Control of Glycoprotein Synthesis BOVINE COLOSTRUM UDP-N-ACETYLGLUCOSAMINE:~-D-MANNOSIDE P2-N-ACETYLGLUCOSAMINYLTRANSFERASE I. SEPARATION FROM UDP-N-ACETYLGLUCOSAMINE:U-D-SPECIFICITY * 11, PARTIAL PURIFICATION, AND SUBSTRATE

Previous work with a lectin-resistant mutant Chinese hamster ovary cell line indicated that the elongation of N-glycosyl oligosaccharides of the “complex” or “N-acetyllactosamine” type requires two distinct UDP-N-acetylglucosamine (G1cNAc):a-D-mannoside p2-N- acetylglucosaminyltransferases that have been termed GlcNAc-transferases I and 11. Direct evidence for the existence of these two enzymes was obtained when chromatography of a 30 to 50% ammonium sulfate cut from bovine colostrum on a CM-Sephadex column resulted in the preparation of each of these enzymes in a form free of the other. At?€inity chromatography on a UDP-hexanolamine-agarose column yielded 11,000-fold purified GlcNAc-transferase I (33% recovery). GlcNAc-transferase I1 was purified 4-fold with a 66% recovery. GlcNAc-transferase was shown to catalyze

where R1 is defined below and RZ is either GlcNAcPl "+ 4(Fucoseal-+ 6)GlcNAcAsn-peptide or GlcNAcPl -+ 4GlcNAcAsn-peptide, although compounds in which Rz is GlcNAc have also been shown to be effective substrates. Reaction 1 has been established when R1 is a mannosyl residue (K, = 0.20 n m ) or Manal -+ 6(Manal "+ 3)Man (K, = 0.12 nm). GlcNAc-transferase I has also been shown to act on substrates in which R1 is a H (K, = 7.4 nm) or Manal -+ 3Man (K, not determined), but definitive identification of the products formed was not carried out. Purified GlcNAc-transferase I also catalyzed the following reaction: UDP-GlcNAc + GlcNAcbl+ ZManal + 3(Manal+ 6)(Ra/31 + 4)Man/31+ 4Rz " -+ GlcNAc/31+ tManal+ 3(GlcNAc/31+ ZManal + 6) (Rs/31 + 4)MaMI + 4Rz + UDP (2) where Rz is GlcNAcPl-+ 4(Fucoseal-+ 6)GlcNAcAsn-* This investigation was supported by the Medical Research Council of Canada. This is Paper IV in a series on Control of Glycoprotein Synthesis; Paper I11 is Ref. 1. Nuclear magnetic resonance spectra were carried out at Brookhaven National Laboratories under the auspices of the United States Department of Energy. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. peptide and R3 is H (K, = 10 mM) but not GlcNAc. These findings indicate that GlcNAc-transferase I prefers to act on the Manal -+ 3ManP terminus but can act on the Manal + 6ManP terminus with an appreciably higher K , if the Manal -+ 3Manp terminus is not available. Purified GlcNAc-transferase I did not act on Manal -+ %Maria, Manal+ 3Mana, or Manal -+ 6Mana termini. GlcNAc-transferase I1 free of GlcNAc-transferase I catalyzed Reaction 2 when RS was H (K, = 0.1 mM) but not when RI was GlcNAc, indicating a possible control function for GlcNAc at R3 in the elongation of complex oligosaccharides.
GlcNAc-transferase I1 catalyzes the following reaction: M a n '

Purification of /32-N-AcetyZglucosaminyltransferase I
It was suggested (2) that GlcNAc-transferases I and I1 are involved in the elongation of asparagine-linked oligosaccharides of the "complex" or "N-acetyllactosamine" type (3). GlcNAc-transferase I was shown by our group to act on a variety of a-mannosyl terminal glycopeptides (2) but subsequent work by others on PhaRI cells and on another similar lectin-resistant cell line (4)(5)(6)(7) indicated that the physiological substrate for GlcNAc-transferase I is probably a (Man)s(GlcNAc)zAsn-X structure. Direct evidence has recently been obtained by Tabas and Kornfeld (8) and in our laboratory (9,lO) that GlcNAc-transferase I must incorporate a GlcNAc residue into (Man)S(GlcNAc)2Asn-X before 2 mannose residues can be released to form GlcNAc-(Man)3(GlcNAc)sAsn-X, the substrate for GlcNAc-transferase 11, as shown in Equation 2. These reactions are discussed in more detail in the accompanying paper (10).
The following report provides direct evidence for the existence of the two transferases; the enzymes have been separated from a common source, bovine colostrum, so that each enzyme preparation is free of the other activity. GlcNAc-transferase I has been purified 11,500-fold but is not completely homogeneous. The recent availability of glycopeptides of known structure and high purity, primarily through the use of preparative high voltage paper electrophoresis in borate as described in the preceding paper (l), has enabled us to carry out extensive substrate specificity studies on both transferases I and 11. These studies have revealed that relatively minor changes in oligosaccharide structure can prevent glycosyltransferase action; such phenomena act as control points which channel glycoprotein assembly along certain well defined pathways. Preliminary accounts of this work have appeared (11,12).

MATERIALS AND METHODS'
Separation of GlcNAc-transferases I and 11-All steps were carried out at 4°C. Bovine colostrum (277 g) was defatted by centrifugation a t 15,000 X gmaX for 20 min, diluted with 3 volumes of cold distilled water, and fractionated by addition of solid ammonium sulfate with constant stirring. Ammonium sulfate precipitates were collected by centrifugation a t 15,000 X gmax for 20 min. The pellet formed between 30 and 50% saturated ammonium sulfate was dissolved in 300 ml of cold 0.1 mM eaminocaproic acid and dialyzed extensively against Buffer C. A fine precipitate formed which was removed by centrifugation as above and the supernatant was applied to a column (3 X 57 cm) of CM-Sephadex C-50 pre-equilibrated in Buffer C. The column was eluted successively with 1,500 ml of buffer C and 600 to 1,200 ml of 0.10, 0.15, and 0.30 M NaCl in Buffer C. Fractions (15 m l ) were collected and monitored for absorbance a t 280 nm and for GlcNAc-transferase activity using as glycosyl acceptor 1 mg of a, acid glycoprotein which had been pretreated with mild acid, P-galactosidase, and 13-N-acetylglucosaminidase as described under "Materials";2 this acceptor detects both GlcNAc-transferases I and I1 (2). Peaks of transferase activity were detected after elution with 0.1 and 0.3 M NaCl (Fig. 1); these peaks were pooled, dialyzed with three changes against Buffer A, and assayed for GlcNAc-transferase I using IgG glycopeptide3 MM (90 nmol/assay), for GlcNAc-transferase I1 using IgG glycopeptide2 MGn (100 nmol/assay), and for galactosyltransferase using GlcNAc as acceptor (20).  (1,2) and these structures are shown in Fig. 2 of the preceding paper (1).
Purification of GlcNAc-transferase I-The GlcNAc-transferase preparation eluted from CM-Sephadex by 0.3 M NaCl in Buffer C contained about 35% of the original GlcNAc-transferase I activity and was used for further purification of this enzyme. From 70 to 95 ml of the dialyzed 0.3 M NaCl eluate was applied to a column (1.3 X 5.0 cm) of UDP-hexanolamine-agarose pre-equilibrated with Buffer A. A flow rate of 1 ml/min was maintained and fractions of 5 ml were collected. The column was washed with Buffer B until the absorbance at 280 nm was about 0.07 (Fig. 2); attempts to obtain more highly purified enzyme by more extensive washing of the UDP-hexanolamine-agarose column led to complete loss of enzyme activity. The column was eluted with 30 ml of 3 mM UDP-GlcNAc in Buffer B and fractions were collected and assayed for GlcNAc-transferase activity using 1 mg of glycoprotein acceptor/assay (Fig. 2). Fractions containing enzyme activity were pooled, and heat-treated (65°C for 15 min) bovine serum albumin was immediately added to a final concentration of 1 mg/ml. This preparation was stored a t 4°C and was used in the studies on GlcNAc-transferase I reported in this paper. Enzyme was dialyzed against Buffer B at 4°C to remove UDP-GlcNAc prior to kinetic study. Fig. 1 shows a typical elution profile on CM-Sephadex for the 30 to 50% ammonium sulfate cut from bovine colostrum. Four protein peaks were obtained. GlcNAc-transferase activity as determined with glycoprotein acceptor was detected primarily in the second (0.10 M NaCl eluate) and fourth (0.30 M NaCl eluate) peaks. Table I shows that GlcNAc-transferase I appears only in the fourth peak; this peak also shows some activity with glycopeptide MGn, but most of the galactosyltransferase activity is eluted in earlier CM-Sephadex peaks ( Table I). The activity towards glycopeptide MGn was retained even by highly purified preparations of GlcNAc-transferase I and it will be shown below that this is due to a low activity of this enzyme towards MGn and not to contamination with GlcNAc-transferase 11. GlcNAc-transferase I1 was most enriched in the second peak eluted from CM-Sephadex (Table I); this preparation also has appreciable galactosyltransferase activity but lacks GlcNAc-transferase I. GlcNActransferase I1 does not adhere to UDP-hexanolamine-agarose under conditions which cause galactosyltransferase to adhere tightly to this affinity adsorbent (16); we have therefore been Fractions (15 ml) were collected and assayed for absorbance a t 280 nm (---). Aliquots (0.010 ml) were assayed for GlcNAc-transferase activity using as acceptor 1 mg of Q, acid glycoprotein previously treated with mild acid to remove sialic acid and with /3-galactosidase and P-N-acetylglucosaminidase (expressed as counts per min/h/0.010 ml (-).

Separation of GkNAc-transferases I and 11-
Fractions pooled for later analysis (Table I)  able to obtain GlcNAc-transferase I1 free of both GlcNActransferase I and galactosyltransferase by passage of the second peak off CM-Sephadex through a column of UDP-hexanolamine-agarose.
Purification of GlcNAc-transferase Z-Passage of the fourth CM-Sephadex peak through a column of UDP-hexanolamine-agarose results in the adsorption of all the GlcNActransferase I activity to the affinity column (Fig. 2). Washing of the column with Buffer B results in removal of a great deal of inactive protein (Fig. 2) and this step alone results in about 68-fold purification (Table 11) of GlcNAc-transferase I. I t was found, however, that extensive washing of the affinity column (Table 111) resulted in poor recovery of GlcNAc-transferase I; washing is therefore stopped as soon as the absorbance of the eluate at 280 nm is 0.07. Under these conditions, recovery of enzyme from the column is nearly quantitative. In one experiment, Buffer B was made 25% in glycerol and used to wash a UDP-hexanolamine-agarose column loaded with GlcNActransferase I until the absorbance at 280 nm of the eluate was 0.002; the recovery of GlcNAc-transferase I from this column was negligible. It therefore appears that the enzyme is unstable at dilute protein concentrations and this has prevented purification to complete homogeneity. Table I1 summarizes a typical GlcNAc-transferase I purification. The final enzyme preparation was obtained in 31 to 33% yield with an enrichment of about 11,OOO. The procedure has been carried out several times and is highly reproducible. In later preparations, the procedure was shortened somewhat by omission of the 0.1 M NaCl elution step; this leads to a less pure preparation of GlcNAc-transferase I1 but does not alter the GlcNAc-transferase I preparation. The CM-Sephadex step can also be accelerated by mixing the crude enzyme solution with CM-Sephadex and pouring the slurry on a column of CM-Sephadex.
The affinity column-purified GlcNAc-transferase I is very unstable at low protein concentrations. In the absence of bovine serum albumin, the enzyme activity decayed a t a rate of 50% per day at 4OC; addition of albumin, as described under "Materials and Methods," prevented any loss of activity on storage at 4°C for at least 1 month.
Polyacrylamide gel electrophoresis in 7.5% sodium dodecyl sulfate in Tris/glycine a t p H 8.8 (23) of the affinity columnpurified GlcNAc-transferase I showed two major bands after staining with Coomassie blue, indicating that the enzyme preparation is not homogeneous. However, the availability for the fist time of preparations of GlcNAc-transferases I and I1 from a common source such that each preparation is free of the other activity prompted us to carry out a detailed substrate specificity study.
The Product of GlcNAc-transferase I Action on ZgG Glycopeptide "" Fig.  3a (upperpanel) shows the radioactivity scan of a high voltage electrophoretogram in borate buffer of the reaction product obtained by the action of affinity columnpurified GlcNAc-transferase I on highly purified IgG glycopeptide MM. The product migrates to the same extent as a sample of authentic IgG glycopeptide MGn. The importance of using highly purified glycopeptide preparations as substrates in studying glycosyltransferase specificities is indicated by the scan shown in Fig. 3b. The glycopeptide used as substrate in this experiment was mainly MGn with only 2 to 3% contamination with glycopeptide MM. Appreciable amounts of MGn are nevertheless formed from MM since, as will be shown below, the conversion of MM to MGn by GlcNAc-transferase I has a much lower K,,, than the conversion of MGn to GnGn by this enzyme. High resolution proton NMR spectroscopy of a relatively large scale preparation of the product formed from MM showed a spectrum identical to that of IgG glycopeptide MGn (Table IV) (Table IV) shows a n extra terminal GlcNAc residue in the product (H-1 signal a t 4.588 ppm and N-acetyl signal at 2.052 ppm); the spin-spin coupling constant J was found to be 8.2 Hz for the H-1 signal from this terminal GlcNAc residue, indicating a P linkage. Evidence that this GlcNAc residue is linked to the a3-linked mannose residue rather than the a6linked mannose residue is the shift of the H-1 signal of the a3-linked mannose from 5.128 ppm in MM to 5.141 ppm in the product (Table IV). There is also a relatively large shift of the H-2 signal of the a3-linked mannose from 4.068 ppm in MM to 4.188 ppm in the product (Table IV), indicating a substitution at C-2 of this mannose residue. The H-1 and H-2 signals of the other 2 mannose residues are identical in MM and in product (Table IV). The /? linkage of the GlcNAc incorporated into MM by GlcNAc-transferase I had previously been established by the use of P-N-acetylglucosaminidase (2). We have also previously shown the GlcNAc-Man linkage in the transferase product to be 1 to 2 by observing that the product adhered to concanavalin A-Sepharose (2,211. Radioactive product obtained by incubating IgG glycopeptide MM and UDP-['4C]GlcNAc was acetylated with nonradioactive acetic anhydride (1) and incubated with endo-P-Nacetylglucosaminidase D (22); at least 85% of the glycopeptide was resistant to cleavage by this enzyme (Table V). The radioactive glycopeptide was also resistant to cleavage by A. A. Grey, S. Narasimhan, H. Schachter, and J. P. Carver, manuscript in preparation. end0-P-N-acetylglucosaminidase CI. This suggests (22, 24) that GlcNAc incorporation was primarily on the a3-linked mannose residue of glycopeptide MM, a result in agreement with the NMR data (Table IV). The product of GlcNActransferase I action on MM therefore appears to be primarily GlcNAcPl + 2Manal+ 3(Manal+ 6)ManBl-4GlcNAcPl -+ 4(Fucal+ 6)GlcNAcAsn-X, i.e. MGn.
The Product of GlcNAc-transferase I Action on IgG Glycopeptide MGn- Fig. 3a (lower panel) shows the scan of an electrophoretogram of the reaction product formed when GlcNAc-transferase I free of GlcNAc-transferase I1 acts on glycopeptide MGn. The product is seen to migrate like standard GnGn," suggesting that both transferases I and I1 convert MGn to GnGn; however, as shown below, the K , of this reaction is much lower for transferase I1 than it is for transferase I. Because of the relatively poor K,,, of GlcNAc-transferase I for MGn, it is especially important to use highly purified MGn when measuring the K,,, (see Fig. 3b). The only other experiment carried out on this product was analysis by concanavalin A-Sepharose (Fig. 4); it is seen that the product adheres to the lectin column and is eluted as a broad peak with methyl-a-D-mannopyranoside, a pattern previously obtained with IgG glycopeptide GnGn (21) and with the product of GlcNAc-transferase I1 action on MGn (2). The product of GlcNAc-transferase I action on MGn therefore appears to be GlcNAcPl-2Manal + 3(GlcNAcP1-2 M a n a l 3 6 ) M a n p l + 4GlcNAcP1-+ 4 ( F u c a l -+ 6)GlcNAcAsn-X, i.e. GnGn.
The Products of GlcNAc-transferase I Action on Oualbumin Glycopeptides-The most extensive product characterization was carried out with the product formed by the action of affinity column-purified GlcNAc-transferase I on highly purified ovalbumin glycopeptide V." Figure 5 (lower panel) shows a radioactivity scan of a high voltage electrophoretogram in borate of this product; the peak is seen to be sharp, indicating homogeneity. It electrophoreses at the same position as ovalbumin glycopeptide 111-A," i.e. GlcNAc-(Man)5(GlcNAc)BAsn, a glycopeptide which differs from the structure suggested below for our product by having an additional GlcNAc residue linked pl-4 to the P-linked mannose residue; we have shown that the presence of this GlcNAc residue does not influence appreciably the mobility of a glycopeptide on borate electrophoresis (1). The product migrates between ovalbumin glycopeptides 111-A and V on gel filtration on Bio-Gel P-6 (10). The high resolution proton NMR spectra5 of glycopeptide V and product are compared in Table IV. The spectra are very similar except for the presence of an additional P-linked GlcNAc residue in the product (H-1 signal a t 4.591 ppm and N-acetyl signal at 2.055 ppm); the ,Ll linkage was deduced from the large coupling constant ( J = 8 Hz) for the H-1 signal. Comparison of the spectra also shows shifts in both the H-1 and H-2 signals from one of the 2 a3-linked mannose residues (Table IV) indicating that GlcNAc is attached to C-2 of one of these residues. NMR spectroscopy does not permit us to determine which mannose residue is substituted.5 Substitution of the Manal -+ 3Manpl -+ 4GlcNAc terminus of glycopeptide V is shown by experiments on the processing of this glycopeptide described in the following paper (10); further proof that it is the Manal -+ 3Manpl -+ 4GlcNAc terminus of glycopeptide V which is substituted with a GlcNAc residue derives from the fact that the product is almost completely resistant to the action of endo-P-Nacetylglucosaminidase D under conditions a t which over 90% of glycopeptide V is cleaved (Table V). About 150 nmol of product were subjected to methylation analysis as previously described (1); the small amount of material available made ' J. P. Carver, A. A. Grey, C. Ceccarini, and P. Atkinson, manuscript in preparation.
quantitation unreliable and only qualitative results are reported in Table VI. The presence of a large peak of 3,4,6-tri-

O-methyl-2,5-di-O-acetyl-hexitol identified by its electron im-
pact mass spectral fragmentation pattern proves that the product has a GlcNAc linked 1 + 2 to a mannose residue; a similar conclusion was suggested by the shift seen in the C-2 hydrogen of the a3-linked mannose residue on NMR spectroscopy. Permethylation analysis of the product (Table VI) also indicates the presence of the 2,3,4,6-tetra-O-methyl and 2,4-di-O-methyl derivatives of partially acetylated hexitol and the 3,4,6-tri-O-methyl and 3,6-di-O-methyl derivatives of partially acetylated N-acetylhexosaminitol consistent with the structure proposed below for the product. However, we also detected a peak with a gas chromatographic retention time and fragmentation pattern consistent with the 2,3,6-tri-Omethyl derivative of partially acetylated hexitol, suggesting that some GlcNAc may have been incorporated in a 1 + 4 linkage to mannose. The presence of a small amount of such a product cannot be ruled out. However, the processing experiments reported in the following paper (10) indicate that most of the product can be converted by the removal of 2 mannose residues to a structure, GlcNAc(Man),-(GlcNAc)nAsn, which adheres strongly to concanavalin-Sepharose, indicating that the terminal GlcNAc residue is linked primarily in a 1 + 2 linkage to mannose (21); we are presently investigating the possibility that our GlcNAc-transferase I preparation may contain some P4-N-acetylglucosaminyltransferase, an enzyme not yet conclusively demonstrated in the literature. The product adheres strongly to concanavalin A-Sepharose and elutes sharply with methyl-a-D-mannopyranoside (10). Endoglycosidase digestion experiments have shown that the product is resistant not only to endo-p-N-acetylglucosaminidase D (Table V) but also to endoglycosidase CI; it is completely susceptible to endo-P-N-acetylglucosaminidase Cn (lo), suggesting that there is little if any GlcNAc linked 1 +

to the Manal
-+ 3ManP terminus (25). The structure proposed for the product is M a n . al+6,

'ManB1+4GlcNAcgl+4GlcNAcAsn
GlcNAcBl+ZMan' Fig. 6 (upper scan) shows an electrophoretogram of [''HIacetylated ovalbumin glycopeptide VI (1). Following incubation with GlcNAc-transferase I and nonradioactive UDP-GlcNAc under standard incubation conditions, there is almost complete conversion to a product with a slower mobility towards the anode on high voltage electrophoresis (lower scan, Fig. 6). This experiment suggests that glycopeptide VI is also a substrate for GlcNAc-transferase I but the only product characterization that has been carried out is analysis on concanavalin A-Sepharose; the product does not adhere to the lectin column but elutes in a broad peak on washing the column with buffer in the absence of methyl-a-D-mannopyranoside (10). The significance of this observation is discussed in the following paper (10).
Kinetic Parameters for GlcNAc-transferase I-Extensive studies on the variation of enzyme velocity with acceptor concentration have been carried out for ovalbumin glycopeptide V, for IgG glycopeptides MM, MGn, and MGn(Gn), and for the trisaccharide Manal + 3Manbl + 4GlcNAc, using the affinity column-purified GlcNAc-transferase I. All five acceptors gave excellent Michaelis-Menten kinetics as indicated by linear l / v uersus l/s plots (data not shown). The kinetic parameters obtained from these plots are summarized in Table VII. It is clear that ovalbumin glycopeptide V is the best substrate, consistent with its being the physiological substrate (10). However, IgG glycopeptide MM is also a good substrate and may under certain situations serve as a substrate in vivo (26). As mentioned earlier, MGn is a rather poor substrate, having a much higher K, than MM; thus, a minor contamination of MGn with MM can lead to serious errors (Fig. 3b). Since the K , of MGn with GlcNAc-transferase I1 is about 0.1 mM (Fig. 7), a K , of 10 mM for this substrate with GlcNAc-transferase I (Table VII) indicates that the latter enzyme is not contaminated with transferase I1 but has a low activity towards MGn. It is interesting that MGn(Gn), which differs from MGn only by having an additional GlcNAc linked pl + 4 to the P-linked mannose residue, is not a substrate for GlcNAc-transferase I; the low activity seen with MGn(Gn) ( Table VI1 and Fig. 7) has recently been shown to be due to a previously undetected trace contamination of MGn(Gn) with As previously determined with Chinese hamster ovary cell extracts (2), ovalbumin glycopeptides I11 (containing a mixture of 111-A, 111-B, and 111-C; see preceding paper, Ref. l), and IV were not acceptors for affinity column-purified GlcNAc-transferase I, i.e. the Manal + 2Mana terminus does not act as an acceptor, nor do Manal + 6Mana or Manal + 3Mana termini act as acceptors; the penultimate mannose must be P-linked for acceptor activity.
The variation of enzyme velocity of affinity column-purified GlcNAc-transferase I with UDP-GlcNAc concentration using a , acid glycoprotein derivative as acceptor also showed excellent Michaelis-Menten kinetics; the K , for UDP-GlcNAc is 0.1 mM. Kinetic experiments previously carried out only with crude enzyme preparations were repeated on the affinity column-purified GlcNAc-transferase I; it was shown that the pH optimum was 6.0, that the enzyme was inhibited by 3 mM EDTA, and that the optimum Mn2+ concentration ranged between 18 and 26 mM.
Studies on GlcNAc-transferase ZI-Preliminary studies were carried out on the 0.1 M NaCl eluate from the CM-Sephadex column. Although this fraction was only about 4fold enriched in GlcNAc-transferase I1 and contained appreciable amounts of galactosyltransferase activity (Table I) the implications of this inhibitory residue are discussed in the following paper (10). The low activity shown by GlcNActransferase I towards MGn(Gn) ( Table VI1 and Fig. 7 ) is due to a trace contamination of MGn(Gn) with MM.6

DISCUSSION
Genetic evidence based primarily on the study of a lectinresistant mutant line of Chinese hamster ovary cells (2) fiist indicated that the incorporation of peripheral GlcNAc residues into the (Man)a(GlcNAc)z core of asparagine-linked oligosaccharides of the complex type required the action of at least two different N-acetylglucosaminyltransferases catalyzing the reactions shown in Equations 1 and 2. To prove the existence of these two enzymes, we set about attempting separation of the activities from a common source. We chose bovine colostrum because previous work on N-acetyl-S. Narasimhan and H. Schachter, unpublished data.
glucosaminyltransferase (27) and other glycosyltransferases (28, 29) had shown colostrum to be a convenient source of soluble transferase. It was felt that colostrum might provide an enzyme source which did not require detergent for purification, unlike intracellular transferases, which usually require detergent for solubilization prior to purification (30-33) with subsequent problems in removal of detergent. The application of ion exchange chromatography led to a ready separation of the two transferases (Fig. 1). GlcNAc-transferase I1 did not adhere to a UDP-hexanolamine-agarose affinity column and could not be highly purified but GlcNAc-transferase I did adhere to the column and was purified about 11,000-fold; thus, the usefulness of this affinity column in the purification of glycosyltransferases (16,32,33) has been extended to yet another enzyme. The final enzyme preparation was not homogeneous. However, we feel that this is related to the instability of the enzyme at dilute protein concentrations, a property shared with other glycosyltransferases. Extensive washing of the affinity column cleared a great deal of nonspecific protein from the column but led to total loss of transferase activity (Table 111). This can probably be reversed by using much larger initial volumes of colostrum (28) so that the amount of protein in the final purified enzyme preparation is appreciably higher; this approach has worked with other transferases (28).
Although not pure, the transferase preparations were useful for kinetic studies since each preparation was free of the other enzyme activity. The identity of the product formed by the action of transferase I on ovalbumin glycopeptide V was established; this product is used in the processing experiments described in the following paper (10). Transferase I was shown to have a preference for t h e M a n a l -+ 3 terminus of MM; this is the same terminus that is available on ovalbumin glycopeptide V. However, if this terminus is not available, as for example in glycopeptide MGn, transferase I can act on the Manal "+ 6 terminus although with a much higher K , and the physiological significance of this reaction is doubtful. It appears certain that the physiological substrate for transferase I is the structure represented by ovalbumin glycopeptide V (see Ref. 10 for evidence and further discussion); however, it is interesting that in a mutant cell line (26) which cannot add mannose to the Manal + 6Manb terminus of the growing lipid-linked oligosaccharide precursor, the structure equivalent to MM can act as the physiological substrate for GlcNActransferase I.
Kinetic parameters for the various glycopeptides have been obtained and indicate that both ovalbumin glycopeptide V and IgG glycopeptide MM are excellent substrates although glycopeptide V is somewhat better. The presence of a GlcNAc linked /3l + 4 to the P-linked mannose residue appears to block the action of both transferases I and 11. This is yet another example of glycosyltransferase substrate specificity and the implications of this and other analogous findings are discussed further in the following paper (10).
The nature of the RZ group in Equations 1 and 2 does not appear to be critical. We have previously shown (2) that the oligosaccharides Manal + 3 ( M a n a l + 6 ) M a n P l 4 4GlcNAc and GlcNAcPl + 2Mana1 -+ 3(Manal -+ 6)ManPl + 4GlcNAc, in which Ri is GlcNAc, are effective substrates for GlcNAc-transferases I and 11, respectively. Also, GlcNActransferase I can act on ovalbumin glycopeptides V and VI which lack a Fucal + 6 residue on the Rz group and on IgG glycopeptide MM in which Rz is GlcNAcPl + 4(Fucal + 6)GlcNAcAsn-peptide, indicating that this fucosyl residue does not influence activity appreciably.