Mucin Biosynthesis: Purification and Characterization of a Mucin ~6N-Acetylglucosaminyltransferase*

We have purified, to apparent homogeneity, a mucin B6N-acetylglucosaminyltransferase (B6GlcNAc trans- ferase) from bovine tracheal epithelium. Golgi membranes were isolated from a 0.25 M sucrose homogenate of epithelial scrapings by discontinuous sucrose gradient centrifugation. The Golgi membranes were sol- ubilized with 1% Triton X-100 in the presence of 1 mM Gawl-3GalNAcabenzyl (Bzl) to stabilize the B6GlcNAc transferase. The solubilized enzyme was bound to a UDP-hexanolamine-Actigel-ALD Superflow affinity column equilibrated with 1 mM GalBl- 3GalNAcaBzl and 5 mM Mn2+. Elution of the enzyme with 0.5 mM UDP-GlcNAc resulted in a 133,800-fold purification with a 1.3% yield and a specific activity of 70 pmollminlmg protein. Radioiodination of the pu- rified enzyme followed by sodium dodecyl sulfate-poly-acrylamide gel electrophoresis and autoradiography revealed a single band at 69,000 Da. Kinetic analyses of the B6GlcNAc tranferase-catalyzed reaction showed an ordered sequential mechanism in which UDP-GlcNAc binds to the enzyme first

Mucin carbohydrate structures are extremely heterogeneous. The number of different oligosaccharides of respiratory tract mucins from one individual can exceed 100 (Roussel, 1985), and their sizes vary from 1 to more than 20 sugar residues per chain. Although most mucins contain the five sugars GalNAc, Gal, GlcNAc, Fuc,' and sialic acid, oligosaccharide chains without GlcNAc are limited in size to pentasaccharides, such as the sialylated oligosaccharides that exhibit blood group A activity found in porcine submaxillary mucins (Carlson, 1968). Thus, GlcNAc transferases, which synthesize these N-acetylglucosaminides, play key roles in the elongation of mucin-type oligosaccharides. Among these GlcNAc transferases, only the PGGlcNAc transferases are responsible for the synthesis of branched mucin-type oligosaccharides (Fukuda et al., 1986). In addition, recent data implicate the mucin core 2 P6GlcNAc transferase as a differentiation marker of human lymphoid cells (Fukuda, 1989). Elevation of this enzyme activity has also been associated with malignant transformation (Yousefi et al., 1991) and Wiskott-Aldrich Syndrome (Higgens et al., 1991).
Portions of this paper (including "Experimental Procedures," part of "Results," Figs. 1 and 5-11, and Tables 111 and IV) 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.

23863
This is an Open Access article under the CC BY license. Columns (4 ml) of UDP-hexanolamine Sepharose 4B were equilibrated with 25 mM Tris-HC1 (pH 7.5), 0.25 M sucrose, and 0.1% Triton X-100 with and without 1 mM Gal@1-3GalNAcaBzl. Microsomal membranes were solubilized with 2% Triton X-100 & 1 mM Galpl-3GalNAcaBz1, and 1 ml of the 100,000 X g supernatant was applied to the column. The columns were eluted with 5 mM UMP f 1 mM Gal@1-3GalNAcaBzl.

Solubilization of P6GkNAc Transferase from the Golgi Membranes
Detergents Effects-A number of detergents were examined for their effect on 06GlcNAc transferase activity as well as their ability to solubilize the enzyme from Golgi membranes. Sucrose mono fatty acid esters of palmitate, stearate, and oleate at concentrations up to 2.5% (w/v) were ineffective at solubilizing the enzyme activity from these membranes. On the other hand, by treating Golgi membranes with 1-5% sucrose laurate, n-dodecyl-fi-D-maltoside, or Triton X-100, we recovered 16-26% of the enzyme activity in the soluble form. Triton X-100 was selected as the solubilizing detergent for all subsequent studies.
Stabilization of Enzyme Activity-To improve the yield of the solubilized P6GlcNAc transferase, numerous osmolytes (Ambudkar and Maloney, 1986) and other agents were examined for their ability to stabilize the enzyme during solubilization with 1% Triton X-100. The most effective stabilizer of the enzyme activity was the synthetic acceptor substrate, Galpl-3GalNAcaBzl. At concentrations >1 mM during the membrane solubilization step, Galpl-3GalNAcaBzl increased the yield of solubilized enzyme from about 21% to approximately 47%.

Purification of B6GlcNAc Transferase
Optimizing Binding Conditions-The binding of B6GlcNAc transferase to the UDP-hexanolamine affinity columns and its recovery depended on the presence of Mn2+ and the synthetic acceptor Galfll-3GalNAcaBzl. As indicated in Table I, the acceptor did not significantly affect the distribution of the enzyme eluted from the column (void volume uersw bound). However, when the acceptor was included in the equilibration buffer, there was a 2-fold increase in the total amount of enzyme bound to and recovered from the column. Inclusion of the acceptor in the elution buffer had little effect on the recovery of enzyme activity. Fig. 2 shows the effect of varying Mn2+ concentrations on the binding of P6GlcNAc transferase to the UDP-hexanolamine column. Inclusion of 5 mM Mn2+ in the column equilibration buffer resulted in a 2-fold increase in the amount of enzyme bound to the gel. Above 5 mM Mn2+, the amount of enzyme bound to the affinity gel decreased.
Affinity Chromatography-UDP-hexanolamine was coupled to several matrices: cyanogen bromide-activated Sepharose 4B, tresyl-activated silica, and monoaldehyde-activated Effect of Mn2+ on the binding of 86GlcNAc transferase to UDP-hexanolamine Sepharose CL-4B. Golgi membranes were treated with 1% Triton X-100 in the presence of 1 mM Gal(31-3GalNAcaBzl. The Mn2+ concentration of the buffer was adjusted to 2-11 mM, and 0.2-ml aliquots were applied to 100 mg of UDPhexanolamine Sepharose CL-4B that had been equilibrated with 25 mM MOPS, 0.25 M sucrose, 0.1% Triton X-100, 0.7 mM Galfi1-BGalNAcaBzl, and Mn2+. The mixtures were incubated at 4 "C for 30 min with gentle agitation. The gel was washed three times (0.2 ml each time) with the equilibration buffer. The 86GlcNAc transferase reaction assay mixture (0.1 ml) was then added directly to the gel and the mixture incubated at 37 "C for 2 h. Products were isolated by C18 solid phase extraction. GlcNAc as the eluent. 06GlcNAc transferase was solubilized in 1% Triton X-100 containing 1 mM GalPBGalNAcaBzl and 5 mM MnC12 and applied to a column that had been equilibrated with the same solution containing 0.1% Triton X-100. Elution with 0.5 mM UDP-GlcNAc was started at fraction 16, and 0.5 M NaCl elution was started at fraction 36. agarose (Actigel-ALD Superflow). The ligand densities obtained on these matricies were 2-4, 10, and 20 pmol/ml, respectively. Recoveries of enzyme activities from these columns were 30, 15, and 80%, respectively. Because UDPhexanolamine-Actigel-ALD Superflow column yielded higher recovery of enzyme activity, it was chosen for the purification of the B6GlcNAc transferase.
Purification Protocol-Golgi membranes were solubilized (as described under "Experimental Procedures") and applied to a 5-ml UDP-hexanolamine-Actigel-ALD Superflow column that had been equilibrated with 1 mM GalB1-3GalNAcaBzl. Approximately 50% of the enzyme activity was recovered in the void volume of the column. Upon reapplication of this material to a fresh affinity column, no further binding was observed (data not shown) suggesting that the column was not overloaded. UDP-GlcNAc (0.5 mM) eluted approximately 50% of the bound P6GlcNAc transferase activity (Fig. 3). The purification after this step was 133,800-fold with a yield of 1.3% and a specific activity of 70 pmol/min/mg protein (Table  11). The remainder of the bound enzyme activity was recovered with 0.5 M NaCl (Fig. 3). The 0.5 M NaCl eluted material   . 4).
Acceptor Competition-The material eluted from the UDPhexanolamine-Actigel-ALD Superflow column with 0.5 mM UDP-GlcNAc was examined for other GlcNAc transferase activities, in particular mucin core 4 PGGlcNAc transferase, mucin core 3 j33GlcNAc transferase, I antigen (GlcNAcB3 (GlcNAcS6)Gal) 86GlcNAc transferase, and i antigen (Glc-NAcj33Gal) P3GlcNAc transferase. Only the mucin core 4 and I antigen P6GlcNAc transferase activities were detected (data not shown). The disposition of the mucin core 4 and I antigen P6GlcNAc transferase activities in relation to the mucin core 2 p6GlcNAc transferase activity on the UDP-hexanolamine-Actigel-ALD Superflow was examined. As shown in Fig. 12, all of these activities coeluted and the ratio of the enzyme activities between the two eluted peaks were essentially identical for all three acceptors, providing evidence that enzyme eluted with 0.5 M NaCl is identical to that eluted with this substrate.
To examine if the observed enzyme activities were exhibited by one enzyme or several enzymes, acceptor competition studies were carried out using the purified enzyme and pairs of the three acceptors: Galj31-BGalNAcaBzl, GlcNAc@l-BGalNAcaPNP, and GlcNAcS1-3Gal/3Me. The results are summarized in Table V. The observed velocities paralleled the theoretical values calculated for one enzyme acting on all three substrates.
Product Identification-The reaction products of j36GlcNAc transferase with the three acceptor substrates GalS1-3GalNAcaBz1, GlcNAc/31-3GalNAcaPNP, and GlcNAcBl-3GalbMe were isolated and analyzed by 'H-NMR as described under "Experimental Procedures." As shown in Table VI, the coupling constant for the newly formed P6GlcNAc structures was 8.5 Hz indicating attachment through a &linkage. Comparison of the chemical shifts and coupling constants obtained for the three products with published data for mucin core 2, Gal~l-3(GlcNAc@l-6)GalNAc-ol (Van Halbeek et al., 1982), mucin core 4, GlcNAc~1-3(GlcNAc~1-6)GalNAc-ol (Brockhausen et al., 19851, and GlcNAc@l-3(GlcNAc~1-6)Gal (Van Halbeek et al. 1982) indicate that the P6GlcNAc transferase catalyzes the transfer of GlcNAc in a @l-6 linkage to the penultimate GalNAc or Gal of the acceptor substrates. This enzyme, therefore, is capable of synthesizing all these /36GlcNAc structures found in mucin-type oligosaccharides.

DISCUSSION
Mucin oligosaccharide chains are assembled in a sequential manner, catalyzed by glycosyltransferases; the product of one enzyme reaction serves as the acceptor for subsequent reaction. The length and complexity of mucin oligosaccharide chains can be regulated by the relative activities of the chainelongation and -termination glycosyltransferases, both of which act on the same acceptor sites (Wingert and Cheng, 1984). As demonstrated by Piller et al. (1988), an increase in B6GlcNAc transferase activity and a decrease in a2,6NeuAc transferase activity, which accompany activation of T-lymphocytes, shift the major sialylated oligosaccharides in a cell surface glycoprotein from tetra-to hexasaccharides. Elevation of P6GlcNAc transferase activity also has been reported in  , . , leukemic cells (Brockhausen et al., 1991;Yousefi et al., 1991) and in T-cells and platelets of patients with Wiskott-Aldrich Syndrome who are at risk for malignant transformation of hematopoietic cells (Higgins et al., 1991). In addition, this enzyme activity is enriched in tracheal goblet Therefore, this 06GlcNAc transferase may play important roles in the normal functions of lymphocytes and their metastatic processes as well as the differentiation of airways secretory cells.
A number of protein isolation procedures (ie. ion exchange P. W. Cheng, unpublished observation. chromatography, chromatofocusing, isoelectric focusing, dye affinity chromatography, immobilized metal affinity chromatography, phase separation, precipitation, lectin affinity chromatography) were examined as potentially useful steps in the purification of P6GlcNAc transferase. All of these procedures resulted in extensive losses of enzyme activity and little or no enzyme purification. Substrate-and product-based affinity columns were also tested for their purification potential. Affinity columns based on the acceptor substrate were prepared by coupling freezing point depression glycoprotein before and after elastase treatment to cyanogen bromideactivated Sepharose 4B or by immobilizing Galpl-3Gal-NAcaBzl on C18 silica. These methods were ineffective in binding the enzyme, even in the presence of UDP. Affinity columns based on the donor substrate, UDP-GlcNAc, and its analogs were effective in the binding and purification of B6GlcNAc transferase. These affinity columns included Hg-UDP-GlcNAc and Hg-UTP thiopropyl-Sepharose, UDP-GlcN Actigel-ALD Superflow, and UDP-hexanolamine type affinity resins. Although all of these affinity columns work equally well, the instability of mercuriated nucleotides and the cost of synthesizing UDP-GlcN (Ropp and Cheng, 1990) have limited the usefulness of these ligands. Our purification of 86GlcNAc transferase therefore was based on the use of UDP-hexanolamine affinity gels.
We devised a rapid two-step purification procedure that resulted in a 133,800-fold purification of the p6GlcNAc transferase with a specific activity of 70 pmol/min/mg protein and an overall yield of 1.3%. After radioiodination, SDS-PAGE, and autoradiographic analysis of the 0.5 mM UDP-GlcNAc, eluted material showed a single band at approximately 69,000 Da. Key factors in the purification of the P6GlcNAc transferase include 1) the stabilization of enzyme activity with GalP1-3GalNAcaBzl during Triton X-100 solubilization and subsequent binding to the affinity column and 2) the elution of the enzyme from the affinity column with 0.5 mM UDP-GlcNAc. However, the instability of the enzyme has thwarted our attempts to scale up the purification procedure and made it necessary to purify the enzyme on the day of study.
Instability of the P6GlcNAc transferase has also been noted in other tissues (Brockhausen et al., 1986;Koenderman et al., 1987;Sekine et al., 1990) but not for enzyme prepared from porcine tracheal epithelium (Sangadala et al., 1991). The reported stability of the porcine enzyme, which has been partially purified to a final specific activity of 2.4 pmol/min/ mg protein, allowed for the use of conventional chromatography and other time-consuming steps that were not possible with the bovine tracheal enzyme. There are several other notable differences in the properties of the porcine and bovine 66GlcNAc transferases. For example, the porcine but not the bovine enzyme binds to acceptor-type affinity columns. In addition, the estimated size of the porcine enzyme, 60,000 Da, is somewhat smaller than the bovine enzyme. We have no explanation for these apparent differences in enzyme properties.
Kinetic analyses of the enzyme revealed that the enzymecatalyzed reaction proceeds by an ordered sequential mechanism where UDP-GlcNAc binds first and UDP leaves last. This type of mechanism has been described for GlcNAc transferases I and I1 that are involved in the synthesis of Nlinked oligosaccharides (Nishikawa et al., 1988;Bendiak and Schachter, 1987). The Michaelis constants of the purified enzyme for UDP-GlcNAc and Galpl-3GalNAccuBzl are 0.36 and 0.14 mM, respectively. A decrease in the K , for UDP-GlcNAc was observed with increasing purity of the enzyme. This may have been due to the removal of enzymes that for Galpl-3GalNAcaBzl remained virtually unchanged throughout the purification. Although the concentration of UDP-GlcNAc in tracheal epithelial cells has not been determined, its concentration in small intestinal cells is approximately 0.5 mM (Zhivkov et al., 1975). The intracellular concentration of UDP-HexNAc in confluent cultures of human colon cancer cells varies from 0.15 mM in differentiated cells to 5.85 mM in undifferentiated cells (Wice et al., 1985). Because the K,,, of bovine tracheal P6GlcNAc transferase for UDP-GlcNAc falls in this concentration range, activity of this transferase is likely to be regulated by altering the physiological concentration of UDP-GlcNAc. Substrate competition analyses with the purified enzyme revealed that the p6GlcNAc transferase can utilize three different mucin-type acceptors: Galpl-BGalNAcaR, GlcNA-c@l-3GalNAcaR, and GlcNAcpl-3GalpR. However, it appears that the core 1 structure is a preferred acceptor ( K , = 0.07-0.14 mM) when compared to core 3 ( K , = 0.26 mM) and i ( K , = 0.53 mM) structures. Proton NMR analyses of the three products confirmed that they all contained a GlcNAc residue linked pl-6 to the penultimate GalNAc or Gal, demonstrating that they are core 2, core 4, and I antigen structures, respectively. This relaxed substrate specificity has also been observed with the crude enzyme from pig gastric mucosa (Brockhausen et al., 1985(Brockhausen et al., , 1986, Novikoff ascites tumor cells (Koenderman et al., 1987), and human ovarian tissue (Yazawa et al., 1986). Although the P6GlcNAc transferase described in this report can form all the P6-branched GlcNAc structures found in mucin-type oligosaccharides, our observation does not preclude the possibility that other B6GlcNAc transferases also are involved in the formation of p6-linked GlcNAc structures. Brockhausen et al. (1991) has reported that leukocytes from normal and leukemic individuals contain the core 2 P6GlcNAc transferase activity but no core 4 or I antigen P6GlcNAc transferase activity, suggesting that the PGGlcNAc transferase from this tissue is specific for the core 1 acceptor. The observed acceptor substrate differences between the PGGlcNAc transferase from leukocytes and those from the other tissues may be due to differential expression of isomeric forms of the enzyme. Employment of highly purified 06GlcNAc transferase from leukocytes may be needed to clarify the apparent differences in acceptor specificity.
The porcine enzyme was also examined for its ability to form the I antigen structure, but no activity was observed. It is surprising that the two acceptors, GlcNAcPl-3Galpl-3GlcNAc/3l-3Gal(Gal~l-4GlcNAc~1-6)GalNAcol and Glc-NAc~1-3Gal~1-3(Gal~l-4GlcNAc/3l-6)GalNAcol, both of which contain the GlcNAcpl-3Gal@-R acceptor structure were poor substrates for this enzyme (Sangadala et al., 1991). It is possible that the presence of the Dl-linked Gal on the P6-linked branch is responsible for changing the acceptor potential, because acceptors without this Gal residue can serve as substrates for P6GlcNAc transferase from pig gastric mucosa (Brockhausen et d., 1986) and Novikoff ascites tumor cells (Koenderman et al., 1987). These results suggest that the P6GlcNAc transferase must act on the p3-linked Gal prior to the formation of P4Gal structure linked to the PGGlcNAc. There have been several reports of an enzyme capable of transferring GlcNAc in a pl-6 linkage to terminal galactose residues (Zielenski and Koscielak, 1983a, 198313;Van den Eijnden et al., 1983;Basu and Basu, 1984;Yazawa et al., 1986;Koenderman et al., 1987). This enzyme activity was not detected in any of our preparations. There have also been reports of a transferase that transfers GlcNAc in a Pl-6 linkage to an internal galactose residue of a polylactosamine chain (Leppanen et al., 1989;Taniguchi et al., 1991). Brockhausen et al. (1986) has shown, with crude P6GlcNAc transferase, that when GlcNAc/3l-3Gal~l-3GalNAcaBzl was used as an acceptor the GlcNAc was transferred to the penultimate Gal, not to the internal GalNAc. We do not yet know if the bovine PGGlcNAc transferase has these activities. Studies are needed to further define the acceptor specificity of this enzyme and the roles of this and other PGGlcNAc transferases in the regulation of mucin-type oligosaccharide structures.