with SDS/ polyacrylamide gel electrophoresis showed single bands, except for N-glycanase-digested EPO, with mobilities expected for each theoretical molecular mass deduced from the structures of polypeptide and N-linked and 0-linked oligosaccharides [l, 31 (and this paper; Fig. 2), indicating that each deglycosylation reaction was completed under the conditions used here. The N-glycanase-digested preparation showed one more minor band which corresponded to fully deglycosylated EPO, suggesting that some of the rHuEPO lacked an 0-linked sugar. Together with the fact that deglycosylated EPOs kept similar properties to that of undigested EPO with respect to interaction with specific monoclonal antibodies against EPO (Table l), deglycosylation reactions with enzymes used here did not cause any damage to the peptide chain of EPO. Experiments with the fully deglycosylated EPO prepared in this way clearly demonstrated that carbohydrates attached to EPO were not essential for expression of EPO activity. It appears that the carbohydrates are required to protect EPO from in vivo inactivation that occurs before the hormone reaches target cells. Asialo-rHuEPO showed a 2.5-fold increase in the in vitro activity, confirming the previous results with sheep plasma EPO [8], human urinary EPO [6], and rHuEPO [6]. EPOs further deglycosylated with P-galactosidase or with both pgalactosidase and p-N-acetylhexosaminidase also showed a similar degree of increase of the in vitro activity. This increase would be due to an increase in affinity of the deglycosylated EPOs for its receptor on the target cells (Fig. 3). However, Nglycanase-digested EPO and fully deglycosylated EPO had similar in vitro activity to that of undigested EPO, although they bound to the receptor with twofold and fourfold higher affinity, respectively, than did the undigested EPO. Dordal et al. [5] reported that human urinary EPO from which most of N-linked sugars were removed formed aggregates during a long incubation. In fact, we measured the affinity of EPOs for the receptor by incubating them with target cells for 3 h at 15"C, while their in vitro activity was measured by incubation with target cells for 20 h at 37°C (incorporation of [3H]thymidine into cellular DNA) or 45 h at 37°C (colony formation). An increase in activity of N-glycanase-digested and fully deglycosylated EPOs may be counterbalanced by inactivation due to aggregation of EPO during incubation for the activity measurement. The other intriguing possibility, that the sugar chains are involved in the efficiency of signal transduction after binding of the ligand to the receptor, remains to be investigated. The carbohydrate of human chorionic gonadotropin is not involved in receptor-binding activity of the hormone but it is required for the expression of the hormonal activity [22]. It is noted that N-glycanase-digested EPO, which still retained terminal sialic acids on the 0-linked sugar, showed a twofold increase in the affinity to the receptor but all other deglycosylated EPOs with no sialic acids gained higher affinity, fourfold higher than the undigested EPO. It appears from these results that the presence of terminal sialic acids decreases the binding affinity of EPO for its receptor.

Not only asialo-EPO but also other partially deglycosylated EPO lost their in vivo activity. Exposure of terminal sugar residues in the deglycosylated EPOs may promote their metabolic clearance by binding to the hepatic receptors; asialylation exposes terminal galactose residues, sialidaselogalactosidase digestion exposes terminal GlcNAc residues, and sialidaselp-galactosidaselP-N-acetylhexosaminidasedigestion exposes terminal mannose residues. The presence of hepatic receptors for these exposed sugars has been shown [23- 251. Fully deglycosylated and N-glycanase-treated EPOs lost their in vivo activity for unknown reasons. These deglycosylated products may be highly susceptible to proteolytic enzymes and thereby be digested before they reach erythropoietic tissues where thay act. Rat liver al-acid glycoprotein that was synthesized in the presence of tunicamycin, an inhibitor of N-glycosylation of protein, was cleared from plasma at a higher rate than the glycosylated protein [26]. According to some studies on the plasma half-life of deglycosylated proteins [27 - 291, the deglycosylated proteins in all cases disappeared faster from the circulation than their glycosylated counterparts. The pathway for removing these deglycosylated proteins from circulation is not clear. Recent studies of the function of the sugar chain in rHuEPO by the use of recombinant DNA techniques indicated that N-linked carbohydrate chains attached to Am24 and Am38 might be involved in the in vitro activity, because the absence of sugar at either position reduced the activity [30]. But this experiment does not exclude the very likely possiblity that EPO molecules that lacked these sites for attachment of sugar may fail to make the conformation appropriate for manifestation of the activity in their biosynthetic pathway. We thank Dr N. Takahashi and Dr A. Murakami for helpful suggestions, Dr K. Yamamoto for the generous gift of endo-P-Nacetylgalactosaminidase and F. Kobayashi for skilful technical assistance.

REFERENCES 1. Jacobs, K. Shoemaker, C. Rudersdorf, R. Neill, S. D. Kaufman, R. J. Mufson, A. Seehra, J. Jones, S. S. Hewick, R. Fritsch, E. F. Kawakita, M. Shimizu, T. & Miyake, T. (1985) Nature 313, 806- 810. 2. Sasaki, H. Bothner, B. Dell, A. & Fukuda, M. (1987) J . Biol. Chem. 262,12059- 12076. 3. Tsuda, E. Goto, M. Murakami, A. Akai, K. Ueda, M. Kawanishi, G. Takahashi, N. Sasaki, R. Chiba, H. Ishihara, H. Mori, M. Tejima, S. Endo, S. & Arata, Y. (1988) Biochemistry 27,56465654. 4. Takeuchi, M. Takasaki, S. Miyazaki, H. Kato, T. Hoshi, S. Kochibe, N. & Kobata, A. (1988) J . Biol. Chem. 263, 36573663. 5. Dordal, M. S. Wang, F. F. &Goldwasser, E. (1985) Endocrinology 114,2293-2299. 6. Goto, M. Akai, K. Murakami, A. Hashimoto, C. Tsuda, E. Ueda, M. Kawanishi, G. Takahashi, N. Ishimoto, A. Chiba, H. & Sasaki, R. (1988) Biotechnology 6, 67-71. 7. Sasaki, R. Yanagawa, S. & Chiba, H. (1987) Methods Enzymol. 147,328-340. 8. Goldwasser, E. Kung, C. K.-13. & Eliason, J. (1974) J . Biol. Chem. 249,4202 - 4206. 9. Yamamoto, K. Fan, J.-Q. Kadowaki, S. Kumagai, H. & Tochikura, T. (1987) Agric. Biol. Chem. 51, 3169-3171. 10. Fan, J.-Q. Kadowaki, S. Yamamoto, K. Kumagai, H. & Tochikura, T. (1988) Agric. Biol. Chem. 52,1715-1723. 11. Goto, M. Murakami, A. Akai, K. Kawanishi, G. Ueda, M. Chiba, H. & Sasaki, R. (1989) Blood 74, 1415-1423. 12. Laemmli, U. K. (1970) Nature 227,680-685.

411 13. Brandan, N. C. Cotes, P. M. & Espada, J. (1981) Br. J . Hematol. 47,461 -468. 14. Iscove, N. N. Sieber, F. & Winterhalter, K. H. (1974) J. Cell. Physiol. 83, 309 - 320. 15. Goldwasser, E. & Gross, M. (1975) Methods Enzymol. 37, 109121. 16. Sasaki, R. Yanagawa, S. Hitomi, K. & Chiba, H. (1987) Eur. J . Biochem. 168,43-48. 17. Aminoff, D. Binkley, W. W. Schaffer, R. & Mowry, R. W. (1970) The carbohydrates (Pigman, W. & Horton, D., ed.) vol. 2B, p. 740 Academic Press, New York. 18. Mikami, H. &Ishida,Y. (1983) BunsekiKagaku32, E207--210. 19. Hase, S. Ibuki, T. & Ikenaka, T. (1984) J . Biochem. (Tokyo) 95, 197-203. 20. Tomiya, N. Kurono, M. Ishihara, H. Tejima, S. Endo, S. Arata, Y. & Takahashi, N. (1987) Anal. Biochem. 163,489-499. 21. Tomiya, N. Awaya, J. Kurono, M. Endo, S. Arata, Y. & Takahashi, N. (1988) Anal. Biochem. 171,73-90. 22. Kalyan, N . K. & Bahl, 0. P. (1983) J . Biol. Chem. 258, 67-74. 23. Morell, A. G. Gregoriadis, G. Scheinberg, I. H. Hickman, J. & Aschwell, G. (1971) J . Biol. Chem. 246, 1461- 1467.

24. Schlesinger, P. H. Doebber, T. W. Mandell, B. F. White, R. Deschryver, C. Rodman, J. S. Miller, M. J. & Stahl, P. (1978) Biochem. J. 176, 103-109. 25. Stahl, P. D. Rodman, J. S. Miller, M. J. & Schlesinger, P. H. (1978) Proc. Natl Acad. Sci. USA 75, 1399- 1403 26. Gross, V. Steube, K. Tran-Thi, T.-A. Haussinger, D. Legler, G. Decker, K. Heinrich, P. C. & Gerok, W. (1987) Eur. J. Biochem. 162,83 -88. 27. Fay, P. J. Chavin, S. I. Malone, J. E. Schroeder, D. Young, F. E. & Marder, V. J. (1984) Biochim. Biophys. Acta 800, 152158. 28. Weber, W. Steube, K. Gross, V. Tran-Thi, T.-A. Decker, K. Gerok, W. & Heinrich, P. C. (1985) Biochem. Biophys. Res. Commun. 126,630-635. 29. Travis, J. Owen, M. George, P. Carrell, R. Rosenberg, S. Hallewell, R. A. & Barr, P. J. (1985) J. Biol. Chem. 260,43844389. 30. Dube, S. Fisher, J. W. & Powell, J. S. (1988) J. Biol. Chem. 263, 17 516- 17521.