Control of Glycoprotein Synthesis LECTIN-RESISTANT MUTANT CONTAINING ONLY ONE OF TWO DISTINCT N- ACETYLGLUCOSAMINYLTRANSFERASE ACTIVITIES PRESENT IN WILD TYPE CHINESE HAMSTER OVARY CELLS*

where K is either -H or -[FucIGlcNAc-Asn-peptide. Wild type cells therefore exhibit at least two GlcNAc-transferase activities, GlcNAc-transferase I acting on structures with 1 or 2 terminal mannose residues, and GlcNAc-transferase II acting on structures with 1 terminal mannose residue and 1 terminal GlcNAc residue. Pha”l cells lack GlcNAc-transferase I activity but possess full GlcNAc-transferase II activity. Preliminary data suggest that both transferases effect the synthesis of GlcNAcpl-2Man linkages. The acceptors for


GlcNAc-transferases
I and II compete for common enzyme active sites in extracts of wild type cells suggesting that the two transferases may share a catalytic subunit.
A line of Chinese hamster ovary cells (Pha"' ) selected for resistance to a phytohemagglutinin from Phaseolus uulgaris has been shown to lack an N-acetylglucosaminyltransferase (GlcNAc-transferase) activity, present in wild type Chinese hamster ovary cells, which transfers GlcNAc from UDP-GlcNAc to an acceptor prepared from human IgG' glycopeptide (1); this acceptor had the structure ManculS[Manal-G]Manpl4GlcNAcpl4[Fuc-]GlcNAc-Asn-peptide (1,2). The mutant Pha"l cells were, however, found to possess significant activity (-30 to 50% of wild type cell extracts) for the transfer of GlcNAc to an apparently similar acceptor prepared by sequential glycosidase treatment of cu,-acid glycoprotein (1). Similar results have been obtained for Chinese hamster ovary cells selected for resistance to the toxin from Ricinus communis (2,31, the agglutinin from wheat germ (31, and an agglutinin from Lens culinaris (3).
It appeared likely that the oc,-acid glycoprotein preparation contained more than one acceptor for GlcNAc transfer and that wild type cells possessed at least two GlcNAc-transferase activities only one of which was deleted in Pha'+ cells. In this paper, we examine the abilities of both wild type and Pha"l cells to transfer GlcNAc to a series of glycopeptides and oligosaccharides.
The results show the presence of at least two GlcNAc-transferase activities in wild type cells, GlcNActransferase I acting on acceptors with 1 or 2 terminal mannose residues and GlcNAc-transferase II acting on acceptors with 1 terminal mannose residue and 1 terminal GlcNAc residue. GlcNAc-transferase I is deleted in Pha"l cells while GlcNActransferase II is present in both wild type and mutant cells. Some of this data has appeared in a preliminary communication (4).
Methyl-cu-n-glucopyranoside was obtained from Pfanstiehl and pronase Grade B from Calbiochem. Sialidase (51, and a mixture of P-galactosidase and /3-N-acetylglucosaminidase (Fraction DE GM2, Ref. 6) were prepared from Clostridium perfringens; the latter preparation showed no detectable sialidase activity under the conditions of glycopeptide digestion described below and was unable to hydrolyzep-nitrophenyl-a-l\i-acetyl-n-glucosaminide,p-nitrophenyl-/3-N-acetyl-n-galactosaminide, and p-nitrophenyl-ol-N-acetyl-n-galactosaminide under standard conditions of assay (6). Jack bean pgalactosidase free of P-N-acetylglucosaminidase activity was a kind gift of Dr. Y.-T. Li, Tulane University.
Human u,-acid glycoprotein (AGP) was sequentially degraded with sialidase, P-galactosidase, and P-N-acetylglucosaminidase (7) to yield the product AGP (SA, Gal, GlcNAc). concentrated, and subjected, in two batches, to gel filtration on a column of Bio-Gel P-10 (2.5 x 90 cm) eluted with water. Glycopeptide was detected as above, and the peaks were pooled, flash-evaporated, and reconstituted with water to a volume of 12 ml. This was our crude glycopeptide fraction.
High .voltage paper electrophoresis of this material at pH 6.5 and pH 9.0 revealed two major and several minor ninhydrin-positive bands. Both major bands were positive with the periodate-benzidine stain for carbohydrate (18); one glycopeptide was neutral and the other acidic. These were separated as described below.  by sequential glycosidase degradation of Preparation I as this structure was the finding that Preparation VI was cleaved by C. described below and as outlined in Fig. 2. perfringens endo-P-N-acetylglucosaminidase (7). Preparation II-An aliquot of Preparation I (21 pm011 was digested with 0.3 unit of C. perfringens /3-galactosidase and 0.3 unit of C. mrfrinpens &Wacetvlalucosaminidase in 6.0 ml of 0.05 M aotassium phosphate (pH 6.0; at 37" for 20 h under a layer of toluene. This treatment released 21 /*mol of gala&se (14) and 18 /*mol of Nacetylglucosamine (15); the incomplete release of GlcNAc, even after repeated glycosidase treatment, was observed in two separate preparations.
The digest was concentrated and purified by gel filtration on Bio-Gel P-10 (2.5 x 90 cm) eluted with water. The carbohydrate composition of this preparation (Preparation II, Table I) and the amounts of galactose and GlcNAc released during digestion indicated that the major structure in this preparation was MS ( Fig. 1 The decasaccharide (8 firno was incubated in 1.5 ml of 0.05 M potassium phosphate buffer (DH 6.0) containine 0.09 unit of C. ner-j^ringens P-galactosidase and 6.07 unit of C. pe$ingens P-N-ac&ylglucosaminidase at 37" for 18 h; a further 0.09 unit of P-galactosidase and 0.07 unit of /3-N-acetylglucosaminidase were added and incubation was continued for 2 h. The degraded oligosaccharide was purified by passage through a column of Bio-Gel P-4 (1.5 x 25 cm) &ted with water; detection on the column was by hexose analysis (13). This material was subjected to a second alvcosidase treatment bv incubation of 7 pmol in 0.5 ml of 0.05 M potassium phosphate buffer (pH 6.0) containing 0.04 unit of P-galactosidase and 0.03 unit of B-Nacetylglucosaminidase at 37" for 17 h and purified as above.'The product (3 pmol) was incubated for a third time in 1.0 ml of 0.05 M potassium phosphate buffer (pH 6.01 containing 0.1 unit of fi-galactosidase and 0.08 unit of P-N-acetylglucosaminidase at 37" for 23 h and purified as above. GlcNAc release was measured (15) after each of the three digestions and found to be 1.8, 0.3, and 0.2 residues/mol, respectively.
The only sugars present in the final oligosaccharide preparation were mannose and glucosamine in a molar ratio 3.0 to 1.3. The structure of the major oligosaccharide in the final preparation was therefore Mancul-3[Mantll-6lManpl-4GlcNAc (9).
Preparation III-An aliquot of Preparation II (10 pmol) was heated at 60" for 45 min to destroy residual glycosidase activities and was subsequently digested with 0.02 unit of sialidase in 1.0 ml of 0.1 M potassium acetate (pH 4.5) over chloroform at 37" for 24 h; a further 0.02 unit of sialidase was added and incubation continued for another 24 h. The free sialic acid content of this digest (21) was then equal to the total sialic acid content (22) indicating complete release of sialic acid. The degraded glvcopeptide was isolated bv gel filtration of the digest on BiGGel P-iO" (2.5 x 90 cm) eluted withwater. Carbohydrate analysis of this preparation (Preparation III, Both digests were subjected to gel filtration on Bio-Gel P-10 columns (2.5 x 90 cm) followed by gel filtration on Bio-Gel P-4 (1.5 x 25 cm) eluted with water. Carbohydrate analysis of the preparation using jack bean P-galactosidase (Preparation IV-l.

Product Identification
The standard N-acetylglucosaminyltransferase incubation (see above), containing wild type cell extract and either 1 mg of AGP (-SA,Gal,GlcNAc) or 0.2 pm01 of Preparation IV-1 or 0.05 pm01 of Preparation VI/O.040 ml, was scaled up 2-to lo-fold and incubated at 37" for 3 to 5 h. The incubation containing glycoprotein acceptor was dialyzed extensively at 4" against 0.1 M NaCl followed by water to remove salts and radioactive low molecular weight compounds. The resulting radioactive glycoprotein product was either hydrolyzed with 4 N HCl at 100" for 4.5 h or treated with C. perfringens /3-Nacetylglucosaminidase in 0.08 M potassium phosphate buffer (pH 6.0) at 37" for 24 h under toluene; these digests were analyzed by high voltage paper electrophoresis (pH 3.6) and descending paper chromatography in butanollpyridineiwatr (45:25:40) for release of radioactive glucosamine and N-acetylglucosamine, respectively. The incubations containing glycopeptide acceptors were subjected to high voltage paper electrophoresis (pH 3.6); radioactive glycopeptide product and free radioactive GlcNAc remained near the origin in this system while radioactive UDP-GlcNAc and GlcNAc-l-phosphate moved toward the anode. The origins were then washed by descending chromatography with 80% ethanol to remove any free radioactive GlcNAc that may have formed during the incubation. The radioactive glycopeptide products were eluted from the papers with water and subjected to gel filtration in water on columns of Bio-Gel P-4 (1.5 x 25 cm). A single radioactive glycopeptide peak was eluted from each of these columns (recovery was over 95%). These glycopeptides were acetylated by incubation with 1.5 pmol of acetic anhydride in 0.05 ml of 0.5% NaHCO,, at room temperature for 1 h (24). Incubations were then flash-evaporated followed by gel filtration on Bio-Gel P-4 (1.5 x 25 cm) eluted with water.
The radioactive acetylated glycopeptides were analyzed with Con AiSepharose by the procedure of Ogata et al. (25). Con AiSepharose columns (0.7 x 5 cm) were washed with 10 ml of 1% bovine serum albumin and then thoroughly equilibrated with at least 30 ml of 0.01 M Tris/HCl (pH 7.5) containing 0.1 M NaCl. Radioactive glycopeptides (2000 to 2500 cpm) were applied to these columns followed by 20 ml of 0.01 M Tris/HCl (pH 7.5) containing 0.1 M NaCl and then by 20 ml of 0.1 M methyl n-n-glucopyranoside in the same buffer. Fractions (1 ml) from the columns were analyzed by adding 0.5.ml aliquots to 10 ml of Aquasol (New England Nuclear) and counting in a liquid scintillation spectrometer.

Presence of Two GlcNAc-transferase Activities in Wild Type
Cells - Table IIA shows the results of GlcNAc-transferase assays of wild type and Pha"l cell extracts using as glycosyl acceptors Preparations VI, IV-l, IV-lb, and IV-2, prepared from IgG. Preparation VI contained primarily the structure MM ( Fig. 1) and behaved in transferase assays as previously reported (11, i.e. Pha"' cells showed no activity with this acceptor while wild type cells had an activity of 4 to 5 nmol/h/ mg of protein. In contrast, Preparations IV-l, IV-lb, and IV-2 all served as excellent glycosyl acceptors for both wild type and Pha"l cell extracts showing GlcNAc-transferase activities of 16 to 22 nmol/h/mg of protein. It appears from this data that wild type cells contain at least two GlcNAc-transferase activities, GlcNAc-transferase I acting on the structure MM and GlcNAc-transferase II acting on a structure present in Preparations IV-l, IV-lb, and IV-2. GlcNAc-transferase I is deleted in Pha"l cells while GlcNAc-transferase II is present at high levels in both wild type and Pha"l cells.

Acceptor
Specificity of GlcNAc-transferase I -GlcNActransferase I activity was identified by its presence in wild type cells and its absence in Pha"l cells. Active acceptors for this enzyme were Preparation VI (Table IIA), the trisaccharide Mancul-3Manpl-4GlcNAc (Table IID) (Table IIE). The high activity with the tri-and tetrasaccharides (Table IID) indicated that neither the amino acid nor fucose contents of MM (Fig. 1) were essential for acceptor activity. However, the trisaccharide showed a higher K,,, (4.2 mM) and a higher V,,,,,, (13 nmolihimg) than the glycopeptide MM (K,,, , 0.24 mM; V,,,:,, , 5.0 nmolihimg) indicating that structures other than the trisaccharide sequence Manotl-SManpl-4GlcNAc were capable of influencing the activity of GlcNAc-transferase I. GlcNAc-transferase I was unable to act on Preparations I and V (Table IIB) indicating that structures GS, GnGn, and GnG ( Fig. 1) were ineffective acceptors. The low activity shown with Preparations II and III (Table IIC) was probably due to the a-Man-terminal compounds MS and MG (Fig. 1). The lack of activity with Preparation IV-la (Table IIC), which appeared to contain the same structures as Preparation III (Table I) (Table I) and the nomenclature used to designate these structures. Every structure is named according to the sugar residues at the nonreducing ends of the left and right branches, respectively, of the oligosaccharide chain. S, sialic acid; G, galactose; Gn, N-acetylglucosamine; M, mannose.
since Preparation III contained material which was both adherent and nonadherent to Con A/Sepharose while Preparation IV-la contained only material nonadherent to Con A/ Sepharose (Fig. 2). It is not known whether GlcNAc-transferase I can act on MGn (Fig. 1) since all preparations enriched in this glycopeptide (Preps. IV-l, IV-lb, and IV-21 contained other o-Man-terminal glycopeptides which may have been responsible for the slightly higher activities seen with wild type cells relative to Pha"l cell extracts (Table IIA).
It is not known whether GlcNAc-transferase I can attach a GlcNAc residue to both the Man&-S-and Man&-6-termini of MM. Unbranched Mancul-GManp-terminal compounds were not available to test this point although it was shown that Mancul-6GlcNAc was not an acceptor. However, according to the data of Baenziger and Kornfeld (20), incomplete glycosylation of human immunoglobulin oligosaccharides is restricted to the Manol-6-branch and, therefore, structures MS and MG (Fig. 1) should both have had Man&-6-termini; since MS and MG appeared to be substrates for GlcNAc-transferase I (Table  IIC), it must tentatively be inferred that GlcNAc-transferase I can probably transfer GlcNAc to Mancxl-6-termini as well as to Mancul-3-termini. The situation for Mancul-2-terminal compounds was ambiguous since some compounds of this type appeared to be acceptors (Table IIE) while others were not  (legend to Table II).
Acceptor Specificit.y of GlcNAc-transferase II -The existence of GlcNAc-transferase II was first suspected because the acceptor AGP (-SA,Gal,GlcNAc) showed transferase activity with both wild type and Pha'+ cells (1) (Table IID), while after three treatments, the oligosaccharide behaved like MM (Table IID), i.e. it was inactive with mutant cell extracts. The only apparent effect of the third glycosidase treatment was the removal of 0.2 residue of GlcNAc from the nonreducing terminus. Since AGP free of sialic acid and gala&se was inactive as an acceptor (11, structure GnGn ( Fig.  1) was unlikely to be an acceptor; this has in fact been verified by showing that Preparation V (a mixture of GnGn and GnG) was inactive with both wild type and Pha"l cell extracts (Table IIB) Pha"' cell extracts were inactive with Preparations VI (Table  IIA), I and V (Table IIB), and II, III, and IV-la ( ases therefore attach GlcNAc in /3 linkage to terminal cymannosyl residues. The nature of the linkages synthesized has been investigated with Con A/Sepharose chromatography as described by Ogata et al. (25). These workers suggested that oligosaccharides and glycopeptides adsorbed to Con AiSepharose columns if they contained at least 2 o-mannosyl residues that were either at the nonreducing terminus or substituted in their carbon 2 positions; any other substitutions would prevent binding to Con A/Sepharose. Fig. 3 and 4 show the results of Con A/Sepharose analysis of the radioactive glycopeptide products of GlcNAc-transferase I and II, respectively. Both transferases produced products which adhered almost completely to Con A/Sepharose; however, while methyl a-glucopyranoside eluted the product of GlcNAc-transferase I as a sharp peak (Fig. 3), the product of GlcNAc-transferase II adhered strongly to the column and was eluted with relative difficulty by the glucoside (Fig. 4). These data suggested that both transferases catalyzed the synthesis of GlcNAcpl-2Man linkages. GS (Fig. 1) in which the sialic acid-containing arm was attached to Mancul-3-terminus; since we have not done this type of structural analysis on our Preparation I and since bovine IgG has been shown (26, 27) to carry an incompletely glycosylated oligosaccharide arm on the Manal-3-branch rather than on the Manotl-6-branch claimed for human immunoglobulins (19,201, the precise nature of the acceptor for GlcNAc-transferase II has not as yet been established. Further, it is not known whether GlcNAc-transferase II is specific for either the Man&-3-or Mancul-6-termini of branched oligosaccharides or whether it can act on either arm; it is to be noted that GlcNActransferase II cannot act on unbranched Manotl-3-terminal compounds (Table IID).
Competition Studies - Table III shows the results of mixed substrate experiments in which wild type cell extracts were assayed with Preparations VI and IV-2 separately and then together in the same tube. It is evident that the enzyme rates in the mixed substrate tubes were not additive as would be expected if GlcNAc-transferase I and II were different and independent enzymes. Rather, the results were compatible with the situation in which the two acceptors competed for common enzyme active sites. Although this type of experiment is not conclusive, the data suggest that GlcNAc-transferases I and II share a common catalytic subunit which is presumably present in both wild type and Pha"i cells; the possibility that a regulatory protein is deleted in the Pha'*l cells is presently under investigation.

DISCUSSION
These studies were motivated by the finding that a line of lectin-resistant (Pha"l ) Chinese hamster ovary cells lacked completely an N-acetylglucosaminyltransferase activity assayed with a glycopeptide derivative of IgG as acceptor (1,2) but showed appreciable transferase activity when the acceptor was prepared from al-acid glycoprotein (1). This apparent discrepancy has now been resolved by the demonstration that wild type cells contain two transferases, GlcNAc-transferases I and II, only one of which (GlcNAc-transferase I) is deleted in Pha'Q cells. GlcNAc-transferase I transfers GlcNAc to Mancul-3[Mancul-GlManpl-4GlcNAc-R (where -R is -H or -[Fuc]-GlcNAc-Asn-peptide) to form GlcNAc/?l-2Manal-3[ManLul-GlManpl-4GlcNAc-R.
We have not shown that GlcNAc-transferase I is specific for the Manal-3-terminus since there is low activity with some Manal-2-terminal compounds (this may be a different transferase) and with the branched structures MG and MS (Fig. 1) which probably have 20); unbranched Mancul-GMan/%terminal compounds were not available for testing as acceptors.
The basis for assigning the Man&-6terminus as the acceptor for GlcNAc in this reaction was the finding by Kornfeld's group (19,20) that human immunoglobulins yielded, on pronase digestion, the asymmetric structure was based on the specificity of Con A/Sepharose columns (25). Since the products of both transferases adhered to the Con A/Sepharose, substitution of the terminal mannose residues at carbon atoms other than carbon 2 appeared highly unlikely. The difficulty observed in eluting the product of GlcNAc-transferase II (Fig. 4) suggested that the structure GnGn ( Fig. 1) adhered more strongly to the column than the structure MGn (Fig. 1); this is in agreement with the finding of Kornfeld and Ferris (28) that GnGn was in fact a more effective competitive inhibitor of Con A binding to guinea pig erythrocytes than either MM or MGn. Table I